Glycemic index ranks carbohydrate foods by how much they raise glucose, and glycemic load adds the portion. Here is how both are measured, what shifts them and what the evidence shows for people with diabetes.
Eaten in portions with the same amount of carbohydrate, legumes and potatoes raise glucose very differently. In the 2021 international tables of tested foods, legumes averaged a glycemic index (GI) of 34 and potatoes 71 1. GI compares the two-hour glucose rise after a carbohydrate food with the rise after the same amount of pure glucose. Glycemic load (GL) adds how much you eat.
A published GI is an average from healthy volunteers eating one food on its own. Your response to a real meal also depends on the portion, the rest of the plate, your physiology and any insulin you take. Here is how GI and GL are measured, what changes them, what trials and guidelines say, and where the numbers fall short.
GI was first published in 1981 2. The method is now set out in an international standard, ISO 26642:2010, which also recommends criteria for classing foods as low, medium or high GI and was last confirmed in 2021 3.
Each volunteer eats a portion of the food containing 50 g of available (digestible) carbohydrate, or 25 g for bulky foods with little carbohydrate 1. Blood glucose is measured while fasting and at 15, 30, 45, 60, 90 and 120 minutes 1, 4. In one UK laboratory’s protocol, the area under each person’s curve above their fasting level was expressed as a percentage of their average response to glucose, taken three times, and the food’s GI was the group average 4.
The reference is glucose, scored as 100. The 2021 tables convert values measured against white bread to this scale by multiplying them by 0.71 1. For the more reliable list in the 2021 tables, each value had to come from at least 10 healthy adults aged 18 to 65 with normal glucose tolerance 1.
On the glucose scale, a GI of 55 or less is low, 56 to 69 is medium and 70 or more is high. The 2021 tables take these cut-offs from the ISO standard 1, 3. GI applies only to foods rich in carbohydrate 2.
GI compares foods gram for gram of carbohydrate, so it ignores how much you eat. GL multiplies a food’s GI by the grams of available carbohydrate in a portion and divides by 100 1, 4.
White rice averaged a GI of 73 in the 2021 tables 1. A bowl holding 45 g of available carbohydrate would have a GL of about 33, half that bowl about 16 and a quarter about 8. Pasta averaged 52, yet a large plate with 60 g of available carbohydrate reaches a GL of about 31.
GL cut-offs are less formal. A widely used scheme rates a serving’s GL as low at 10 or less, medium at 11 to 19 and high at 20 or more 4. A 2007 UK study that applied it credits the scheme to a 2003 book co-written by Jennie Brand-Miller, whose University of Sydney group compiles the international tables 1, 4. The ISO standard’s classification criteria are for GI 3.
Carbohydrate that is digested and absorbed quickly gives a high GI, and carbohydrate handled slowly gives a low one 2. The International Society for Pediatric and Adolescent Diabetes (ISPAD) lists cooking and preparation, a food’s physical form, the type of starch, and the fat and protein eaten with it 5. The 2015 international GI consensus adds that less gelatinized starch, more viscous fibre and more fructose all lower GI 2.
In the 2021 tables, potatoes averaged 84 as instant mash, 79 as regular mash, 73 boiled and 49 when cooked and refrigerated overnight 1. Cereal products, whole-grain versions included, varied widely, which the authors suggested came from differences in manufacturing 1.
Particle size counts even among whole grains. At the University of Otago in New Zealand, 15 adults with type 2 diabetes (average A1c 7.5%, or 58 mmol/mol) ate four nutrient-matched breads made entirely from whole-grain wheat, in random order 6. Across three breads made with roller-milled flour and added grains, the largest glucose response was about 1.7 times the smallest, and responses fell as the grain particles got larger 6.
Ripeness changes fruit. In 10 people with type 2 diabetes, over-ripe banana produced a four-hour glucose response about 70% larger than the same weight of under-ripe banana, which the authors linked to the under-ripe fruit’s higher starch content 7.
