PCOS Glycemic Index Table: 150 Foods, With Sources

2% milk is rarely GI-tested separately; whole (27-38, middle 32) and skim (32-37) bracket it tightly, so 32 is an interpolation, not a measured 2%-specific value. Lactose-driven; low regardless of fat level.

University of Sydney GI database (glycemicindex.com) tested agave products: 'light' agave nectar (90% fructose) GI 11, 'premium' agave nectar GI 19; broader published range ~10-27 by brand and processing. 15 is the middle of the Sydney-tested pair — brand/processing is the dominant variable, this is a representative middle, not a single-product figure.

Tani Y, Tokuda M, Nishimoto N, Yokoi H, Izumori K, "Allulose for the attenuation of postprandial blood glucose levels in healthy humans: a systematic review and meta-analysis", PLOS ONE 2023;18(4):e0281150 — pooling 8 human experiments, allulose does not raise blood glucose with a single intake, and 5 g/10 g doses taken with carbohydrate reduced postprandial glucose iAUC by roughly 13-14% (SMD -0.28 and -0.26). This makes the 0 a measured absence of glycemic response rather than an assumption. Metabolism: about 80% is absorbed and excreted unchanged in urine, the remainder passes to the large intestine with negligible fermentation. FDA final guidance (Oct 2020, "The Declaration of Allulose and Calories from Allulose on Nutrition and Supplement Facts Labels") permits excluding allulose from Total Sugars and Added Sugars while still counting it under Total Carbohydrate — so the label carbohydrate figure is NOT available carbohydrate; available carbohydrate is 0 and GL is 0. Allulose is absent from the University of Sydney international GI tables (Atkinson 2008, Diabetes Care 31:2281-2283; Atkinson 2021, AJCN Suppl. Tables 1-2) because there is no glycemic response to index. Serving weight caveat: granulated allulose is close to table sugar in bulk density (~12-12.5 g/tbsp); the exact gram weight does not affect any number in this row since GI is 0 and fiber and protein are 0.

Composition: USDA FoodData Central, Foundation Foods FDC 2262074 (Almond butter, creamy, published 2022) — 21.2 g carbohydrate, 9.7 g fibre, 20.8 g protein, 53.0 g fat per 100 g; 2 tbsp = 32 g, giving 3.7 g available carbohydrate, 3.1 g fibre and 6.7 g protein. GI recorded as n/a: neither almonds nor almond butter appear in the NUTS section of Supplemental Table 1 or Supplemental Table 2 of Atkinson FS, Brand-Miller JC, Foster-Powell K, Buyken AE, Goletzke J, "International tables of glycemic index and glycemic load values 2021", Am J Clin Nutr 114:1625-1632 — only cashews, chestnuts and peanuts have been tested — and the paper states that nuts are so low in carbohydrate that most varieties cannot be tested. At 3.7 g available carbohydrate per serving this is far below the 25-50 g test portion the ISO method requires. Consistent with the existing almonds row in this table.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (the University of Sydney GI database and all major published GI tables only test foods with meaningful available carbohydrate), so GI is recorded as n/a rather than 0: it is undefined, not measured-as-zero. Confirmed absent from both Foster-Powell et al. 2002 and Atkinson et al. 2008 international tables. Composition from McCance & Widdowson's Composition of Foods Integrated Dataset (CoFID) 2021, Public Health England, 'Anchovies, canned in oil, drained': 25.2g protein, 0.0g carbohydrate, 0.0g fibre per 100g. Note: the CoFID figure is for the drained product, which is the form specified in serving_desc; salt content is high and varies substantially by brand, so copy should not state a salt figure.

USDA FoodData Central #173469; ~0.1g carb per tablespoon, GI is undefined for non-carbohydrate condiments.

University of Sydney GI database (glycemicindex.com) — standard reference apple value (commonly cited 34-40 range), among the most consistently low-GI whole fruits measured.

UNSWEETENED applesauce only. No single Sydney-tested value found; composition-based and secondary GI databases disagree (cited 35 to 53); 45 is the middle of that range. Sweetened/commercial applesauce with added sugar is NOT covered by this figure and runs meaningfully higher.

GI 34±3 — Atkinson FS, Foster-Powell K, Brand-Miller JC, "International Tables of Glycemic Index and Glycemic Load Values: 2008", Diabetes Care 2008;31(12):2281-2283, item 948 "Apricots, raw, NS" (Australia): 120 g serve, 9 g available carbohydrate, GL 3, glucose reference, n=10. The 120 g serving and the 9 g available-carbohydrate figure are taken directly from that tested portion, so the arithmetic reproduces the published GL exactly (34 x 9/100 = 3.06, rounded to 3). DISCLOSED DISCREPANCY: scaling USDA composition instead would give (11.1 - 2.0) x 1.20 = 10.9 g available carbohydrate and a GL of 4; the tested-portion figure is used because it is the basis on which the GI itself was measured. Honest range on the GI: the ISO 26642:2010-compliant 2021 update (Atkinson et al., AJCN 2021, Suppl. Table 1) lists the same Australian value of 34±3 (item 1083) alongside a Canadian 2018 value of 42±5 (item 1084); GL still rounds to 3 at GI 42, so the serving-level number is stable across the disagreement. Composition from USDA FoodData Central SR Legacy 171697 "Apricots, raw": per 100 g, 11.1 g carbohydrate, 2.0 g fiber, 1.4 g protein, 9.24 g sugars; USDA lists one apricot at 35 g, so 120 g is roughly 3-4 whole fruit. Related but distinct entries, not covered by this row: apricot halves canned in fruit juice GI 51±5 (2008 item 949) and dried apricots GI 30-32 (2008 items 950-952).

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

EFSA Panel on Food Additives, "Scientific Opinion on the re-evaluation of aspartame (E 951) as a food additive", EFSA Journal 2013;11(12):3496: aspartame is completely hydrolysed in the gut to phenylalanine, aspartic acid and methanol and is used at milligram levels, so the sweetener itself supplies no available carbohydrate. GI recorded as n/a rather than 0 because aspartame has never been GI-tested — it has no entry in Atkinson FS et al., "International tables of glycemic index and glycemic load values 2021: a systematic review", Am J Clin Nutr 2021 — so a published 0 would assert a measurement that does not exist. Per-packet aspartame content (~40 mg) is back-calculated from FDA's published statement that a 60 kg adult would need roughly 75 tabletop packets to reach the 50 mg/kg bw/day acceptable daily intake (60 x 50 = 3000 mg; 3000/75 = 40 mg). SERVING CAVEAT: a 1 g tabletop packet is mostly dextrose or maltodextrin bulking agent, contributing roughly 0.9 g carbohydrate, which is glycemically trivial but not literally zero; bulking-agent content varies by product, so copy must not describe a packet as carbohydrate-free. Aspartame supplied as a tablet or liquid has no bulking agent.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

USDA FoodData Central SR Legacy 168322 (Pork, cured, bacon, pre-sliced, cooked, pan-fried): 1.7g carbohydrate, 0g fiber, 33.92g protein, 35.09g fat per 100g. Serving from that record's USDA food_portion entry, 1 slice = 11.5g, so 2 slices = 23g. GI reported as 'n/a' rather than 0: the serving supplies ~0.4g available carbohydrate, far below the 25-50g ISO 26642:2010 test portion, and bacon appears nowhere in the 2021 International Tables (Atkinson et al.) — same convention already used in this CSV for eggs and cheese. Sugar-cured, maple and honey-cured varieties carry somewhat more carbohydrate than this generic entry; USDA also lists reduced-sodium (167712) and baked (167914) preparations with slightly different profiles.

University of Sydney GI database (glycemicindex.com) — white bagel GI 72, one of the standard high-GI reference bakery items. Serving is the 70g portion used in published GI tables (GL 25); a whole modern deli bagel is ~100g, so real-world GL is often ~35+. Whole-wheat bagels run somewhat lower but remain high-GI territory.

Atkinson FS, Foster-Powell K, Brand-Miller JC, 'International Tables of Glycemic Index and Glycemic Load Values: 2008', Diabetes Care 31(12):2281 — entry 1083, 'Baked Beans in Tomato sauce', a commercial Australian canned product, GI 40+/-5, glucose reference, n=10, 150g serving, 20g available carbohydrate, GL 8. Fibre and protein from McCance & Widdowson's Composition of Foods Integrated Dataset (CoFID) 2021, Public Health England, 'Baked beans, canned in tomato sauce' (4.9g AOAC fibre, 5.0g protein per 100g). Disputed: tested baked-bean products span GI 40-57 in the same table — canned tomato-sauce beans reheated tested at 57 (entry 1084), ham/barbecue/curry/sweet-chilli sauces at 46-53 (entries 1079-1082), plain canned Canadian and commercial vegetarian haricot products at 40 (entries 1076-1077). A University of Sydney summary page separately cites 68 for a 250mL serving, which does not match the compiled table; the peer-reviewed table value is used here. 40 is the plain tomato-sauce tested value, not a universal one. Generated copy must not name any manufacturer or product brand.

University of Sydney GI database (glycemicindex.com) — ripe banana. Unripe/green banana is substantially lower (~30); ripeness is the dominant variable.

University of Sydney GI database / international GI tables (Atkinson) — pearled barley boiled, cited consistently at 25-28; 28 used. One of the lowest-GI grains measured, driven by beta-glucan and intact starch structure. Quick/flaked barley products run higher — this figure is for boiled pearled grains.

Foster-Powell K, Holt SHA, Brand-Miller JC, Am J Clin Nutr 2002;76:5-56, entry 297 "Basmati, white, boiled": GI 58 ± 8, 150g serving, 38g available carbohydrate, GL 22 (also carried in the University of Sydney GI database, glycemicindex.com, Sydney University's Glycemic Index Research Service, n=9, 1995). AUDIT NOTE: the 38g available-carbohydrate figure used here is the source's own paired value, kept so the GI and GL come from one measurement — a USDA-derived estimate would instead give ~41.7g for 150g and GL ~24. Real spread in the Sydney database: GI 50 (Ranawana et al., Int J Food Sci Nutr 2009;60 Suppl 4:99-110), 56/60/63 across Chinese, Asian-Indian and Malay test groups (Tan et al., Br J Nutr 2015;113:1228-36), 57 (Kataoka et al., Diabet Med 2013;30:e101-7), 59 and 65 (Sydney, unpublished). Fiber and protein from USDA FoodData Central SR Legacy #168878 (rice, white, long-grain, regular, enriched, cooked) as a proxy — USDA holds no basmati-specific Foundation or SR Legacy entry.

University of Sydney GI database / published tables — boiled/canned beetroot GI 64. Raw beetroot is cited around 32 — cooking roughly doubles the GI by breaking down fibre structure, so preparation is the dominant variable. 64 represents the cooked form people actually eat; GL stays low because a serving carries little carbohydrate.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Published international GI tables (Foster-Powell/Atkinson) — boiled black beans ~30, with some studies citing as low as 20; 30 is the conservative-higher representative. Canned beans run somewhat higher than home-boiled. Fiber+protein density is what keeps legume GI low.

Composition: USDA FoodData Central, SR Legacy FDC 173759 (Cowpeas, common (blackeyes, crowder, southern), mature seeds, cooked, boiled, without salt) — 20.8 g carbohydrate, 6.5 g fibre, 7.73 g protein per 100 g; 1 cup = 171 g, giving 24.5 g available carbohydrate, 11.1 g fibre and 13.2 g protein. GI = 41 is the mean of the three boiled-cowpea entries tested in healthy adults in Atkinson FS, Brand-Miller JC, Foster-Powell K, Buyken AE, Goletzke J, "International tables of glycemic index and glycemic load values 2021", Am J Clin Nutr 114:1625-1632, Supplemental Table 2: 33 (Canada 1981, n=6), 38 (Nigeria 1989, n=10) and 52 (Nigeria 1989, n=10). GL = 41 x 24.5/100 = 10.0. DISPUTED: the spread is wide and all three entries are non-ISO (Supplemental Table 2 covers studies with method deviations from ISO 26642:2010). A fourth entry (GI 50, Canada 1983) was measured in type 2 diabetic subjects and was excluded; two Philippine 2009 entries for cowpeas boiled, cooled overnight and microwaved measured GI 7 and 11, which is a cook-chill resistant-starch effect and not a value for freshly boiled peas.

