Low-glycemic diet

• Lower HbA1c and fasting glucose in people with diabetes or prediabetes.
• Reduce large rises in blood sugar after meals.
• Lower the risk of developing type 2 diabetes.
• Improve blood lipids and lower cardiovascular risk.
• Help with appetite control and keeping weight off.

The glycemic index. In 1981, David Jenkins and colleagues fed 62 common foods to small groups of healthy volunteers and measured blood glucose for 2 hours. They expressed each food's response as a percentage of the response to the same amount of carbohydrate taken as glucose [13]. Dried legumes (31%) and dairy foods (35%) raised blood sugar least. Fat and protein in a food blunted the rise; fiber and sugar content, on their own, did not predict it.

Cutoffs. The University of Sydney's GI group, which keeps the international GI database, classes foods as low GI at 55 or less and high GI at 70 or more. The 2021 international tables list more than 4,000 tested items [14].

Glycemic load. GI compares equal amounts of carbohydrate, not normal portions. Glycemic load fixes that: GI multiplied by the grams of available carbohydrate in a serving, divided by 100. A GL of 10 or less is low, 11 to 19 intermediate and 20 or more high [15]. Watermelon and a doughnut have similar GIs, but a serving of watermelon has a GL of about 8 and a medium doughnut about 17.

What it is not. A low-glycemic diet is not a low-carbohydrate diet. It changes the type of carbohydrate, not necessarily the amount. It is also not a calorie plan: fat lowers a food's glucose response, so some high-fat processed foods score low.

The 2021 international tables found that dairy products, legumes, pasta and fruit were usually low GI and gave consistent values worldwide. Breads, breakfast cereals and rice came in high, medium and low GI versions, and whole-meal products were not reliably low, presumably because of how they were made. Most potatoes were high GI, although a few low GI varieties now exist [14].

What changes a food's GI?

• Processing. Cereal products, including whole-grain ones, vary widely in GI, presumably because of how they are made [14].
• Ripeness. Diabetes UK lists overripe bananas as medium GI, while most other whole fruit is low.
• The rest of the meal. Fat and protein lower the glucose response [13], which is why AumniVIDA pairs carbohydrate with protein, fat or fiber rather than removing it.
• Portion size. A low GI food in a large portion can still carry a high glycemic load [15].

AumniVIDA's glucose stability targets

• Fiber at least 14 g per 1,000 kcal and whole grains at least 90 g a day.
• Added sugar at most 5% of energy, and no sugar-sweetened drinks.
• At least 20 g of protein and at most 60 g of carbohydrate per meal, judged across most meals rather than every one.
• Last meal at least 3 hours before sleep.

The app estimates glycemic load from logged meals and labels it as an estimate, because published GI values vary by brand, ripeness and preparation.

In people with diabetes. A meta-analysis of 29 trial comparisons in 1,617 people with type 1 or type 2 diabetes found that low GI or GL diets lowered HbA1c by 0.31 percentage points versus higher GI diets, with high-certainty evidence, plus small drops in fasting glucose [1]. An earlier Cochrane review of 11 trials found a drop of 0.5 points and fewer hypoglycemia episodes in one trial [2]. In the largest single trial, 210 adults with type 2 diabetes lowered HbA1c by 0.50 points on low GI advice versus 0.18 on high cereal fiber advice over 6 months [3]. These are small but clinically meaningful changes; see HbA1c.

For prevention. In 65,173 US women, the highest glycemic load fifth had 47% higher risk of type 2 diabetes, and high glycemic load combined with low cereal fiber more than doubled risk [4]. The Lancet's 2019 series of reviews put this in context: fiber and whole grains were linked to 15 to 30% lower risk of diabetes, heart disease and death, while low GI or GL diets showed smaller or no reductions, with low to very low certainty [5]. Sugar-sweetened drinks are the clearest single target: each daily serving was linked to 13% more diabetes after adjusting for body fat [16].

The practical message is that fiber, whole grains and fewer sugary drinks carry most of the benefit, and GI helps choose between carbohydrate foods.

Losing weight. The 2023 Cochrane review included 10 trials in 1,210 people. Compared with higher GI diets, low GI or GL diets changed weight by 0.82 kg, with a confidence interval that crossed zero, which the reviewers judged as probably little or no difference [8]. All trials had unclear or high risk of bias, and none reported on mortality.

Keeping weight off. The European Diogenes trial first had 773 adults lose an average 11.0 kg on an 800 kcal diet, then assigned them to one of five maintenance diets for 26 weeks. Low GI groups regained 0.95 kg less than high GI groups, and high protein groups 0.93 kg less than low protein groups. Fewer people dropped out of the low GI and high protein arms [9]. That points toward a combination of more protein and lower GI, which overlaps with the high-protein diet.

In diabetes. Low GI diets lowered body weight by 0.66 kg on average [1].

The mechanism people cite. High GI meals cause faster rises and falls in glucose and insulin, which some researchers argue drive hunger and fat storage [17]. In rats kept at the same body weight, high GI starch produced 71% more body fat [10]. Human trials have not shown an effect of that size.

Risk factors in diabetes. The 2021 BMJ meta-analysis found reductions in LDL cholesterol (0.17 mmol/L), non-HDL cholesterol (0.20 mmol/L), ApoB (0.05 g/L) and triglycerides (0.09 mmol/L), with no change in HDL cholesterol or diastolic blood pressure [1]. See LDL cholesterol, ApoB and triglycerides.

Events. In the PURE cohort of 137,851 adults on five continents, the highest GI diets were linked to more major cardiovascular events or death than the lowest: HR 1.51 in people with heart disease and 1.21 in people without it [6]. For glycemic load, the link was not significant in people without heart disease.