As cooked starch cools, some of its chains pack into tight double helices that digestive enzymes struggle to break down in the small intestine. This fraction is called resistant starch 8.
At Poznan University of Medical Sciences in Poland, 32 adults with type 1 diabetes using insulin pumps ate a rice meal with 50 g of carbohydrate twice, in random order 8. One serving was freshly cooked. The other had been refrigerated for 24 hours and reheated, which raised its resistant starch from 7.5 to 12.0 g per 100 g 8. With the cooled rice, glucose peaked at 9.9 mmol/L (178 mg/dL) instead of 11.0 mmol/L (198 mg/dL), rose by 2.7 instead of 3.9 mmol/L (49 vs 70 mg/dL), and the area under the curve was about 60% smaller 8. It also brought more lows (see the insulin section).
Viscous, gel-forming fibre blunts the rise. In a 1978 experiment, guar gum added to a 50 g glucose load flattened the response more than the other fibres tested. Guar broken down so that it no longer thickened lost the effect, and the more viscous a fibre, the larger the drop in peak glucose 9.
Acidity works in the same direction. In 10 healthy volunteers, white bread eaten with vinegar raised glucose about a third less than the same bread without it, and insulin rose less too, which the authors suggested was due to slower stomach emptying 10.
Fat and protein eaten with a carbohydrate food lower its glucose response, and so its GI, by 25% to 50%, with high-GI foods falling more than low-GI ones 1. The American Diabetes Association (ADA) notes that fat and protein affect both early and delayed post-meal glucose, apparently in proportion to the amount 11.
A 2021 meta-analysis by researchers at the University of Toronto and St Michael’s Hospital, done to inform the nutrition guidelines of the European Association for the Study of Diabetes (EASD), pooled randomized controlled trials lasting three weeks or more 12. It found 29 trial comparisons with 1,617 people with type 1 or type 2 diabetes. Most were middle-aged, had overweight or obesity and had moderately controlled type 2 diabetes treated with drugs or insulin 12. Trials lasted a median of 12 weeks (range 3 to 52), and the low-GI diets had a median GI of 49 against 63 for the comparison diets 12.
Compared with higher-GI diets, low-GI or low-GL diets lowered A1c by 0.31 percentage points (about 3 mmol/mol) on top of usual treatment 12. The authors rated this finding as high certainty, though results varied a lot between trials, and noted that it meets the 0.3-point threshold the European Medicines Agency proposes as clinically relevant 12. Fasting glucose fell by 0.36 mmol/L (6 mg/dL), a change they judged trivial 12. LDL cholesterol, triglycerides, body weight (by about 0.7 kg) and several other risk markers also fell slightly 12. Results were similar in type 1 and type 2 diabetes, although only three comparisons, with 165 people, involved type 1 12.
One included trial, at St Michael’s Hospital in Toronto, randomized 210 people with type 2 diabetes on glucose-lowering medication to six months of low-GI dietary advice or advice to eat more cereal fibre 13. A1c fell by 0.50 percentage points with low GI and by 0.18 with high cereal fibre 13.
Better glucose can bring lows if medication stays the same. In that trial, 6 of 106 people on the low-GI diet and none of 104 on the comparison diet had hypoglycemic symptoms or low readings and needed their glucose-lowering drugs reduced 12. Over 24 to 52 weeks, though, Diabetes Canada cites evidence that low-GI diets reduced hypoglycemic events in adults and children with type 1 diabetes 14.
In plain terms: Swapping higher-GI carbohydrate foods for lower-GI ones lowered A1c by about 0.3 percentage points in trials, on top of usual treatment. That is a modest effect, and it came as part of a whole eating pattern.
Diabetes Canada’s 2018 nutrition therapy guideline recommends that adults with diabetes choose low-GI carbohydrate foods to help optimize glycemic control (grade B, level 2 evidence, for both type 1 and type 2 diabetes), with lower-grade evidence for benefits to LDL cholesterol and cardiovascular risk 14. It gives legumes, whole grains, fruit and vegetables as examples of low-GI foods 14.