USDA FoodData Central, SR Legacy FDC 173946 (Blackberries, raw) — 9.61 g carbohydrate, 5.3 g fibre, 1.39 g protein, 4.88 g sugars, 43 kcal per 100 g; 1 cup = 144 g. That leaves 4.3 g available carbohydrate per 100 g, i.e. 6.2 g per cup. GI recorded as n/a because blackberries have never been GI-tested: they appear in neither Supplemental Table 1 nor Supplemental Table 2 of Atkinson FS, Brand-Miller JC, Foster-Powell K, Buyken AE, Goletzke J, International tables of glycemic index and glycemic load values 2021, Am J Clin Nutr 114:1625-1632, which together list over 4,000 items. The widely repeated blackberry GI of 25 traces only to third-party aggregator sites with no underlying measurement and was deliberately not used. The structural reason no value exists: an ISO 26642:2010 test portion of 25-50 g available carbohydrate would require roughly 0.6-1.2 kg of blackberries. NO GI OR GL FIGURE MAY APPEAR IN GENERATED COPY FOR THIS FOOD — do not state one, do not estimate one, and do not substitute a category or whole-fruit average in its place. The fibre-to-available-carbohydrate ratio is the substantive fact here.

Commonly-cited value (~55) from secondary GI databases; blackstrap specifically has NOT been independently Sydney-tested — treat as an estimate anchored to molasses generally. Carb content per tbsp also varies by source (11-15g); 13g used. Lower GI than white sugar (65) but still a concentrated sugar for verdict purposes.

USDA FoodData Central SR Legacy, "Cheese, blue" (FDC 172175): 2.34 g carbohydrate, 0 g fiber, 21.4 g protein per 100 g; 1 oz = 28.35 g per the USDA portion table, giving 0.7 g available carbohydrate and 6 g protein per serving. GI recorded as n/a because non-carbohydrate foods are not GI-tested: the University of Sydney GI database and the Atkinson/Foster-Powell/Brand-Miller international tables index only foods with meaningful available carbohydrate, and blue cheese appears in neither. At 0.7 g available carbohydrate per serving, an ISO 26642:2010 test portion of 25-50 g available carbohydrate is unreachable.

International GI tables 2021 (Atkinson) — wild blueberries (Canada) GI 53±7, tested by Glycemic Index Laboratories. Cultivated blueberries were not separately confirmed in this pass; wild-berry figure used as the representative value, low-GI either way.

McCance & Widdowson's The Composition of Foods Integrated Dataset (CoFID) 2021, Public Health England, item 12-344 'Cheese, Brie, with outer rind removed' — carbohydrate 'Tr' (trace, <0.1g/100g), starch 0.0g, total sugars 'Tr', AOAC fibre 0.0g, protein 20.3g and fat 29.1g per 100g. Scaled to 30g: protein 6.1g, fibre 0g, available carbohydrate effectively 0g. GI is recorded as 0 under the convention for foods with no appreciable available carbohydrate; 0 here means 'not applicable', NOT 'measured as zero' — GI is not assigned to non-carbohydrate foods in the University of Sydney GI database or in any published international table (Foster-Powell 2002, Atkinson 2008, Atkinson 2021), which test only foods carrying meaningful available carbohydrate. Generated copy must not imply Brie was GI-tested.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

University of Sydney GI database (glycemicindex.com) — boiled brown rice. The white-vs-brown GI gap (73 vs 68) is real but smaller than commonly assumed.

Not independently GI-tested — brown sugar is sucrose with trace molasses, so the universal sucrose reference value (65, see white-sugar row) is applied. The 'brown sugar is gentler on blood sugar' assumption has no GI basis; it is metabolically equivalent to white sugar.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Published international GI tables (Foster-Powell/Atkinson) — boiled buckwheat groats GI 54, GL 16 for the standard 150g serve. Despite the name it is a low-to-moderate GI pseudo-grain; buckwheat flour products (noodles, pancakes) are a different, generally higher-GI story not covered by this row.

University of Sydney GI database / published tables — boiled bulgur GI 48, consistently cited across grain-GI comparisons. Cracked-but-intact wheat structure keeps it well below couscous and most breads made from the same wheat.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

NOT directly GI-tested in published clinical work — composition-based estimates range 31-39 across secondary databases; 35 is the middle. Lactose-driven and compositionally close to milk (see whole-milk row, 32), so low-GI is a safe directional call even without a measured figure.

Commonly-cited boiled value (~51) across GI databases; boiled/steamed preparation only — roasting runs meaningfully higher, same pattern as sweet potato. GL stays modest because squash is mostly water; verdict copy should lean on the low GL rather than the borderline GI.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

USDA FoodData Central, SR Legacy 172336 "Oil, canola": per 100 g, 100 g total fat, 0 g carbohydrate, 0 g protein, 0 g fibre. Fatty acid profile per 100 g: 7.36 g saturated (4.3 g palmitic), 63.3 g monounsaturated (61.7 g oleic 18:1), 28.1 g polyunsaturated of which 19.0 g linoleic (n-6) and 9.14 g alpha-linolenic (n-3); also 17.5 mg vitamin E and 71.3 µg vitamin K. USDA portion "tbsp" = 14 g, used as the serving and consistent with the existing oil rows in gi_table.csv. GI recorded as n/a: canola oil contains no carbohydrate to test and has no entry in Foster-Powell et al. 2002 or Atkinson et al. 2021. Fatty-acid figures deliberately taken from USDA rather than from industry bodies, which publish slightly different percentages.

Foster-Powell et al. 2002 international table, Am J Clin Nutr 76:5-56 — item 435 'Rockmelon/Cantaloupe, raw (Australia)': GI 65 +/- 9, 120g serving, 6g available carbohydrate, GL 4; single Australian study, 8 healthy subjects, bread reference. Fibre and protein from Health Canada Canadian Nutrient File code 1721 (Melon, cantaloupe, raw: 0.69g fibre, 0.84g protein, 8.16g carbohydrate per 100g), scaled to 120g. Caveat: CNF composition implies ~9g available carbohydrate in 120g rather than the 6g used in the GI test, which would give GL ~6; melon sugar concentration varies materially with ripeness and cultivar, and the literature also reports a value of 70 at higher sugar concentration. The published GI/GL pairing is kept here rather than mixing sources.

Boiled carrots ~39, but carrots are the classic example of GI measurement noise: published range spans ~16 (raw) to ~92 (an outlier early study), with boiled clustering 32-49. 39 is the commonly-cited boiled middle. Flag: verdict copy should note the infamous range and that GL is tiny regardless — carrots barely carry carbohydrate.

University of Sydney GI database (glycemicindex.com) — cashews (salted) GI 22±5, one of the few nuts with enough carbohydrate to be formally tested. Unlike most nuts (listed n/a in this table), cashews carry ~9g carbs per 30g, but GI and GL are both firmly low.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

University of Sydney GI database / Foster-Powell 2002; boiled/steamed cauliflower GI ~15, glycemic load 1.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Published values disagree sharply by variety: the widely-quoted 22 is from an old Canadian SOUR-cherry measurement, while dark SWEET cherries (the kind people actually snack on) tested at 63±6 in international tables. 43 is the middle — sources disagree by variety, this is the middle. Flag: verdict copy must not repeat the '22, one of the lowest-GI fruits' claim for sweet cherries.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

'The Effect of Cooking and Cooling Chickpea Pasta on Resistant Starch Content, Glycemic Response, and Glycemic Index in Healthy Adults', Metabolites 2024;14(11):585 (PMC11596303): GI 39 for freshly cooked chickpea pasta against a 50g glucose reference in 12 healthy adults aged 18-65; the same pasta cooled and reheated tested at GI 33, with resistant starch rising from 1.83 to 3.65 g/100g. Composition from the same paper: 53.5g available carbohydrate, 21.0g protein, 11.5g dietary fibre per 100g dry pasta, scaled here to a 57g dry portion (30.5g available carbohydrate). Wheat-pasta comparison figures for the notes are from USDA FoodData Central 'Pasta, dry, enriched' (13.04g protein, 3.2g fibre per 100g, i.e. ~7.4g protein and ~1.8g fibre per 57g dry). Caveats: single study on one commercial 100%-chickpea product; retail products that blend in rice, lentil or pea flour would change all four numbers, and the cooled-and-reheated figure (33) is a genuinely different value that generated copy should not blur into the 39.

University of Sydney GI database / international GI tables — boiled chickpeas GI 28, GL 8 for the standard 150g serve. Canned chickpeas run somewhat higher (~38) than home-boiled; hummus and roasted-chickpea snacks are separate preparations not covered by this row.

USDA FoodData Central, 'Spices, cinnamon, ground' (FDC ID 171320): per 100g, 80.6g carbohydrate of which 53.1g is fiber, plus 3.99g protein; USDA lists 1 tsp as 2.6g, giving ~0.7g available carbohydrate, 1.4g fiber and 0.1g protein per teaspoon. No measured GI exists for cinnamon and none is obtainable at culinary amounts; GI recorded as n/a per the non-carbohydrate-food convention used in this table, not as a measured value. Clinical context for the generator: Kort DH, Lobo RA. 'Preliminary evidence that cinnamon improves menstrual cyclicity in women with polycystic ovary syndrome: a randomized controlled trial'. Am J Obstet Gynecol 2014;211(5):487.e1-6 — improved menstrual cyclicity over 6 months, but no change in insulin resistance or serum androgens. Contrasting result: Hajimonfarednejad M, Nimrouzi M, Heydari M, Zarshenas MM, Raee MJ, Jahromi BN. Phytother Res 2018;32(2):276-283 — improved fasting insulin and HOMA-IR. The two disagree; present it as unsettled and do not state or imply any amount to take. CATEGORY NOTE: this row is filed under 'sweetener' only because the permitted category list has no 'spice' option — cinnamon contains no sugar and sweetens nothing, and copy must never call it a sweetener or a sugar substitute.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate). Borderline call: canned coconut milk does carry ~3g carbohydrate per 120mL, but that is far below the level at which GI testing is feasible or meaningful, so it is honestly treated as not-assigned rather than 'low-GI'. Sweetened carton 'coconut milk beverage' with added sugars is NOT covered by this row.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

University of Sydney GI database (glycemicindex.com) — tested coconut sugar at 54. The widely-marketed GI of 35 comes from a small Philippine (FNRI) study and is treated here as an outlier; 54 vs white sugar's 65 is a modest difference, not the dramatic one marketing implies.

USDA FoodData Central SR Legacy #171956 (fish, cod, Atlantic, cooked, dry heat): 0g carbohydrate, 0g fiber, 22.8g protein per 100g. GI recorded as n/a rather than 0 — cod has no appreciable available carbohydrate and is therefore absent from both the University of Sydney GI database and Foster-Powell et al. 2002 (Am J Clin Nutr 2002;76:5-56), which only test foods with meaningful available carbohydrate. GI is undefined for this food, not measured as zero, and no GI or GL figure should appear in generated copy. Pacific cod (SR Legacy #175178) is nutritionally close.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

USDA FoodData Central, SR Legacy FDC 169599 "Gelatins, dry powder, unsweetened": 85.6 g protein, 0 g carbohydrate, 0 g fibre per 100 g — used as the compositional reference because hydrolysed collagen peptides are the same protein enzymatically cut into shorter chains, giving 8.6 g protein per 10 g scoop. GI recorded as n/a because there is no available carbohydrate to test; collagen appears nowhere in Atkinson FS, Brand-Miller JC, Foster-Powell K, Buyken AE, Goletzke J, "International tables of glycemic index and glycemic load values 2021", Am J Clin Nutr 114:1625-1632, Supplemental Tables 1 or 2. Protein per scoop is a conservative floor: commercial peptide powders are typically 90-97% protein by weight, so a 10 g scoop may deliver slightly more than 8.6 g.