The counterweight. OmniCarb fed 163 adults four complete DASH-style diets for 5 weeks each. On a high carbohydrate diet, lowering GI from 65 to 40 reduced insulin sensitivity by 20% and raised LDL cholesterol from 139 to 147 mg/dL; it did not change HDL, triglycerides or blood pressure [7]. The authors concluded that within an already healthy diet, choosing foods by GI may not improve cardiovascular risk factors. Lowering total carbohydrate did lower triglycerides.

Why the results differ. OmniCarb tested GI alone in an already healthy diet; the cohorts and many diabetes trials changed whole patterns, adding legumes, intact grains and fiber. The Lancet reviews found that fiber and whole grains, more than GI itself, carried the clearest benefit [5].

The personalized nutrition finding. Researchers at the Weizmann Institute tracked 46,898 meals in 800 adults with continuous glucose monitors and found large differences between people eating identical meals. An algorithm using blood tests, habits, body measures and gut microbiome predicted individual responses, and diets built from it lowered post-meal glucose in a small blinded trial [18]. In the UK PREDICT 1 study of 1,002 adults, glucose responses to identical meals varied with a coefficient of variation of 68% [19]. Even people with normal standard tests spent 15% of the day in the prediabetic range on a monitor [20].

Does personalization improve outcomes? In 225 adults with prediabetes, an algorithm-personalized diet cut time above 140 mg/dL (7.8 mmol/L)7.8 mmol/L (140 mg/dL) by 1.3 hours a day versus 0.3 hours on a Mediterranean diet, and HbA1c by 0.16 versus 0.08 percentage points over 6 months [21]. The difference was statistically significant and small.

The limits.

• Noise within a person. When the same white bread was tested three times, GI values were 78, 60 and 75, and variation within each person (42.8%) exceeded variation between people (17.8%) [22]. A 2026 reanalysis of 382 adults found that about 90% of individual responses were predicted by each person's response to glucose scaled by the food's GI, within normal day-to-day variation [23].
• Meal composition still matters most for glucose. In PREDICT 1, a meal's macronutrients explained more glucose variation (15.4%) than the gut microbiome (6.0%) [19].
• Device error. In 15 healthy adults, a continuous glucose monitor read about 0.9 mmol/L higher than finger-prick samples and overestimated time above 7.8 mmol/L about fourfold [12].
• Stable traits. Repeated n-of-1 tests in 176 young adults did find a personal glycemic sensitivity that stayed stable over 2 years [24], so some people really do run higher.

A monitor can show which meals raise your glucose, but a single spike is not a diagnosis, and HbA1c and fasting glucose remain the outcomes that matter.

In OmniCarb, the most informative contrast was the low GI, low carbohydrate diet (40% of energy from carbohydrate) against the high GI, high carbohydrate diet (58%). It lowered triglycerides from 111 to 86 mg/dL (23%) but did not change insulin sensitivity, LDL cholesterol, HDL cholesterol or blood pressure [7]. Within the low carbohydrate diets, switching to low GI lowered triglycerides by only 5%.

• The ketogenic diet removes most starchy foods, fruit and legumes, which makes GI largely irrelevant and lowers glucose spikes by removing carbohydrate.
• The Mediterranean diet is naturally low to moderate GI because it relies on legumes, vegetables, fruit and pasta, and it served as the comparison diet in the personalized nutrition trial [21].
• The anti-inflammatory diet shares the emphasis on whole grains, legumes and plant variety.

Acne. Young men on a low glycemic load diet had 23.5 fewer acne lesions after 12 weeks versus 12.0 fewer on a control diet. They also lost 2.9 kg, so the authors could not separate diet from weight loss [11]. The trial included only males aged 15 to 25.

PCOS and fertility. Low GI diets are often recommended for polycystic ovary syndrome because insulin resistance is common in it. We did not find trials strong enough to grade here, so this page leaves PCOS ungraded.

Energy, mood and cravings. Popular glucose programs promise better energy, fewer cravings and less inflammation by flattening glucose spikes. The 2023 Cochrane review found one small trial on mood, with very uncertain results [8]. Critics such as the diabetes dietitian Nicola Guess argue that consumer monitor readings mislead healthy people, and device studies support caution [12].

Animal results. In rodents, high GI starch caused more body fat and damaged insulin-producing cells at equal body weight [10]. These experiments support the mechanism [17] but cannot grade an effect in people.

• Choosing high-fat processed snacks just because fat lowers their GI.
• Ignoring portion size: a large portion of a low GI food can carry a high glycemic load.
• Treating every glucose spike on a monitor as harmful, or changing diet based on a few readings.
• Assuming whole-meal bread or cereal is low GI; many are not.
• Cutting fruit, legumes or whole grains to avoid any rise in glucose, which loses fiber.
• Changing glucose-lowering medication without your clinician when readings improve.

• You take insulin or sulfonylureas: better glucose control can raise hypoglycemia risk, so ask your clinician how to adjust doses.
• You have type 1 diabetes: GI changes affect insulin dosing; plan changes with your diabetes team.
• You have kidney disease with a potassium or phosphorus limit: check legume, dairy and fruit portions with a dietitian.
• You have a history of eating disorders: strict food rating and glucose tracking can feed restriction; get support first.
• You are pregnant or have gestational diabetes: follow the plan from your care team.

• Gas and bloating when legumes and fiber increase quickly; build up over a few weeks.
• Food anxiety or overly restrictive eating from chasing flat glucose curves.
• Hypoglycemia in people on insulin or sulfonylureas if doses are not adjusted.
• Weight gain if low GI foods high in fat replace lower calorie foods.