ISPAD’s 2022 nutrition guideline for children and teens states that using GI adds benefit for glucose management beyond counting total carbohydrate alone (grade B) 5.
The EASD’s 2023 European recommendations state that a low-GI or low-GL diet can improve glucose and intermediate cardiometabolic risk factors in people with diabetes. They add that markers of carbohydrate quality beyond fibre may also guide management, provided advice on added sugars, sodium and saturated fat is also met 15.
The ADA’s 2026 Standards of Care are more cautious. They call the research on GI and GL in diabetes complex, with differing definitions of low and high GI and mixed effects on fasting glucose and A1c. One systematic review found no significant effect on A1c, while others found reductions of 0.15 to 0.5 percentage points 11. The ADA notes that low-GI eating goes hand in hand with higher-fibre eating, and recommends emphasizing minimally processed, nutrient-dense, high-fibre carbohydrate, with at least 14 g of fibre per 1,000 kcal (grade B) 11.
A published GI is an average, and individual tests scatter widely around it. At Tufts University, researchers measured the GI of white bread three times in each of 63 adults free of chronic disease. Results varied by about 20% between repeat tests in the same person and 25% between people, and adding people, repeats or longer sampling did not reduce the spread 16. The volunteers’ insulin responses and A1c explained 15% and 16% of the variation, which the authors took as a sign that glycemic status shapes the result 16.
The researchers behind the 2015 international GI consensus, whose listed funders included several food companies and GI testing services, state that a correctly used method separates high-GI from low-GI foods with 95% certainty. They also accept that published values for the same food can vary so widely, for some foods, that you cannot know the GI of the one on your plate 2.
People also respond differently to identical meals. At the Weizmann Institute of Science in Israel, a week of continuous glucose monitoring in 800 people, covering 46,898 meals, showed wide variation between people, which the authors said limits the value of universal dietary advice 17.
GI describes single foods, and the consensus group advises against measuring the GI of mixed meals. A meal’s GI can be calculated from its parts, but that figure is not necessarily expected to predict the glucose response, because the amounts and types of fat, protein and carbohydrate also matter 2.
A low GI does not make a food nutritious. Fat and fructose both lower GI 1, 2, and savoury snack foods averaged 60 in the 2021 tables, below breads at 64 1. The consensus group says low-GI and low-GL diets should always be judged within an otherwise healthy diet, alongside fibre and whole grains 2.
In a trial by researchers at Brigham and Women’s Hospital, Harvard and Johns Hopkins, 163 overweight adults with systolic blood pressure of 120 to 159 mm Hg ate four fully provided DASH-style diets for five weeks each, in random order 18. On the high-carbohydrate diets, switching from high-GI (65) to low-GI (40) carbohydrate lowered insulin sensitivity by 20% and raised LDL cholesterol by 6%, with no change in HDL cholesterol, triglycerides or blood pressure 18. The authors concluded that within an overall DASH-type diet, choosing foods by GI may not improve cardiovascular risk factors or insulin resistance 18. Volunteers were enrolled on the basis of weight and blood pressure, so the results may not apply to people with diabetes.
The 2021 tables, compiled at the University of Sydney, list more than 4,000 items, 61% more than the 2008 edition 1. About 2,100 values meet the ISO method and form the more reliable list. About 1,900 come from less robust tests, such as small groups or wide error margins 1.
The table summarizes the reliable list for common food groups, from Table 1 of the 2021 tables, with 32 (legumes) to 214 (breads) products per group 1. Individual products, brands and tests vary widely around these averages.
| Food group | Average GI | Tested low (≤55) | Tested high (≥70) |
|---|---|---|---|
| Legumes | 34 | 94% | 0% |
| Dairy products | 35 | 95% | 0% |
| Fruit | 51 | 72% | 6% |
| Pasta | 52 | 64% | 8% |
| Savoury snacks | 60 | 46% | 34% |
| Breakfast cereals | 61 | 37% | 30% |
| Breads | 64 | 29% | 35% |
| Rice | 67 | 28% | 38% |
| Potatoes | 71 | 14% | 58% |
Rice products ranged from 19 to 116, potatoes from 35 to 103 and breads from 24 to 100 1. Averages also differed by region. Rice averaged 74 in Asia, 65 in Australia, 61 in Europe and 60 in North America, and potatoes 77 in Australia, 73 in Europe and 67 in North America 1.