GI 61±6 — Atkinson FS, Foster-Powell K, Brand-Miller JC, "International Tables of Glycemic Index and Glycemic Load Values: 2008", Diabetes Care 2008;31(12):2281-2283, item 782 "Milk, condensed, sweetened" (Australia; 100 g serve, 54 g available carbohydrate, GL 33, glucose reference, n=11). The same value is carried in the ISO 26642:2010-compliant tier of Atkinson FS et al., "International tables of glycemic index and glycemic load values 2021: a systematic review", AJCN 2021, Supplemental Table 1 item 936. Composition from USDA FoodData Central SR Legacy 171275 "Milk, canned, condensed, sweetened": per 100 g, 54.4 g carbohydrate (all of it sugars), 0 g fiber, 7.91 g protein, 8.7 g fat; USDA lists 1 fl oz = 38.2 g, so a 2 tbsp pour carries ~20.7 g available carbohydrate. GL = 61 x 20.7/100 = 12.6, rounded to 13. Caveat: this row is sweetened condensed milk only — unsweetened evaporated milk is a separate product and this GI does not transfer to it.

Composition: USDA FoodData Central, SR Legacy FDC 172716 (Cookies, chocolate chip, commercially prepared, regular, higher fat, enriched) — 65.4 g carbohydrate, 2.0 g fibre, 5.1 g protein, 32.9 g sugars per 100 g; a 30 g serving gives 19.0 g available carbohydrate, 0.6 g fibre and 1.5 g protein. GL = 59 x 19.0/100 = 11.2, rounded to 11. GI = 59 is DISPUTED and DERIVED, not a single published figure. It is the mean of the 14 generic (non-reformulated) sweet biscuit entries in Atkinson FS, Brand-Miller JC, Foster-Powell K, Buyken AE, Goletzke J, "International tables of glycemic index and glycemic load values 2021", Am J Clin Nutr 114:1625-1632, Supplemental Tables 1 and 2, covering shortbread 64, vanilla wafers 78, rich tea 55 and 40, digestives 59 and 39, oatmeal biscuit 54, arrowroot 62 and 69, a plain sweet biscuit 79, shredded wheatmeal 62, "Biscuit, NS" 70, petit-beurre 53 and a rotary-moulded biscuit 47 — range 39 to 79, mean 59.4, rounded to 59. The published category mean in the same paper's Table 1 is lower (Cookies: 49±9, n=135, 84% classified low-GI), but that dataset is dominated by commercial biscuits reformulated and submitted by manufacturers for GI certification, so it understates a plain everyday cookie and was not used as the headline value. SCOPE LIMIT for generated copy: treat this GI as a rough midpoint of a genuinely heterogeneous category, not as a property of any particular cookie, and say so.

University of Sydney GI database (glycemicindex.com) — boiled sweet corn representative value (~52); published range is wide (~37-67 by variety and origin per international tables). This row is sweet corn only — popcorn and cornmeal products are separate foods.

University of Sydney GI database (glycemicindex.com) — cornflakes are a classic high-GI reference food (~81, published range ~72-93 by country/brand); consistently among the highest-GI breakfast cereals tested.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

University of Sydney GI database / international GI tables — boiled couscous (~65). Despite its whole-grain image, couscous is a refined-wheat pasta granule and behaves closer to white bread than to intact whole grains.

McCance & Widdowson's The Composition of Foods Integrated Dataset (CoFID) 2021, Public Health England, item 16-392 'Crab, white meat, purchased cooked' — carbohydrate 'Tr' (trace, <0.1g/100g), starch 'Tr', total sugars 'Tr', AOAC fibre 0.0g, protein 20.5g and fat 0.3g per 100g. Scaled to 85g: protein 17.4g, fibre 0g, available carbohydrate effectively 0g. Cross-check within the same dataset: item 16-441 'Crab, purchased cooked' gives 19.7g protein and item 16-391 'Crab, brown meat' gives 18.8g protein per 100g, all with trace carbohydrate — the zero-carbohydrate finding is consistent across every crab entry. GI recorded as 0 by the stated convention for foods with no appreciable available carbohydrate; it means 'not applicable' rather than a measured zero, since the published international tables do not assign GI to non-carbohydrate foods. IMPORTANT SCOPE LIMIT for the generator: this row covers plain crab meat only. Imitation crab stick (surimi) is bound with starch and sugar, carries meaningful available carbohydrate, and is NOT described by these figures — it needs its own slug if it is to be covered.

Plain white-flour cracker representative value — University of Sydney-derived sources cite ~63-78 for plain crackers (water crackers ~78, soda crackers ~74); 70 is a representative middle for refined-flour crackers. Whole-grain/seed crackers can run substantially lower and are NOT covered by this figure — highly brand-variable category.

McCance & Widdowson's The Composition of Foods Integrated Dataset (CoFID) 2021, Public Health England, item 12-551 'Cheese, spreadable, full fat, soft, white' — the UK descriptor for full-fat cream cheese — 3.0g carbohydrate per 100g, all of it sugars (lactose; starch 0.0g), 0.0g AOAC fibre, 5.3g protein and 24.4g fat per 100g. Scaled to 30g: 0.9g available carbohydrate, 0g fibre, 1.6g protein. GI recorded as 0 by the stated convention for foods with no appreciable available carbohydrate; no measured GI for cream cheese exists in Foster-Powell et al. 2002, Atkinson et al. 2008, or Atkinson et al. 2021, and 0 here means 'not applicable' rather than measured. CAVEAT WORTH CARRYING INTO COPY: US-labelled cream cheese is typically listed a little higher in carbohydrate than the UK figure (roughly 4-6g per 100g), and whipped, low-fat and flavoured tubs differ again — flavoured and 'fruit' varieties can carry added sugar that changes the picture materially. No USDA entry is cited for the US figure, so it is flagged here rather than sourced, and the 0.9g per 30g above is a low-end figure for a plain full-fat product. Comparison within CoFID: quark (12-174) gives 14.6g protein and ricotta (12-176) 9.4g per 100g, both far above cream cheese.

Foster-Powell, Holt & Brand-Miller, International table of glycemic index and glycemic load values 2002, Am J Clin Nutr 76:5-56 — bakery products item 10: GI 67, 57g serving, 26g available carbohydrate, GL 17; a single commercial Canadian bakery croissant tested in 13 type 1/2 diabetic subjects against a white-bread reference (the 67 quoted here is the glucose-scale value from the table, not the bread-scale figure). Fibre and protein from USDA FoodData Central SR Legacy 174987 (Croissants, butter: 2.6g fibre, 8.2g protein per 100g), scaled to 57g. Caveats: this is one tested product, not a mean of studies, and croissant weight varies widely by bakery — GL scales directly with the pastry's weight. Reassuringly, USDA-derived available carbohydrate for 57g (~24.6g) rounds to the same GL of 17.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Commonly-cited value (~23) for 70%+ cacao dark chocolate — low because fat and cocoa solids dominate and available carbohydrate per serving is small. Applies to 70%+ cacao only; sweeter 50-60% 'dark' chocolates run higher, and milk chocolate is a separate food.

Published GI value (~42) is for deglet noor variety; medjool-specific GI was not independently confirmed in this research pass — medjool has a notably higher sugar density than deglet noor, so treat this figure as a LOWER-BOUND estimate and flag for independent medjool-specific verification.

Atkinson FS et al., International tables of glycemic index and glycemic load values 2021 (Am J Clin Nutr), Supplemental Table 1, entry 61: 'Doughnut', Oman 2020, GI 75+/-3, n=12, 50g available-carb test portion, ISO 26642:2010-consistent. CAVEATS: this is the only ISO-consistent plain-doughnut entry in the table, the study used a glucometer rather than a lab enzymatic method, and the serving size is recorded as 'NS' (not stated). It converges well with the classic ~76 value cited for cake-type doughnuts in earlier Foster-Powell tables, which raises confidence. The regional Emirati doughnut-cakes in the same table (entry 1976 Khanfaroosh GI 45+/-3, entry 1977 Legemat GI 44+/-4, UAE 2017) are different foods and are NOT represented by this value. Composition from USDA FoodData Central SR Legacy 172758 (Doughnuts, yeast-leavened, glazed, enriched): 47.93g carbohydrate, 2.1g fiber, 6.14g protein per 100g; serving from that record's USDA food_portion entry, 1 medium (approx 3-3/4 inch dia) = 60g. Cake-type doughnuts differ materially in composition (SR Legacy 174990: 47.06g carb, 24.93g fat per 100g).

Kamchansuppasin A. et al., 'Glycaemic index and glycaemic load of commonly consumed Thai fruits', International Food Research Journal 28(4): 788-794 (2021) — white dragon fruit (Hylocereus undatus) GI 51.9 ± 4.7 against glucose = 100, n = 12 healthy subjects, 25g available-carbohydrate test load; the paper's Table 2 gives 13.26g total carbohydrate, 1.70g dietary fibre and 11.56g available carbohydrate per 100g. This is the study behind dragon fruit's first appearance in the GI tables (Atkinson et al., 'International tables of glycemic index and glycemic load values 2021', AJCN, which names dragon fruit among the new Asian-fruit entries). Protein 1.0g/100g from the peer-reviewed characterisation of Hylocereus species (PubMed 33774520), which reports 0.90-1.1g per 100g fresh weight for H. undatus. TWO CAVEATS: (1) this row is the WHITE-fleshed fruit; red-fleshed H. polyrhizus tested higher in the same study at GI 57.2 ± 6.4 with a medium GL, so a red-flesh card would read differently. (2) The paper's own GL for its 160g serving is 8.8 versus the 10 computed here, because the paper appears to apply edible-portion weight rather than a flat 160g of flesh; the difference does not change the low-to-moderate verdict.

University of Sydney GI database / international tables — dried apricots ~31, among the lowest-GI dried fruits tested (secondary sources cite up to ~42). Low GI but sugar-dense — portion size drives the real-world load.

No University of Sydney-tested value found in this research pass; secondary GI databases cluster at ~56-60 for dried mango, and 60 is used as the representative middle. Most commercial dried mango is sugar-added, which plausibly pushes it above 70 — treat as an estimate, flagged for independent verification.

SCOPE — READ FIRST: every figure in this row describes DRIED SWEETENED cranberries, not fresh. The slug should be renamed to 'dried-cranberries' before import; if it is published under a bare 'cranberries' identifier, the copy must state the dried sweetened form in the first sentence and must not attach GI 62 or GL 19 to fresh cranberries. GI: University of Sydney GI database (glycemicindex.com) — "Cranberries, sweetened, dried": GI 62 (USA, Sydney University's Glycemic Index Research Service, unpublished observations, n=10, 2011). Composition from USDA FoodData Central SR Legacy #171723 (cranberries, dried, sweetened): 82.8g carbohydrate, 5.3g fiber, 0.17g protein, 72.6g total sugars per 100g; 1/3 cup = 40g. Fresh raw cranberries have NO GI value in either the Sydney database or Foster-Powell et al. 2002 — verified by direct search of both — as they are too tart and too low in available carbohydrate to be tested by the standard protocol (USDA SR Legacy #171722, cranberries, raw: 12.0g carbohydrate and 3.6g fiber per 100g, i.e. ~8.4g available carbohydrate per 100g). No GI or GL may be stated or estimated for the fresh fruit. Cranberry juice cocktail is a third distinct case, GI 52-68 across three tested products in Foster-Powell entry 402.

USDA FoodData Central, SR Legacy 172411 "Duck, domesticated, meat only, cooked, roasted" — per 100 g: 23.5 g protein, 11.2 g total fat (3.95 g saturated), 0 g carbohydrate, 0 g fiber. For contrast, USDA SR Legacy 172409 "Duck, domesticated, meat and skin, cooked, roasted" is 19.0 g protein and 28.4 g fat per 100 g, so skin status changes fat far more than it changes protein. GI recorded as n/a: duck is absent from the University of Sydney international GI tables (Atkinson et al., Diabetes Care 2008;31:2281-2283; Atkinson et al., AJCN 2021 Suppl. Tables 1-2) because foods without meaningful available carbohydrate are not GI-tested.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

White English muffin representative value — cited at ~69-77 across GI sources (international tables list English muffin bread at 77); 70 used as conservative-middle. Whole-wheat versions run somewhat lower.

