What Is the Glycemic Index? GI, GL and the Insulin Index

How GI is measured

Researchers give about ten people a portion of a single food containing 50g of available carbohydrate, measure blood sugar several times over two hours and plot the area under the curve. They compare it with each person's response to 50g of pure sugar in a drink, the reference food. The ratio, multiplied by 100 and averaged across the group, is the GI, according to the University of Sydney's GI service.

The concept comes from Jenkins and colleagues, who tested 62 common foods in 1981. The 2008 international tables compiled by the University of Sydney list more than 2,480 food items (Atkinson et al., Diabetes Care 2008). The method is standardised, and GI is widely used in diabetes and prediabetes eating advice, including our prediabetes eating guide.

High, medium, low

• Low GI (55 or less): kidney beans (about 24), chickpeas (28), lentils (32), apples (36), full-fat milk (39), boiled carrots (39), oranges (43), bananas (51), rolled oat porridge (55).
• Medium GI (56 to 69): boiled sweet potato (about 63), boiled brown rice (68).
• High GI (70 or more): boiled white rice (about 73), whole wheat bread (74), white bread (75), watermelon (76), boiled potato (78), instant oat porridge (79).

These are average values across laboratories and varieties from the 2008 international tables, so read them as approximate. Two surprises stand out: whole wheat bread sits almost level with white bread, and instant oats sit far above rolled oats. That is why our low insulin breakfast guide treats the form of the oat as more important than the label on the box.

Where GI falls short

We want to be direct about the limitations:

1. It ignores portion. Watermelon has a high GI (about 76) but only about 7g of available carbohydrate per 100g, according to USDA FoodData Central, so a normal serving delivers little sugar. Glycemic load (GL = GI × grams of carbohydrate ÷ 100) corrects for this.
2. It assumes single-food meals. Real meals combine carbohydrates with protein, fat and fibre, which can soften the blood sugar rise. A high-GI food eaten alone behaves differently from the same food eaten as part of a meal with chicken and vegetables.
3. It varies between individuals. In the 2015 Cell study by Zeevi and colleagues, 800 people wearing sensors that track blood sugar continuously showed widely different blood sugar responses to identical meals.
4. It leaves out protein and fat foods. Chicken, eggs, salmon, butter and cheese have no GI value because they barely raise blood sugar on their own. A plan built only on GI ranks has nothing to say about a large part of most plates, and nothing to say about insulin.

Ready to stop guessing what to eat?

Glycemic Load: fixing GI's biggest flaw

Glycemic Load (GL) corrects the most glaring GI weakness by accounting for portion size. The formula is: GL = (GI × grams of carbohydrate per serving) ÷ 100. A food with a high GI but low carbohydrate content per serving ends up with a low glycemic load, meaning its actual blood-sugar impact is modest.

We argue glycemic load is a meaningful improvement over raw GI, but still incomplete: GL assumes single-food servings and ignores the protein, fat and fibre that strongly shape how the body responds to a real meal. In one study of 11 healthy adults (Hätönen et al., British Journal of Nutrition, 2011), mashed potato had a glycemic index of 108 on its own and 54 when eaten with chicken breast, rapeseed oil and salad, and the authors concluded that calculating a mixed meal's GI from its parts is imprecise. Insulin told a different story: adding chicken breast raised the meal's insulin response index from 118 to 148, while rapeseed oil lowered it. Protein also calls in glucagon, the hormone that counterbalances insulin, which sugar on its own pushes down (Westphal et al., 1990). That gap between blood sugar and insulin is what the insulin index measures.

Glycemic index vs insulin index: what is the difference?

The glycemic index measures how much a food raises blood sugar, while the insulin index measures how much insulin the body releases for an equal-calorie portion of it, and the two disagree most for protein-rich foods and dairy. The insulin index comes from a 1997 University of Sydney study that fed 38 foods in 1000 kJ (240 kcal) portions to groups of 11 to 13 healthy adults, measured insulin for two hours and set white bread at 100.

Overall, blood sugar and insulin scores were highly correlated (r = 0.70). But protein-rich foods and bakery products, which are rich in fat and refined carbohydrate, produced insulin responses that were disproportionately higher than their blood sugar responses (Holt et al., American Journal of Clinical Nutrition 1997). Same calories, different insulin.

Dairy is the clearest mismatch. In a 2001 Swedish study of regular and fermented milk products, glycemic indexes were low, 15 to 30, yet insulin indexes were 90 to 98, not significantly different from white bread (Östman et al., American Journal of Clinical Nutrition 2001). By GI alone, milk looks almost inert. By insulin, it does not.

For whole meals, insulin data predicted better than carbohydrate counting. In a 2009 study of 13 mixed meals with the same calories, the food insulin index of the ingredients predicted each meal's insulin response (r = 0.78) better than glycemic load (r = 0.68), and carbohydrate content alone was not a significant predictor (Bao et al., American Journal of Clinical Nutrition 2009).

What to use instead

Neither index alone describes a mixed plate, and the variable that matters in practice is the response to the meal you actually ate. RealFoods built its Weight Impact Score on both lines of research. Weight Impact is RealFoods' 1 to 10 score for how strongly a meal is likely to raise blood sugar and call in insulin, where lower is better: 1 to 2 is Burning Fat, 3 to 4 Mostly Neutral, 5 to 6 Fat-Burning Paused and 7 to 10 Keeping Fat. It is a modelled estimate from the meal's foods, not a medical measurement. We argue a meal-level estimate is the more practical everyday tool, because it scores the whole plate and adjusts for the portion you logged, and we read it as a pattern across a week rather than a lab value.

White rice has a high GI, and on its own it scores 8 out of 10 per 100g in RealFoods' published estimates. Eaten as part of a meal with chicken and broccoli, the same rice is estimated to land lower, because protein and fibre arrive with it. GI only tells you about the rice. We argue the meal is the more useful unit. See how 23 common foods rank on the Weight Impact Score for practical examples.

Common questions

What is the glycemic index in simple terms?

The glycemic index (GI) ranks carbohydrate foods by how much they raise blood sugar over two hours compared with pure sugar, which is set at 100. Foods at 70 or more are high GI, 56 to 69 medium, and 55 or less low GI, according to the University of Sydney's GI service.

What's the difference between glycemic index and glycemic load?

Glycemic index ranks the blood sugar response to a fixed 50g of available carbohydrate. Glycemic load multiplies a food's GI by the grams of carbohydrate in the portion you eat and divides by 100, so it accounts for portion size. Watermelon has a high GI of about 76, but 100g holds only about 7g of available carbohydrate, so its glycemic load is about 5.

What is the difference between the glycemic index and the insulin index?

The glycemic index measures the blood sugar response to a food's carbohydrate; the insulin index measures the insulin response to an equal-calorie portion of the whole food. In the 1997 study that introduced the insulin index, the two were highly correlated (r = 0.70), but protein-rich foods and bakery products produced more insulin than their blood sugar response predicted. Milk products are the clearest mismatch: in a 2001 study their glycemic indexes were 15 to 30, yet their insulin indexes were 90 to 98, close to white bread.

Is GI the same as the Weight Impact Score?

No. Glycemic index measures the blood sugar response to the carbohydrate in a single food. The Weight Impact Score is RealFoods' 1 to 10 modelled estimate for a whole meal, built on glycemic and insulin index research, that looks at the whole plate rather than one food. We argue the meal-level estimate is more useful for everyday decisions, with the caveat that it is an estimate, not a measurement.

Continue reading: how blood sugar drives weight gain, the Weight Impact Score of 23 popular foods, or how insulin resistance develops.