The authors caution that the listed foods are those laboratories tested for research or commercial reasons, which may not represent what is sold where you live. Assigning a GI requires local values, ideally brand-specific ones, because manufacturers process foods, especially cereal products, in different ways 1. Where a food was tested several times, the tables list a mean under the individual results, so you can see the spread 1. That makes the tables better for comparing options within a group, such as two breads, than for predicting your own response. Dairy products, legumes, pasta and fruit usually test low, with consistent values around the world 1.
ISPAD’s 2022 guideline notes that the timing and type of insulin delivery may be adjusted depending on a food’s GI. It says earlier insulin with high-GI foods may blunt the spike after eating, and a combination-type bolus may help with lower-GI foods 5. It also reports that a bolus 15 to 20 minutes before eating, instead of immediately before, improves post-meal glucose 5.
The Poznan rice study shows the other side. Participants dosed for both meals the same way, with their pumps’ bolus calculators and their own settings, 10 minutes before eating, and the doses did not differ 8. Within three hours there were 12 episodes of glucose below 3.9 mmol/L (70 mg/dL) after the cooled rice, against 3 after the fresh rice 8. The authors concluded that cooling rice reduces the rise but raises the risk of lows with a standard insulin dose 8.
The ADA recommends education on how carbohydrate, fat and protein affect glucose, tailored to each person’s needs and insulin plan, to optimize mealtime dosing 11. For people on fixed insulin doses, it recommends consistent carbohydrate intake in timing and amount 11. A switch to lower-GI foods can change that balance. Changes to insulin timing, bolus type or dose are decisions to make with your care team and to check against your own data.
Glycemic index (GI) compares the two-hour glucose rise after a carbohydrate food with the rise after the same amount of pure glucose. Because it compares foods gram for gram of carbohydrate, it ignores how much you eat. Glycemic load (GL) adds the portion.
On the glucose scale, a GI of 55 or less is low, 56 to 69 is medium and 70 or more is high 1, 3. A widely used scheme rates a serving's glycemic load as low at 10 or less, medium at 11 to 19 and high at 20 or more 4.
Multiply a food's glycemic index (GI) by the grams of available carbohydrate in a portion and divide by 100 1, 4. White rice averaged a GI of 73 1. A bowl with 45 g of available carbohydrate would have a glycemic load of about 33, and half that bowl about 16.
In a 2021 meta-analysis of 29 trial comparisons with 1,617 people with diabetes, low-GI or low-GL diets lowered A1c by 0.31 percentage points more than higher-GI diets, on top of usual treatment 12. That is a modest effect. Better glucose can bring lows if medication stays the same.
A published GI is an average, and individual tests scatter widely around it. When Tufts researchers measured white bread's GI three times in each of 63 adults free of chronic disease, results varied by about 20% between repeat tests in the same person and 25% between people 16.
Your own data are a better guide than a table to how a food behaves for you. Pick two versions of something you eat often, such as two breads or fresh and cooled rice, in portions with the same carbohydrate, and log each a few times. In Glyve, Meal Scan photographs a meal, a nutrition label or a barcode and estimates carbs, protein, fat and fibre, and Glyve shows a glycemic index estimate. Every estimate is a draft you review and edit before saving. The History graph shows meals on the same timeline as glucose, with your target range shaded, so you can compare the first two hours and look for later lows or rises. Log insulin too. Glyve shows what happened together; it doesn’t tell you what caused it. Bring what you find to your care team.
How we write about diabetes. Spotted a mistake? Email info@glyve.ca and we’ll check it against the source.
Glyve keeps glucose, food, insulin, exercise, sleep and stress on one timeline.
For iPhone, iPad and Apple Watch.