Livesey G, "Health potential of polyols as sugar replacers, with emphasis on low glycaemic properties", Nutrition Research Reviews 2003;16:163-191 — pooled human glycaemia data assign erythritol a glycaemic index of 0 and an insulinaemic index of 2 (only mannitol shares the GI of 0 among polyols). This 0 is a measured value, not the not-assigned convention used for untested non-carbohydrate foods. Roughly 90% is absorbed and excreted essentially unchanged in urine, so it is not available carbohydrate despite being declared as carbohydrate (sugar alcohol) on labels; available carbohydrate for the serving is therefore 0 and GL is 0 by definition. Erythritol has no entry in Atkinson et al. 2021 or Foster-Powell et al. 2002, as expected for a polyol assessed separately. DISPUTED, but not on glycaemia: EFSA, "Re-evaluation of erythritol (E 968) as a food additive", EFSA Journal 2023;21:e8430, set an acceptable intake anchored to a human no-observed-adverse-effect level for the laxative effect; and Witkowski M et al., "The artificial sweetener erythritol and cardiovascular event risk", Nature Medicine 2023 (doi:10.1038/s41591-023-02223-9) reported an association between circulating erythritol and major adverse cardiovascular events plus platelet aggregation effects in preclinical models. Neither finding changes the GI value.

International GI tables (per glycemicindex.com) list pearled farro at 63 — the only formally-tested figure. The widely-repeated 'farro is low-GI (~40)' claim refers to untested estimates for whole intact emmer; most retail farro is pearled, so 63 is the honest representative. Verdict copy should not repeat the low-GI marketing claim.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Fresh figs: commonly-cited ~51 but formal GI testing of fresh figs is scarce — treat as an estimate. Dried figs ARE tested (~61, University of Sydney) and are a materially different, more sugar-dense food; this row covers fresh only.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Published values span ~63 (Atkinson international tables) to ~75 (Foster-Powell 2002); 70 used as representative middle. Frying method, cut and potato variety all move the number — directionally high-GI regardless of exact figure.

Commonly-cited value (~51), comparable to regular ice cream — the 'healthier than ice cream' framing is not supported on the glycemic axis. Highly brand-variable (sugar content swings widely by brand and flavor); treat as directionally moderate, not precise.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

USDA FoodData Central / non-carbohydrate beverage; GI is undefined for herbal infusions without available carbohydrate.

International tables list gluten-free white bread at ~80 and gluten-free multigrain at ~79; a 2021 systematic review (Foods, PMC7996770) found ~61% of 132 gluten-free bread samples were high-GI. 79 is representative for standard rice/corn-starch GF bread — formulation varies enormously by brand, and 'gluten-free' typically means HIGHER GI than regular bread, not lower.

USDA FoodData Central, 'Cheese, goat, soft type' (FDC ID 173435): 18.5g protein, 0g carbohydrate, 0g fiber, 0g sugars per 100g; USDA lists 1 oz as 28.35g. Cross-checked against Health Canada Canadian Nutrient File code 98 ('Cheese, goat, soft': 18.52g protein, 0g carbohydrate per 100g). GI recorded as n/a per the non-carbohydrate-food convention used in this table — ripened cheese retains essentially no lactose and is not GI-tested. Note for the generator: firmer styles are more protein-dense (USDA FDC ID 173433, 'Cheese, goat, semisoft type': 21.6g protein, 0.12g carbohydrate per 100g), so this row understates aged goat cheeses. The A1/A2 beta-casein point commonly attached to goat dairy comes from digestive-tolerance research in general populations, not from PCOS trials — it must not be presented as a PCOS finding.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (the University of Sydney GI database and all major published GI tables only test foods with meaningful available carbohydrate), so GI is recorded as n/a rather than 0: it is undefined, not measured-as-zero. Confirmed absent from both Foster-Powell et al. 2002 and Atkinson et al. 2008 international tables. Composition from McCance & Widdowson's Composition of Foods Integrated Dataset (CoFID) 2021, Public Health England, 'Cheese, Gouda': 25.3g protein, carbohydrate recorded as trace (Tr), 0.0g fibre per 100g. Slug follows the existing '<name>-cheese' pattern already in the repo (cheddar-cheese, feta-cheese, mozzarella-cheese, parmesan-cheese, ricotta-cheese) and the 30g serving matches those rows so the cheese cards stay comparable.

Representative value for typical oat-based granola with added sweeteners — no single Sydney-tested 'granola' figure exists; published muesli/granola-family values span roughly ~40-66 depending on sugar, oat form and fat content. 60 is a middle for sweetened commercial granola; low-sugar/high-nut recipes run meaningfully lower — highly brand-variable category.

International tables' muesli-bar-with-dried-fruit value (~61) used as representative; a gluten-free muesli bar tested at ~50. Syrup binders, fruit purees and protein/fiber content make this category extremely brand-variable — treat as a category midpoint, not a per-product figure.

University of Sydney / international tables — published grape values span ~46-59 (green seedless ~54, red seedless ~50); 53 used as representative middle. Moderate GI but easy to over-portion; glycemic load scales fast with serving size.

Commonly-cited plain Greek yogurt value (~14), among the lowest GI foods measured — protein content is the driver. Sweetened/flavored versions are NOT covered by this figure (see regular-yogurt, pending_research).

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

University of Sydney GI database (glycemicindex.com) — boiled green peas ~48 (cited range ~39-54 across tests). Fiber and protein content keep GL low despite the mid-range GI.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Passos TU, Sampaio HAC, Sabry MOD, Melo MLP, Coelho MAM, Lima JWO. 'Glycemic index and glycemic load of tropical fruits and the potential risk for chronic diseases'. Food Science and Technology (Campinas) 2015;35(1):66-73 (SciELO) — guava GI 12±11.3, GL 1±1.3, FAO/WHO protocol, glucose reference, n=6, 95g test portion. Composition from USDA FoodData Central 'Guavas, common, raw' (FDC ID 173044): 14.3g carbohydrate, 5.4g fiber, 2.55g protein per 100g, so 8.9g available carbohydrate; USDA lists 1 fruit (edible portion) as 55g. IMPORTANT CAVEAT: the GI figure is weak evidence — small sample and a standard deviation almost as large as the mean, and guava does not appear in the Atkinson 2008 international tables at all. The low-GL conclusion is nevertheless robust, because even a GI of 40 would put this serving at GL 4; copy should lean on the glycemic load and the fiber, not on the precision of '12'.

Matvaretabellen, the Norwegian Food Composition Table (Norwegian Food Safety Authority, Norwegian Directorate of Health and University of Oslo), entry "Cheese, Halloumi", drawing on the Swedish National Food Agency food database: 22.3g protein, 0.9g carbohydrate, 0g fiber, 21.9g fat per 100g. GI recorded as n/a rather than 0 — no appreciable available carbohydrate (~0.5g per serving, far below an ISO 26642:2010 test portion), and halloumi appears in neither the University of Sydney GI database nor Foster-Powell et al. 2002. GI is undefined here, not measured as zero, and no GI or GL figure should appear in generated copy. Caveat driving the medium confidence: halloumi has no entry in USDA FoodData Central's Foundation or SR Legacy datasets (verified by direct query — zero hits outside branded products), so composition rests on European national tables, and reported carbohydrate for halloumi varies from roughly 0 to about 2g per 100g between them, reflecting genuine batch variation in residual lactose. Sodium was not sourced here and should be treated as pending_research if the generator wants to cite a figure.

USDA FoodData Central, 'Ham, sliced, restaurant' (FDC ID 746952, Foundation Foods): 19.6g protein and 2.36g carbohydrate per 100g; USDA lists 1 slice as 16.2g. GI recorded as n/a per the non-carbohydrate-food convention used in this table — cured meats are not GI-tested. Caveat: this row describes plain sliced ham. Sweet-cured products carry meaningfully more sugar — Health Canada Canadian Nutrient File code 1148 ('Deli-meat, ham, cooked, sliced') lists 5.82g carbohydrate and 1.83g total sugars per 100g, and honey/maple-glazed products run higher still. Sodium also varies several-fold across products and is the more relevant variable than carbohydrate for this food.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

USDA FoodData Central, SR Legacy FDC 175117 (Fish, herring, Atlantic, cooked, dry heat) — 23.0 g protein, 0 g carbohydrate, 0 g fibre, 11.6 g fat, 203 kcal per 100 g; 1 fillet = 143 g, 1 oz = 85 g. GI recorded as n/a: not a significant carbohydrate source, and GI is not assigned to such foods per the convention used throughout this table and in the University of Sydney / Atkinson 2021 international tables, which only test and list foods with meaningful available carbohydrate. This row describes plain cooked Atlantic herring; pickled (FDC 175118), kippered (FDC 173668) and raw (FDC 175116) are separate USDA entries with different composition, and the pickled version in particular is not covered by these figures.

Mean of published honey-variety GI studies (range ~35-85 by floral source, per multiple honey-specific GI studies incl. Mexican honey varieties, PMC10572678). 61 is a commonly-cited conservative-middle value, not a single-variety figure — sources disagree by variety, this is the middle.

University of Sydney GI database (glycemicindex.com) — chickpea hummus tested at ~6, one of the lowest GI values in the database. Fat+fiber+low available carb per serving makes GL effectively zero; store-bought recipes vary but stay very low.

Atkinson international GI tables (2008/2021) — regular ice cream mean ~51, but the cited brand range is wide (~37-80) depending on fat and sugar content; premium high-fat runs lower, low-fat/high-sugar runs higher. 51 is the published mean, not a per-brand figure.

GI 51 = "Instant noodles, mean of three foods" in Atkinson FS, Brand-Miller JC, Foster-Powell K, Buyken AE, Goletzke J, "International tables of glycemic index and glycemic load values 2021: a systematic review", Am J Clin Nutr 2021 — individual study values 48±8, 52±7 and 52±5, all glucose-referenced, ISO methodology, and two of the three tested the noodles including broth, which matches the whole-packet serving used here. Composition from USDA FoodData Central, SR Legacy 171177 "Soup, ramen noodle, any flavor, dry": per 100 g, 60.26 g carbohydrate by difference, 2.9 g fibre, 10.17 g protein, 17.6 g fat, 1860 mg sodium. USDA's own portion weights give 81 g for the noodle block and 5.8 g for the flavour sachet, hence the 87 g as-sold serving. Available carbohydrate = 52.3 - 2.5 = 49.8 g; GL = 51 x 49.8/100 = 25.4, rounded to 25. Sodium works out to roughly 1600 mg for the packet. CAVEAT: USDA's entry is a composite of the whole dry product, so applying per-100 g values across block-plus-sachet is a close approximation rather than an exact split. The older Foster-Powell K, Holt SHA, Brand-Miller JC, Am J Clin Nutr 2002;76:5-56 gave a mean of 47 for the same food on a 180 g cooked / 40 g available-carbohydrate serving (GL 19); 51 is the newer, larger-evidence figure and is used here.

Atkinson FS, Foster-Powell K, Brand-Miller JC, 'International Tables of Glycemic Index and Glycemic Load Values: 2008', Diabetes Care 31(12):2281 — entry 1024, 'Strawberry jam (Australia)', GI 51+/-10, bread reference, n=9, 30g serving, 20g available carbohydrate, GL 10. Unchanged from Foster-Powell et al. 2002 (Am J Clin Nutr 76:5-56) entry 437, i.e. replicated across both editions. Fibre and protein from CoFID 2021, 'Jam, stone fruit' (69.3g available carbohydrate, 1.0g AOAC fibre, 0.4g protein per 100g), which independently corroborates the 20g-per-30g carbohydrate figure (69.3 x 0.3 = 20.8g). CoFID's 'Jam, fruit with edible seeds' is nearly identical on carbohydrate (69.0g) but its fibre is recorded as not determined, which is why the stone-fruit entry is used for fibre. Caveat: this row represents standard sugar-set jam only. In the same 2008 table, 100%-fruit and no-added-sugar spreads tested at 26-46 (entries 996, 1019, 1020), and CoFID reduced-sugar jam has less than half the carbohydrate (31.9g per 100g) — a single 'jam' verdict does not transfer across product types and the card should say which one it is describing.

Foster-Powell K, Holt SHA, Brand-Miller JC 2002 (Am J Clin Nutr 76:5-56) entry 288, and Atkinson FS, Foster-Powell K, Brand-Miller JC 2008 (Diabetes Care 31:2281) entries 539 and 1742 — 'Jasmine rice, white long grain, cooked in rice cooker', GI 109+/-10, glucose reference, n=12, 150g cooked serving, 42g available carbohydrate, GL 46. Replicated across both editions of the table. Independently supported by Reader et al., 'Glycemic index of American-grown jasmine rice classified as high', Int J Food Sci Nutr 2014;65(4), which measured GI 96-116 across four brands (n=12, 50g available carbohydrate, glucose reference) and found no difference between American- and Thai-grown jasmine. Disputed at the low end: Atkinson 2008 entry 538 lists a branded Australian jasmine rice at GI 89+/-4 on the same 150g/42g serving, so the full tested range across products is 89-116. 109 is used as it is the replicated value and the one carried by both table editions; the 89 outlier should be acknowledged in copy. Fibre and protein from CoFID 2021, 'Rice, white, long grain, boiled in unsalted water' (0.5g AOAC fibre, 2.8g protein per 100g) — CoFID has no jasmine-specific entry, and jasmine is a white long-grain aromatic rice. The 150g/42g serving matches the existing white-rice and brown-rice rows in gi_table.csv, so the three rice rows stay directly comparable. Generated copy must not name any manufacturer or product brand.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

University of Sydney GI database (glycemicindex.com) — boiled red kidney beans ~24 (cited range ~19-29 across preparations); canned runs slightly higher. Among the lowest-GI carbohydrate staples measured.

USDA FoodData Central, 'Kimchi' (FDC ID 2710077, FNDDS 2021-2023): 2.4g total carbohydrate, 1.6g fiber, 1.1g protein per 100g, so ~0.8g available carbohydrate per 100g serving; USDA lists 1 cup as 150g. No published glycemic index exists for kimchi — it has never been GI-tested because the standard protocol requires a 50g available-carbohydrate test portion, which is unreachable for this food. GI recorded as n/a per the non-carbohydrate-food convention used in this table, NOT as a measured value, and generated copy must not imply it was measured. Product variability is the real caveat: seasoning pastes differ in added sugar and rice flour, and sodium varies widely between brands.

University of Sydney GI database / Atkinson tables — green kiwifruit cited in the ~47-58 range across tests; 52 used as representative middle. Gold varieties and riper fruit sit toward the top of the range.

NOT formally GI-tested in the University of Sydney database or Atkinson tables — 18 is a secondary-database estimate (logifoodcoach.com) consistent with kombucha's low residual sugar (~2-8g per 240ml, heavily brand- and fermentation-dependent). Treat as directionally low, not a measured value; check the label's sugar content per brand.

Limited primary GI data — lactase hydrolysis converts lactose to glucose+galactose, which published dairy-GI comparisons (e.g. PMC9914410) suggest nudges GI slightly above regular milk (~32) while staying low; secondary databases cite ~30-46 by fat level. 40 is a conservative middle of that range, flagged as an estimate rather than a Sydney-tested value.

USDA FoodData Central SR Legacy, "Lamb, composite of trimmed retail cuts, separable lean and fat, trimmed to 1/4\" fat, choice, cooked" (FDC 172480): 0 g carbohydrate, 0 g fiber, 24.52 g protein per 100 g. GI recorded as n/a because non-carbohydrate foods are not GI-tested (University of Sydney GI database; Atkinson/Foster-Powell/Brand-Miller international tables), and lamb appears in neither. The composite cut is the general reference; protein varies modestly by cut, with lean-only leg (shank half, roasted, FDC 172487) higher at 28.17 g per 100 g and untrimmed cuts lower, so trim level is the main variable. The 100 g serving matches the existing meat rows in gi_table.csv.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Atkinson international GI tables — lentils mean ~32 across types (cited range roughly 21-52 by color and preparation; red split lentils cook down faster and run higher than intact green/brown). 32 is the published cross-type mean.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

USDA FoodData Central, SR Legacy 168626 "Beef, variety meats and by-products, liver, cooked, braised": per 100 g, 29.1 g protein, 5.13 g carbohydrate by difference (essentially residual glycogen), 0 g fibre, 6.54 mg iron, 9440 µg RAE vitamin A (9430 µg as retinol). Scaled to the USDA 85 g (3 oz) portion: 24.7 g protein, 4.4 g available carbohydrate, 0 g fibre, ~8000 µg RAE preformed vitamin A. GI recorded as n/a: liver has no published GI, appearing in neither Foster-Powell et al. 2002 nor Atkinson et al. 2021, and 4.4 g of available carbohydrate is far below the 25-50 g test dose GI methodology requires. n/a rather than 0 because the carbohydrate present is glycogen and therefore genuinely glucose-yielding — a published 0 would assert a measured zero response that does not exist. AMBIGUITY FLAG: "liver" here means beef liver, the usual reference in composition tables. Chicken liver (USDA SR Legacy 171061) is materially different — 0.87 g carbohydrate, 24.5 g protein and 3980 µg RAE vitamin A per 100 g — so unqualified "liver" copy averages two distinct foods; the ficha should state which one it means.

Health Canada Canadian Nutrient File code 3210 (Crustaceans, lobster, american/northern, boiled or steamed): 19.0g protein, 0g carbohydrate, 0g fibre, 0g sugars per 100g, scaled to a 100g cooked portion. GI is not assigned to foods without meaningful available carbohydrate in the University of Sydney GI database or any published GI table; 0 is used per this task's stated convention and means 'not applicable' rather than a measured zero. Note: spiny/rock lobster is a separate CNF entry (codes 3106/3174) with a slightly different profile; this row is the northern/American species, which is what 'lobster' denotes in most searches.

USDA FoodData Central SR Legacy 175120 "Fish, mackerel, Atlantic, cooked, dry heat" — per 100 g: 23.8 g protein, 17.8 g total fat (4.18 g saturated), 0 g carbohydrate, 0 g fiber, 83 mg sodium. GI recorded as n/a: absent from the University of Sydney international GI tables (Atkinson et al., Diabetes Care 2008;31:2281-2283; Atkinson et al., AJCN 2021 Suppl. Tables 1-2) because foods without meaningful available carbohydrate are not GI-tested. Species caveat sourced to the joint US FDA/EPA "Advice About Eating Fish" chart (2017 revision, Federal Register 82 FR 6571): Atlantic mackerel and chub mackerel appear under "Best Choices" (lowest mercury), while king mackerel appears under "Choices to Avoid" (highest mercury). This distinction must not be flattened in generated copy.

University of Sydney GI database (glycemicindex.com) — ripe mango ~51, but ripeness is the dominant variable: cited range ~41-60 from firm-unripe to very ripe. 51 is the representative mid-ripeness value.

Published value for PURE maple syrup (~54, cited as 54±6); pancake-style 'maple-flavored' syrup is a different product and runs higher (~68). Modestly below white sugar (65) — a real but small advantage, not a free pass.

Health Canada Canadian Nutrient File code 7458 (Margarine, stick, canola and/or soybean oil): 1.39g carbohydrate, 0g fibre, 0.16g protein per 100g — that is 0.2g carbohydrate and 0.02g protein in a 14g tablespoon, neither a meaningful available-carbohydrate load nor a meaningful protein contribution. GI is not assigned to foods without meaningful available carbohydrate in the University of Sydney database or any published GI table; 0 is used per this task's stated convention. Caveat: 'margarine' spans stick, non-hydrogenated tub, plant-sterol and reduced-fat spread formulations (CNF codes 7458-7476, 7574-7575) whose fat composition differs substantially. They do not differ meaningfully on carbohydrate, so the glycemic fields hold across the category, but any fat-quality statement in generated copy must name which formulation it refers to.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

USDA FoodData Central SR Legacy #171009 (salad dressing, mayonnaise, regular): 0.57g carbohydrate, 0g fiber, 0.96g protein per 100g; 1 tbsp = 13.8g. GI recorded as n/a rather than 0 — mayonnaise has no appreciable available carbohydrate (~0.1g per tablespoon, far below the 25-50g an ISO 26642:2010 test requires) and appears in neither the University of Sydney GI database nor Foster-Powell et al. 2002. GI is undefined here, not measured as zero, and no GI or GL figure should appear in generated copy. Worth flagging for the generator: light/reduced-fat mayonnaise is a materially different product (SR Legacy #173594, salad dressing, mayonnaise, light) because starch or sugar replaces the removed fat, so its carbohydrate per tablespoon is not negligible.

Atkinson international GI tables list chocolate at ~40±3; milk chocolate specifically is commonly cited 40-49 depending on brand sugar content. 43 used as representative middle — brand formulation moves this figure.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

USDA FoodData Central SR Legacy 174217 (Mollusks, mussel, blue, cooked, moist heat): 7.39g carbohydrate, 0g fiber, 23.8g protein, 4.48g fat per 100g. Serving from that record's USDA food_portion entry, 3 oz = 85g, giving ~6.3g available carbohydrate. GI reported as 'n/a' — but for a DIFFERENT reason than the other non-GI rows here, and generated copy must not flatten the distinction: mussels do carry appreciable carbohydrate as glycogen (more per 100g than whole milk), yet no GI has ever been published. Mussels appear nowhere in the 2021 International Tables, and an ISO 26642:2010 test would require roughly 700g of mussel meat to reach a 50g available-carbohydrate portion, which is not a realistic test food. The glycemic load is therefore NOT COMPUTABLE rather than negligible. Do not state, imply, or estimate a GI or GL for this food.

University of Sydney GI database (glycemicindex.com) — "Naan bread, heated in microwave for 25 sec": GI 71 (Australia, Sydney University's Glycemic Index Research Service, unpublished observations, n=10, 2013). Composition from USDA FoodData Central SR Legacy #171845 (bread, naan, plain, commercially prepared, refrigerated): 50.4g carbohydrate, 2.2g fiber, 9.62g protein per 100g; 1 piece = 90g. Caveats: the GI rests on a single commercial product tested once, and Foster-Powell et al. 2002 contains no naan entry at all (only chapatti variants, e.g. wholemeal roti at GI 45 per Radhika et al., Br J Nutr 2010;103:1642-7) — so the GI is less firmly established than the composition. Restaurant naan of 150-200g would carry roughly double the glycemic load shown.

Foster-Powell et al. 2002 international table, Am J Clin Nutr 76:5-56 — item 547 'Corn chips', mean of 3 studies: GI 63 +/- 10, 50g serving, 26g available carbohydrate. GL recomputed here as 63 x 26/100 = 16; the table's own printed mean GL of 17 is an average of the three studies' individual GLs (11, 18, 21), which is why it differs by one point. Individual study GIs ranged 42 to 74 — a wide spread. Fibre and protein from Health Canada Canadian Nutrient File code 4127 (Snacks, tortilla chips, plain: 5.4g fibre, 7.1g protein per 100g), scaled to 50g. Two caveats: (a) CNF composition for generic plain tortilla chips implies ~31g available carbohydrate per 50g rather than the 26g of the tested products, which would put GL nearer 20 — treat 16 as a lower bound; (b) the tested food is plain corn chips, while 'nachos' as served is a composite dish not covered by this figure.

Atkinson FS, Foster-Powell K, Brand-Miller JC, 'International Tables of Glycemic Index and Glycemic Load Values: 2008', Diabetes Care 31(12):2281 — entry 993, 'Nectarines, raw (Australia)', GI 43+/-6, glucose reference, n=10, 120g serving, 9g available carbohydrate, GL 4. Not present in the 2002 edition of the table; this is the single published measurement. Fibre and protein from CoFID 2021, 'Nectarines, flesh and skin, raw' (1.3g AOAC fibre, 1.0g protein per 100g). Minor discrepancy worth knowing: CoFID gives 9.8g available carbohydrate per 100g, so the same 120g serving would be ~11.8g and the GL ~5 rather than 4. The GI table's own tested serving figure (9g) is used here so that GI, carbohydrate and GL are internally consistent; either figure leaves the load clearly low.

Commercial oat milk is commonly cited at ~69 — enzymatic processing breaks oat starch into maltose (GI 105), so oat milk runs far higher than whole oats (55) and other milks. Strongly brand-dependent (published citations span roughly 60 to 80+ for sweetened versions); 69 is the widely-cited representative for standard commercial oat milk.

University of Sydney GI database (glycemicindex.com) — rolled oats representative value. Steel-cut runs lower (~42), instant runs higher (~79); processing level materially changes this, rolled oats used as the general reference.

USDA FoodData Central SR Legacy, "Okra, cooked, boiled, drained, without salt" (FDC 169261): 4.51 g carbohydrate, 2.5 g fiber, 1.87 g protein per 100 g; 1/2 cup slices = 80 g per the USDA portion table, hence 160 g per cup, giving 3.2 g available carbohydrate, 4 g fiber and 3 g protein. GI recorded as n/a: available carbohydrate is ~3 g per cup, below the level at which GI is measured, and okra appears nowhere in Atkinson/Foster-Powell/Brand-Miller 2008 Table A1. This matches how the existing broccoli and spinach rows in gi_table.csv are handled. IMPORTANT SCOPE LIMIT for any generated copy: okra has been tested for glycemic effect in randomised controlled trials and pooled analyses (Jafari et al., Experimental Physiology 2025, systematic review and meta-analysis across the diabetes spectrum; and a randomised double-blind placebo-controlled trial in type 2 diabetes, PMID 37507536), but these used concentrated okra powder or extract, not the cooked vegetable. That trial evidence must not be presented as applying to okra as eaten.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

University of Sydney GI database (glycemicindex.com) — whole orange ~43 (cited range ~31-51 by variety). Whole fruit only; orange juice is meaningfully higher per serving because fiber and portion control are lost.

USDA FoodData Central, 'Mollusks, oyster, eastern, wild, cooked, moist heat' (FDC ID 171980): 11.4g protein, 5.45g carbohydrate, 0g fiber, 1.23g total sugars per 100g; USDA lists 6 medium oysters as 42g and 3 oz as 85g. GI recorded as n/a per the non-carbohydrate-food convention used in this table — oysters have never been GI-tested. FLAG FOR THE GENERATOR: the carbohydrate here is stored glycogen, which is unusual for an animal protein, so do not describe oysters as 'zero carb' even though the amount per serving is far too small to produce a meaningful glycemic load. Raw oysters are less concentrated (Health Canada Canadian Nutrient File code 3118, eastern wild raw: 5.71g protein, 2.72g carbohydrate per 100g) because cooking drives off water; breaded fried oysters are a different food and are not covered by this row.

Atkinson FS, Foster-Powell K, Brand-Miller JC. International Tables of Glycemic Index and Glycemic Load Values: 2008. Diabetes Care 2008;31(12):2281-2283, Table A1 — 'Pancakes, homemade' GI 66±9, 80g serve, 26g available carbohydrate, GL 17 (glucose reference, n=8). The same table spans a wide range: gluten-free packet mix 61±6, shake mix 67±5, wheat-flour pancakes (China) 80±4, and buckwheat gluten-free packet mix 102±11 — recipe and mix choice move this substantially, and the 'healthier' buckwheat/gluten-free versions tested highest. Fiber and protein from USDA FoodData Central 'Pancakes, plain' (FDC ID 2708304: 35.32g carbohydrate, 2.1g fiber, 7.41g protein per 100g), which independently corroborates ~26g available carbohydrate in an 80g serving. These figures are for plain pancakes only — syrup, honey, butter or fruit compote are not included and would change the picture materially.

Indian Food Composition Tables 2017 (IFCT), National Institute of Nutrition, Indian Council of Medical Research (ICMR), Hyderabad — official ICMR-NIN publication, nin.res.in/ebooks/IFCT2017.pdf. Food code L003, Paneer, n=6 regions: available carbohydrate (column CHOAVLDF, carbohydrate available by difference) 2.41+/-0.12 g, protein (PROTCNT) 18.86+/-0.75 g, total fat (FATCE) 24.78+/-0.17 g, moisture 51.96+/-0.76 g per 100g. No dietary fibre reported (dairy). Verified internally: the listed components sum to 99.99g/100g and the stated energy of 1278 kJ reconciles with these macros, confirming correct column assignment. IFCT is used because paneer is absent from USDA SR Legacy. GI reported as 'n/a' rather than 0: ~2.4g available carbohydrate per 100g is far below the 25-50g ISO 26642:2010 test portion, and paneer appears nowhere in the 2021 International Tables — same convention as the existing cheese row in this CSV.

Atkinson international GI tables / University of Sydney GI database — papaya cited tightly at ~56-60; 59 is the published mean. One of the higher-GI whole fruits, though low available carb per serving keeps GL moderate.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Al dente white pasta, cited range ~42-48 across sources; 45 used as central value. Overcooked pasta runs meaningfully higher — degree of cooking is the dominant variable, more than pasta type.

University of Sydney GI database (glycemicindex.com) / International Tables of GI and GL 2008 — fresh raw peach, GI 42 (commonly cited mean; individual tests vary widely). Fibre and protein from USDA FoodData Central, 'Peaches, yellow, raw' (1.5g fibre, 0.91g protein per 100g). GL computed for the 150g serving from ~8.0g available carbohydrate per 100g (9.5g total minus 1.5g fibre): 42 x 12.1g / 100 = 5. Canned peaches in heavy syrup test higher and are not what this row describes.

Commonly-cited low value (~14) for smooth peanut butter without added sugar. Fat-dominant foods generally have less stable/reliable GI measurement than carb-dominant foods — treat as directionally low-GI rather than a precise figure. Added-sugar peanut butter brands will differ.

Not independently GI-tested in this research pass — using peanut butter's commonly-cited low value as a reasonable proxy given compositional similarity (same legume, minimal processing difference for GI purposes). Flag for independent verification.

Atkinson FS et al., International tables of glycemic index and glycemic load values 2021 (Am J Clin Nutr), Supplemental Table 1, entries 1160 and 1161: 'Pear, Williams, raw, under ripe/crunchy' GI 24+/-3 and 'Pear, Williams, raw, ripe/soft' GI 33+/-4 (both Australia 2015, n=10, 25g available-carb portion, ISO 26642:2010-consistent). 33 used as the ripe/typical-eating-state value; ripeness is the dominant variable and the 24-33 span should be reflected in generated copy. Note the older Foster-Powell-era tables commonly cite 38 for raw pear — the ISO-consistent 2021 value is lower, and 33 is preferred here as the methodologically stronger figure. Composition from USDA FoodData Central SR Legacy 169118 (Pears, raw): 15.23g carbohydrate, 3.1g fiber, 0.36g protein per 100g; serving from the same record's USDA food_portion entry, 1 medium = 178g. Available carb = (15.23 - 3.1) x 1.78 = 21.6g.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

USDA FoodData Central SR Legacy 168558 "Pickles, cucumber, dill or kosher dill" — per 100 g: 2.41 g carbohydrate, 1.0 g fiber, 0.5 g protein, 1.07 g sugars, 809 mg sodium. USDA lists 1 large (4 inch) = 135 g, giving ~1.9 g available carbohydrate, 1.4 g fiber, 0.7 g protein and ~1092 mg sodium per pickle. Not GI-tested: absent from both Atkinson et al. 2008 (Diabetes Care 31:2281-2283) and the 2021 ISO-compliant update (AJCN, Suppl. Tables 1-2), since ISO 26642:2010 testing requires a portion containing 25 or 50 g of available carbohydrate, unreachable for pickles. Recorded as n/a accordingly. Contrast for the fork: USDA SR Legacy 169378 "Pickles, cucumber, sweet (includes bread and butter pickles)" is 21.2 g carbohydrate and 18.3 g sugars per 100 g, so this row must not be generalised to sweet pickles. The acetic-acid/postprandial-glucose literature is separate from GI and is not represented by these numbers.

University of Sydney GI database (glycemicindex.com) — fresh pineapple mean ~59, cited range ~51-66 by ripeness and origin. 59 is the representative middle; canned-in-syrup is a different, higher product.

Atkinson international GI tables — boiled pinto beans ~39; canned runs slightly higher (~45). Higher than kidney beans (24) but still solidly low-GI with substantial fiber and protein per serving.

Cited low range ~15-28 across GI sources; 22 used as representative middle. Like other fat-dominant foods, GI measurement is less stable/reliable than for carb-dominant foods — treat as directionally low-GI rather than a precise figure.

Foster-Powell K, Holt SHA, Brand-Miller JC, 'International table of glycemic index and glycemic load values: 2002', Am J Clin Nutr 76:5-56 — entry 147, 'Pita bread, white (Canada)', GI 57, 30g serving, 17g available carbohydrate, GL 10. Serving scaled here to one standard 60g pita using CoFID 2021, 'Bread, pitta, white' (55.1g available carbohydrate, 9.1g protein, 2.3g AOAC fibre per 100g); the two sources agree closely on carbohydrate density (56.7 vs 55.1 g per 100g), so the scaling is sound — 60g x 55.1/100 = 33g available carbohydrate. Disputed by product: Atkinson et al. 2008 (Diabetes Care 31:2281) lists UK mini white pita at GI 68 (entry 295) and UK wholemeal pita at 56 (entry 296), so the white-pita range across tested products is 57-68 and the white-vs-wholemeal gap is small. 57 is the long-standing reference value and the more conservative of the two white-pita measurements.

Peer-reviewed kefir GI testing (J Am Coll Nutr 2012, PMID 23378456) centered on flavored kefirs (48-60); plain kefir is cited at ~25-36 across secondary GI databases. 32 is a conservative middle for PLAIN kefir — sweetened/flavored versions are materially higher and not covered by this figure.

Atkinson FS, Foster-Powell K, Brand-Miller JC. International Tables of Glycemic Index and Glycemic Load Values: 2008. Diabetes Care 2008;31(12):2281-2283, Table A1, entry 1859 — 'Ripe plantain (Musa paradisiaca), peeled, boiled 10 min (Jamaica)' GI 66±2, glucose reference, n=10. The same source gives the widest spread of any food in this batch: green plantain boiled 39±4 (entry 1857), green fried 40±3 (1858), ripe boiled 66±2 (1859), ripe fried 90±6 (1860). Ripeness and frying — not the food itself — decide the outcome, and a verdict written without naming both variables would be wrong. Composition from USDA FoodData Central 'Plantain, cooked, no added fat' (FDC ID 2709558): 31.8g carbohydrate, 1.7g fiber, 1.3g protein per 100g, 1 cup = 200g, giving 30.1g available carbohydrate in this serving. GL computed from that composition (66 x 30.1/100 = 20); note that the Jamaican study's own 120g test portion listed only 19g available carbohydrate for a published GL of 13, i.e. a less carbohydrate-dense plantain than the USDA reference, so 20 is the conservative figure.

Foster-Powell K, Holt SHA, Brand-Miller JC, "International table of glycemic index and glycemic load values: 2002", Am J Clin Nutr 2002;76:5-56, entry 432 "Plum, raw, NS": mean GI 39 ± 15 across two studies, 120g serving, 12g available carbohydrate, GL 5. Genuinely disputed value: the two underlying studies returned GI 24 and GI 53, both small and conducted in type 2 diabetic subjects, so 39 is a midpoint rather than a settled figure — the University of Sydney database lists both entries separately and adds a GI 32 result for Bau Kul (Zizyphus mauritiana), which is a jujube rather than a true Prunus plum and was excluded here. Fiber and protein from USDA FoodData Central SR Legacy #169949 (plums, raw): 11.4g carbohydrate, 1.4g fiber, 0.7g protein per 100g; 1 fruit (2-1/8 in dia) = 66g.

No primary University of Sydney value for fresh arils — 35 is the commonly-cited fresh-fruit figure, with sources spanning ~18-53 (the 53 belongs to pomegranate JUICE, which is Sydney-tested and NOT covered by this row). Flag: aril value is an estimate from secondary sources, weaker provenance than most rows in this table.

Atkinson international GI tables mean for plain popcorn (65±5); individual studies range ~55-89, so treat as a middle for plain air-popped. Buttered/kettle/movie preparations are not covered by this figure.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Foster-Powell/Atkinson international tables, plain salted potato chips (~54-57 across cited studies); frying fat blunts the response vs boiled potato (78). Low-ish GI here does not make this a low-glycemic-quality food — verdict copy should not read 56 as an endorsement.

University of Sydney GI database (glycemicindex.com) — HIGHLY variety-dependent: cited range ~53 (Carisma) to ~101 (some Russet Burbank/baked prep). 78 is a conservative-middle representative for BOILED white potato, excluding baked-prep extremes. Flag: verdict copy should note variety/prep matters more than for most foods in this table.

EXTREMELY brand-variable — measured bars span roughly GI 20s (low-carb/artificially-sweetened, e.g. NuGo Slim 26-31) to 60s (sugar-syrup-based); University of Sydney-tested Extend bars run 32-41. 40 is a representative middle for a mainstream whey-based bar, NOT a figure for any specific product; fiber/protein (3g/15g) are likewise representative of a 50g bar and vary widely by brand.

Atkinson FS et al., International tables of glycemic index and glycemic load values 2021 (Am J Clin Nutr), Supplemental Table 1. Two ISO 26642:2010-consistent entries for the same US producer's pitted prunes DISAGREE: entry 1168 'Prunes, pitted' GI 29+/-4 (USA 2000, n=10) and entry 1169 'Prunes, pitted, Californian' from the same producer GI 40+/-6 (USA 2009, n=10). 35 is used here as the midpoint of the two, following the conservative-middle convention already used in this CSV for honey and whole milk; note that 35 is therefore a derived value that appears in no source verbatim, and the disagreement should be visible in copy rather than hidden. Flag for generated copy: the widely-quoted 'prunes are GI 29' claim cherry-picks the lower of two conflicting measurements of the same product and should not be presented as settled. At GI 29 the GL would be 8; at GI 40 it would be 11, so the load is low across the whole disputed range. Composition from USDA FoodData Central SR Legacy 168162 (Plums, dried (prunes), uncooked): 63.88g carbohydrate, 7.1g fiber, 2.18g protein per 100g; serving from that record's USDA food_portion entry, 1 pitted prune = 9.5g, so 5 prunes = ~48g. Available carb = (63.88 - 7.1) x 0.48 = 27.3g. Generated copy must not name any manufacturer or product brand.

International GI tables (three studies) list boiled pumpkin at 64; some sources cite up to 75 for roasted/pureed prep. High-GI/low-GL case: so little available carbohydrate per serving that GL stays low (~6 per cup) despite the high GI — verdict copy should lead with GL.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Composition: Livsmedelsverket (Swedish Food Agency) Livsmedelsdatabasen, food no. 75 'Kvarg farskost fett 1%' (quark, fresh cheese, 1% fat) — 12.7 g protein, 3.6 g available carbohydrate, 2.9 g sugars, 75 kcal per 100 g, no dietary fibre. A national food-composition database was used because quark has no USDA FoodData Central SR Legacy or Foundation entry. Cross-check within the same database: food no. 3243 (Kvarg naturell, 0.2% fat) gives 10.0 g protein and 5.2 g available carbohydrate per 100 g, so protein varies by product tier and 19.1 g per 150 g is representative of the common 1% plain version rather than universal. GI recorded as n/a: quark appears in neither Supplemental Table of Atkinson FS, Brand-Miller JC, Foster-Powell K, Buyken AE, Goletzke J, International tables of glycemic index and glycemic load values 2021, Am J Clin Nutr 114:1625-1632, and at ~5 g available carbohydrate per serving it sits far below the 25-50 g portion an ISO 26642:2010 GI test requires. Same treatment as the existing cottage-cheese row, which is compositionally the closest food already in this table. Widely quoted quark GI figures of 30, 35 or 45 come only from third-party aggregator sites and were deliberately not used.

University of Sydney GI database (glycemicindex.com) — cooked quinoa 53, consistently cited across published tables; the fiber+protein package is what separates it from white rice (73) more than the GI number alone.

USDA FoodData Central, SR Legacy 169276 "Radishes, raw": per 100 g, 3.4 g carbohydrate by difference, 1.6 g fibre, 1.86 g sugars, 0 g starch, 0.68 g protein. USDA portion "cup slices" = 116 g, used as the serving. Available carbohydrate = (3.4 - 1.6) x 1.16 = 2.1 g; fibre 1.9 g; protein 0.8 g. Cross-checked against USDA Foundation Foods 2747665 "Radishes, red, raw" (4.06 g carbohydrate, 1.3 g fibre, 0.66 g protein per 100 g), consistent. GI recorded as n/a — radish appears in neither Foster-Powell et al. 2002 nor Atkinson et al. 2021, and available carbohydrate is negligible per serving. IMPORTANT LOOKUP TRAP: the only near-match in the Atkinson 2021 tables is "Chinese carrot cake / turnip cake / daikon cake, pan-fried shredded radish with plain rice flour" (Singapore, GI 77±8, GL 27) — that is a rice-flour dish and must NOT be used as a GI for raw radish.

University of Sydney GI database (glycemicindex.com) — raisins 64 (cited range ~49-66 across studies). Drying concentrates sugar, so GL per 30g (~14) is roughly what a whole cup of fresh grapes delivers — portion size is the story here.

USDA FoodData Central SR Legacy 173592 (Salad dressing, ranch dressing, regular): 5.9g carbohydrate, 0g fiber, 4.69g total sugars, 1.32g protein, 44.54g fat per 100g. Serving from that record's USDA food_portion entry, 1 tablespoon = 15g, so 2 tbsp = 30g. GI reported as 'n/a' rather than 0: a 30g serving supplies only ~1.8g available carbohydrate, and ISO 26642:2010 GI testing requires a test portion containing 25-50g of available carbohydrate — roughly 850g of dressing — so no GI is determinable, and none is published (ranch dressing appears nowhere in the 2021 International Tables). This is the USDA generic entry; sugar content varies meaningfully by brand and reduced-fat versions typically carry more carbohydrate (SR Legacy 173593).

GI recorded as n/a: fresh raspberries have no published glycaemic index measurement and none is obtainable at a realistic portion. Verified absent from both Foster-Powell K, Holt SHA, Brand-Miller JC 2002 (Am J Clin Nutr 76:5-56) and Atkinson FS, Foster-Powell K, Brand-Miller JC 2008 (Diabetes Care 31:2281): every raspberry entry in those tables is a manufactured product — raspberry yoghurts, fruit spreads, smoothies, ice cream, coffee cake, fruit bars — and none is the raw fruit. GI and GL must NOT be inferred from those product entries, must not be estimated, and no GI or GL figure may appear in generated copy for this food. At 5.1g available carbohydrate per 100g, an ISO 26642:2010 test portion of 25-50g available carbohydrate would require roughly 0.5-1.0 kg of raspberries, which is why no value exists. Composition IS verified and usable: CoFID 2021, Public Health England, 'Raspberries, raw' — 5.1g available carbohydrate, 3.7g AOAC fibre, 0.8g protein per 100g, giving 6.4g available carbohydrate, 4.6g fibre and 1.0g protein for the 125g serving. Treated the same way as the blackberries row, which is the identical case. The fibre-to-available-carbohydrate ratio, not the GI, is the substantive fact here.

University of Sydney GI database (glycemicindex.com) — canned refried pinto beans, reheated: GI 38 (Sydney University's Glycemic Index Research Service, unpublished observations, n=10, 2006). Composition from USDA FoodData Central SR Legacy #174296 (refried beans, canned, vegetarian): 13.5g carbohydrate, 4.7g fiber, 5.28g protein per 100g; 1 cup = 242g. Note: Foster-Powell et al. 2002 has no standalone refried-bean entry, only composite tortilla dishes; plain pinto beans in that table test at GI 39 (dried, boiled) and 45 (canned in brine), so refried preparation does not appear to push the value up despite the mashing.

Plain (non-Greek) yogurt cited ~19-41 across sources (Foster-Powell 2002: sweetened low-fat 33, artificially-sweetened 14; Atkinson mean 41); 36 used as a middle for PLAIN regular yogurt. Sweetened fruit yogurts carry materially higher GL from added sugar and are not covered by this row.

University of Sydney GI database (glycemicindex.com) — puffed rice cakes 82 (cited range 77-91), among the highest-GI snack foods tested; the 'diet food' halo is not supported by the glycemic data. Flavored/coated varieties differ.

Atkinson FS, Foster-Powell K, Brand-Miller JC. International Tables of Glycemic Index and Glycemic Load Values: 2008. Diabetes Care 2008;31(12):2281-2283, Table A1 — 'Rice noodles, dried, boiled' GI 61±6, 180g serve, 38g available carbohydrate, GL 23 (glucose reference, normal subjects). The same table lists 'Rice noodles, freshly made, boiled' at GI 40±4 and 'Rice vermicelli' at GI 58; dried vs fresh is the dominant variable, and 61 is used here as the supermarket-typical reference. Fiber and protein from USDA FoodData Central 'Rice noodles, cooked' (FDC ID 168914: 24.0g carbohydrate, 1.0g fiber, 1.79g protein per 100g; 1 cup = 176g). Caveat: USDA composition implies ~41g available carbohydrate per 180g rather than the 38g the GI study used, so real-world GL may run slightly above 23.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Wholemeal rye representative value (~58) from international GI tables; the rye category spans ~41 (true rye-kernel pumpernickel) to ~68 (light/refined rye), so flour coarseness and kernel content matter more than the word 'rye' on the label.

USDA FoodData Central, SR Legacy 174603 "Salami, Italian, pork": per 100 g, 21.7 g protein, 1.2 g carbohydrate by difference, 0 g fibre, 37 g fat, 1.52 mg iron. Scaled to a 30 g serving: 6.5 g protein, 0.4 g available carbohydrate, 0 g fibre. GI recorded as n/a — salami appears in neither Foster-Powell et al. 2002 nor Atkinson et al. 2021, consistent with never having been tested, and 0.4 g of available carbohydrate per serving is far below any testable dose. NOTE: Italian pork salami is used as the general reference. Other USDA entries differ materially (cooked beef salami SR 172935 and cooked turkey salami SR 172937 are lower in fat and higher in water), so per-gram protein, fat and sodium should not be treated as fixed across salami styles.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

GI recorded as n/a: no published GI measurement exists for seitan, and none can be produced at a realistic portion. Verified absent from Foster-Powell K, Holt SHA, Brand-Miller JC 2002 (Am J Clin Nutr 76:5-56), Atkinson et al. 2008 (Diabetes Care 31:2281), and CoFID 2021 — all three checked directly. At ~4g available carbohydrate per 100g, an ISO 26642:2010 test portion of 25-50g available carbohydrate is unreachable, so GI is undefined rather than zero. Composition from the Norwegian Food Composition Table (Matvaretabellen, Norwegian Food Safety Authority / Norwegian Directorate of Health), entry 'Seitan, wheat gluten': 28.4g protein, 4g available carbohydrate, 6g fibre per 100g; primary reference given as 'Nutrient content in plant-based protein products intended for food composition databases', Journal of Food Composition and Analysis 106 (2022). CAVEATS: (1) third-party sites quote GI 15 and GI 64 for seitan — mutually contradictory, untraceable to any study, and neither may be used or approximated in copy; (2) prepared seitan varies materially by brand and recipe (added chickpea flour, soy sauce, yeast), so composition is representative rather than exact; (3) the 6g fibre figure is high for a near-pure gluten product and most likely reflects commercial products containing added flours — copy should not lead with the fibre number or present it as characteristic of seitan generally.

USDA FoodData Central SR Legacy, "Oil, sesame, salad or cooking" (FDC 171016): 0 g carbohydrate, 0 g fiber, 0 g protein, 100 g fat per 100 g. The USDA portion table gives 1 tablespoon = 13.6 g, rounded to 14 g for consistency with the existing olive-oil, coconut-oil, butter and ghee rows in gi_table.csv. GI recorded as n/a because non-carbohydrate foods are not GI-tested (University of Sydney GI database; Atkinson/Foster-Powell/Brand-Miller international tables), and sesame oil appears in neither. Toasted and cold-pressed sesame oils do not differ in carbohydrate content, so this row covers both.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Foster-Powell 2002 lists skim milk at 32; Atkinson 2008 mean runs slightly higher (~37). 32 used to stay consistent with the whole-milk row's sourcing — removing fat barely moves milk's GI. Lactose-driven; low regardless of exact value.

USDA FoodData Central, SR Legacy 171257 "Cream, sour, cultured": per 100 g, 4.63 g carbohydrate by difference, of which 3.41 g sugars (lactose) and 0 g starch, 2.44 g protein, 0 g fibre, 19.4 g fat. Scaled to a 30 g (2 tbsp) serving: 1.4 g available carbohydrate, 0.7 g protein, 0 g fibre. Cross-checked against USDA Foundation Foods 2346387 "Cream, sour, full fat" (5.56 g carbohydrate, 3.07 g protein, 18.0 g fat per 100 g), which is consistent. No published GI exists: a standard test dose of 25-50 g available carbohydrate would require roughly 540-1080 g of sour cream, so it has never been tested and appears in neither Foster-Powell et al. 2002 nor Atkinson et al. 2021. Recorded as n/a rather than 0 because the sugar present is the same lactose as in milk (GI around 32), so the true contribution is very small and slow rather than literally none — copy must not claim sour cream is "zero carb".

Cited consistently at ~54 across bread-GI comparison sources; ~20 points lower than standard white bread (75) due to fermentation slowing starch digestion. The 'sourdough is much healthier' framing is directionally correct but the gap is moderate, not dramatic.

University of Sydney GI database (glycemicindex.com) — soy milk ~34 (cited range ~17-44 across formulations). GI tests used sweetened products; UNSWEETENED soy milk has so little carbohydrate (~1-4g/cup) that GL is near zero and the GI figure barely applies.

USDA FoodData Central / non-carbohydrate beverage; GI is undefined for herbal infusions without available carbohydrate.

GI = 63: Atkinson FS, Brand-Miller JC, Foster-Powell K, Buyken AE, Goletzke J, "International tables of glycemic index and glycemic load values 2021", Am J Clin Nutr 114:1625-1632, Supplemental Table 1 (ISO 26642:2010-consistent), entry 714 "Spelt, pearled (farro)", Italy 2009, GI 63±10, 9 healthy subjects, glucose reference. Corroborated twice: the same table's "Spelt bread, mean of four foods" = 61, and Marques C, D'auria L, Cani PD et al., "Comparison of glycemic index of spelt and wheat bread in human volunteers", Food Chemistry 2007;100:1265-1271, which measured white spelt bread at GI 93±9 against wheat white bread as the 100 reference — no significant difference from wheat, and roughly 65 on the glucose scale. Composition: USDA FoodData Central, SR Legacy FDC 169746 (Spelt, cooked) — 26.4 g carbohydrate, 3.9 g fibre, 5.5 g protein per 100 g; 150 g gives 33.7 g available carbohydrate, 5.9 g fibre and 8.3 g protein. GL = 63 x 33.7/100 = 21.2, rounded to 21. CAVEAT: the USDA entry does not state whether it is pearled or wholegrain spelt while the GI entry is specifically pearled, so the fibre figure may run high for the exact form that was GI-tested; also "farro" is sometimes emmer rather than true spelt, though the source table classifies this entry as spelt.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

USDA FoodData Central, SR Legacy 170495 "Seaweed, spirulina, dried": per 100 g, 57.5 g protein, 23.9 g carbohydrate by difference, 3.6 g fibre, 3.1 g sugars, 7.72 g fat, 28.5 mg iron. USDA portion "tablespoon" = 7 g, used as the serving. Available carbohydrate = (23.9 - 3.6) x 0.07 = 1.4 g; fibre 0.3 g; protein 4.0 g. GI recorded as n/a — spirulina appears in neither Foster-Powell et al. 2002 nor Atkinson et al. 2021, and 1.4 g of available carbohydrate per serving is far below a testable dose. CAVEATS: USDA's entry is a single composite for dried spirulina and commercial products vary; and the metabolic evidence base is not PCOS-specific — published human trials on lipids and insulin sensitivity are in other populations, while the PCOS-labelled comparison against metformin is a rat model, so generated copy must not imply spirulina has been trialled in PCOS.

USDA FoodData Central, 'Mollusks, squid, mixed species, raw' (FDC ID 174223): 15.6g protein, 3.08g carbohydrate, 0g fiber per 100g; USDA lists 3 oz as 85g. Cross-checked against Health Canada Canadian Nutrient File code 3125 ('Mollusks, squid (calamari), mixed species, raw': 15.58g protein, 3.08g carbohydrate per 100g). GI recorded as n/a per the non-carbohydrate-food convention used in this table — squid has never been GI-tested and its small carbohydrate fraction is stored glycogen, not starch or sugar. Caveat: these figures describe plain squid only. Battered fried calamari (USDA 'Mollusks, squid, mixed species, cooked, fried', FDC ID 171982) adds a wheat-flour coating and is not covered by this row.

Health Canada Canadian Nutrient File code 6144 (Beef, loin, top sirloin steak, boneless, lean and fat, 0mm trim, cooked, broiled): 28.79g protein, 0g carbohydrate, 0g fibre per 100g, scaled to a 100g cooked portion. GI is not assigned to foods without meaningful available carbohydrate in the University of Sydney GI database or any published GI table; 0 is recorded here by the stated convention and means 'not applicable' rather than a measured zero. Note: cut and trim level change the fat and energy density considerably — top sirloin is used as a mid-range representative — but the carbohydrate figure stays at zero across every cut, so the glycemic fields hold for the whole category.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

University of Sydney GI database (glycemicindex.com) — single published study (40±7), so less-replicated than staple foods; Sydney lists GL 1 per 120g serving. Low available carbohydrate keeps GL trivially low at any realistic portion.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Atkinson FS, Brand-Miller JC, Foster-Powell K, Buyken AE, Goletzke J. International tables of glycemic index and glycemic load values 2021 (Am J Clin Nutr), Supplemental Table 1, entry 673: 'Japanese Style Sushi rice, white', a commercial Australian product, Australia 2005, GI 85+/-10, n=10, 50g available-carb test portion, ISO 26642:2010-consistent. CAVEAT: single study with a wide SD (+/-10) on one product. Composition from USDA FoodData Central SR Legacy 168882 (Rice, white, short-grain, enriched, cooked; 28.73g carb, 2.36g protein per 100g); fiber taken from SR Legacy 169753 (long-grain cooked, 0.4g/100g) because the short-grain cooked entry reports no fiber value. IMPORTANT SCOPE NOTE: these figures are for PLAIN cooked sushi-style rice; the vinegar-sugar-salt seasoning adds a small and highly variable amount of sugar that is not modeled here. Entries 2049 and 2050 of the same table give ASSEMBLED sushi at GI 45+/-3 (brown rice, vegetable and tofu) and 48+/-8 (salmon) — the rice and the finished roll must not be conflated in generated copy. Generated copy must not name any manufacturer or product brand.

Boiled sweet potato range ~46 (30min boil) to ~61 (8min boil) across cited sources; 54 is the midpoint. Baking is a materially different, much higher-GI preparation (~94) — NOT covered by this row; do not conflate boiled and baked verdicts.

HEAVILY prep-dependent, sparse direct testing: fried sweet potato measured at 76±7 (Jamaican cultivar GI study, PMC3205609); boiling runs ~41-61 and baking ~82-94, so 76 covers the fried/fries preparation only — do not reuse the boiled sweet-potato verdict (54) for fries. Fiber/protein are representative for frozen oven-baked fries.

Health Canada Canadian Nutrient File code 47 (Cheese, swiss/emmental): 26.93g protein, 0g fibre and 5.38g carbohydrate-by-difference per 100g, of which only 1.32g is sugar — aged Swiss retains almost no lactose, and the by-difference figure overstates its true available carbohydrate. Scaled to a 30g slice: 8.1g protein, 0g fibre, at most ~1.6g carbohydrate. GI is not assigned to foods without meaningful available carbohydrate in the University of Sydney GI database or in any published GI table; 0 is recorded here by the stated convention and means 'not applicable' rather than a measured zero. Generated copy must not imply Swiss cheese was GI-tested.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

USDA FoodData Central SR Legacy 175177 (Fish, tilapia, cooked, dry heat): 0g carbohydrate, 0g fiber, 26.15g protein, 2.65g fat per 100g. Serving from the same record's USDA food_portion entry: 1 fillet = 87g. GI reported as 'n/a' rather than 0: tilapia contains zero available carbohydrate and appears nowhere in the 2021 International Tables (Atkinson et al.), consistent with the standard convention — and with the existing convention in this CSV for eggs, coffee and cheese — that GI is only determined for foods with meaningful available carbohydrate. A GI of 0 would falsely imply the food was tested and produced no glycemic response; GI is undefined here, not zero.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Foster-Powell/Atkinson international tables, corn chips 63±10 (individual studies ~42-74) — brand, frying oil, and masa processing move the number, so treat 63 as a category middle rather than a per-brand figure.

USDA FoodData Central SR Legacy #173718 (fish, trout, rainbow, farmed, cooked, dry heat): 0g carbohydrate, 0g fiber, 23.8g protein, 7.38g fat per 100g. GI recorded as n/a rather than 0 — trout has no appreciable available carbohydrate and is absent from both the University of Sydney GI database and Foster-Powell et al. 2002. GI is undefined for this food, not measured as zero, and no GI or GL figure should appear in generated copy. Farmed vs wild comparison verified directly: wild rainbow trout (SR Legacy #175155) is 22.9g protein and 5.82g fat per 100g, so the farmed fish is the fattier of the two; mixed-species trout (#172004) is 26.6g protein and 8.47g fat.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

USDA FoodData Central SR Legacy, "Alcoholic beverage, distilled, vodka, 80 proof" (FDC 174818): 0 g carbohydrate, 0 g fiber, 0 g protein; 1.5 fl oz = 42 g per the USDA portion table. GI recorded as n/a: there is no available carbohydrate to measure, and vodka appears nowhere in the Atkinson/Foster-Powell/Brand-Miller international tables (2008, Diabetes Care 31:2281-2283). Important mechanistic point for the generator: ethanol's effect on blood glucose is not a glycemic-index phenomenon at all — it acts through suppressed hepatic gluconeogenesis, so a GI value could not describe it even in principle, and n/a should not be read as 'no metabolic effect'.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

University of Sydney GI database (glycemicindex.com) — watermelon 76, the classic high-GI/low-GL food: only ~8-11g available carbohydrate per cup keeps GL at ~4-6 for a realistic serving (Sydney lists GL 4 per 120g). Verdict copy should lead with GL, not the scary-looking GI.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

Universal standard reference value for sucrose — one of the most consistently-cited GI figures across every published table, effectively uncontested.

University of Sydney GI database (glycemicindex.com) — standard reference value, one of the most consistently-cited GI figures in the literature.

Foster-Powell 2002 international tables — white chocolate (Milky Bar, Nestle) 44±6; single-brand measurement, so treat as representative rather than universal. Fat blunts the glucose response, but this is a sugar-dense food: low-ish GI should not read as an endorsement in verdict copy.

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).

University of Sydney GI database (glycemicindex.com) — boiled white rice, standard reference value cited consistently across GI literature.

Range 27-38 across cited sources (Sydney-style databases and published dairy GI tables); 32 used as conservative middle. Lactose-driven; low regardless of exact value.

Cited at ~69 in bread-GI comparison sources; meaningfully higher than sourdough (54) and only modestly lower than white bread (75) — the 'wheat bread is automatically much better' assumption overstates the real gap.

Commonly cited at 57 across GI databases, though some sources list ~45 and direct published testing is thinner than for true rices (wild rice is botanically a grass seed) — 57 used as the conservative-higher representative; flag as less-replicated than white/brown rice figures.

GI 13 (insulinaemic index 11, glucose = 100) — Livesey G, "Health potential of polyols as sugar replacers, with emphasis on low glycaemic properties", Nutrition Research Reviews 2003;16(2):163-191, which re-analysed the published glycaemia/insulinaemia literature and expressed it in GI/II terms (erythritol 0/2; xylitol 13/11; sorbitol 9/11; maltitol 35/27; versus sucrose 65/43). DISPUTED VALUE: the GI of 7 circulated widely online is not traceable to a peer-reviewed source, so Livesey's 13 is used; a value of 7 would give GL 1 rather than 2. Xylitol is absent from the University of Sydney international tables (Atkinson 2008, Diabetes Care 31:2281-2283; Atkinson 2021, AJCN Suppl. Tables 1-2), so no ISO 26642:2010 test exists for it. Serving and carbohydrate load from USDA FoodData Central Branded 392084 (granulated xylitol): serving size 4.0 g = 1 tsp, 100 g carbohydrate per 100 g, 0 g fiber, 0 g protein — so 1 tbsp = 12 g and the full 12 g is the declared carbohydrate load. GL = 13 x 12/100 = 1.6, rounded to 2. Contested safety context (not a GI matter): Witkowski M et al., "Xylitol is prothrombotic and associated with cardiovascular risk", European Heart Journal 2024;45(27):2439-2452, with published correspondence disputing the model and interpretation in the same journal (Eur Heart J 2025;46(3):326 and 46(3):328; 46(27):2705 and 46(27):2707).

Not a significant carbohydrate source — GI is not assigned to non-carbohydrate foods per standard convention (University of Sydney GI database and all major published GI tables only test/list foods with meaningful available carbohydrate).