Glucotypes

Treat glycemic-index tables as a guide to the average person, not a forecast for you. If you have a continuous glucose monitor, a few matched food tests can hint at which staples hit you hardest, but small differences can't be told apart; without one, whole-food, low-glycemic-load eating is the safer default.

What a glucotype is, and what it isn't

The glycemic index (GI) scores how much a food's carbohydrate raises blood glucose, as a percentage of the rise from pure glucose, averaged across a group of tested people. The question here is how well that average describes any one person.

The word "glucotype" comes from a Stanford study in which 57 people without a diabetes diagnosis wore continuous glucose monitors (CGMs), wearable sensors that track glucose around the clock. The researchers sorted people by their most frequent pattern of glucose variability — low, moderate or severe — and called those three patterns glucotypes. Sixteen people whose standard tests classed them as normal reached prediabetic or diabetic glucose levels after at least one standardised meal, though only 30 of the 57 ran the standardised meals.^[1]

The "rice-spiker" and "bread-spiker" labels you may have heard come from commercial messaging, not from that study. It defined patterns of variability, not which staple a person spikes on. The glucotype classification itself was later formally challenged on how far it generalises, and it has not been independently replicated in people without diabetes.

The case for personal responses

Even when the test is done carefully, GI is a poor personal predictor. In a tightly controlled study, white bread averaged a GI of 62, but individual results scattered by about 15 points either way — low-GI for some people, medium for others and high for the rest — and one person's own result varied by around 20% from test to test. The investigators concluded that GI "is unlikely to be a good approach to guiding food choices" for individuals.^[2]

The best-known study of personal responses followed 800 people through about 47,000 meals. Responses to identical standardised foods varied widely between people, while the group average still tracked published GI values reasonably well (a correlation of about 0.69). The strongest predictors of a person's response were their baseline glucose status — HbA1c, the three-month average of blood sugar, and waking glucose — followed by age, body-mass index, liver enzymes, blood pressure and gut bacteria. A computer model built on these predicted individual responses better than carbohydrate counting did: its predictions correlated with the measured responses at 0.68, against 0.38 for counting carbohydrates.^[3]

The same study also put a ceiling on how well any method can do. When one person ate the same standardised meal twice, their two responses correlated at only 0.71 to 0.77 — so a person's own repeat result only roughly matches their first.

The case against: mostly noise

A 2026 reanalysis of 382 healthy adults and more than a thousand paired tests pushes back on the whole framing. About 90% of responses predicted from each person's own response to pure glucose, scaled by the food's average GI, fell inside that person's own test-to-test range, and simulated data reproduced the observed spread between people without any person-specific settings at all. That suggests much of the apparent individuality is ordinary day-to-day variation in how well a person handles glucose. The analysis also gives a practical rule: two foods need to differ by at least 15 GI units before one person can reliably tell their responses apart.^[4]

Its senior authors are long-standing defenders of the glycemic index, so read it as a counterweight rather than a settlement. Either way, the studies on both sides point to the same practical lesson: a single test of a single food says little, because the same person's response varies so much from day to day.

What hasn't been shown

That individual responses differ is well established. That acting on those differences changes long-term health is not: no trial has shown that eating to your own glucose curve, rather than to a sound general diet, prevents disease or extends life. Whether commercial personalised-nutrition programmes add anything is covered in Continuous glucose monitors.

What to do

1. Read GI tables as averages. Two foods less than 15 GI units apart can't be reliably told apart in one person.
2. If you have a CGM, test a few staples over a two-week trial. Eat your normal meals and add a few matched tests — white rice, white bread, pasta and boiled potato, each with the same amount of carbohydrate. Repeat a test before trusting it. A consistently higher peak hints at which staple to pair with vegetables, protein, vinegar or food order.
3. Don't let the screen run your eating. Reviewers have raised, as a worry rather than a finding, that watching your trace too closely could feed obsessive eating.^[5] Read Continuous glucose monitors before over-reading a healthy trace.
4. Without a CGM, don't guess at your type. Whole-food, low-glycemic-load eating is the safer default; the levers that work for most people are on Glycemic index.

Further reading

• Hall H et al. Glucotypes reveal new patterns of glucose dysregulation. PLOS Biology 2018.^[6]
• Zeevi D, Korem T, Segal E, Elinav E, et al. Personalized Nutrition by Prediction of Glycemic Responses. Cell 2015.^[7]
• Matthan NR, Ausman LM, Lichtenstein AH, et al. Estimating the reliability of glycemic index values and potential sources of methodological and biological variability. Am J Clin Nutr 2016.^[8]
• Della Corte KA et al. Individual glycemic responses scale predictably with the glycemic index: evidence against person-specific food effects. Am J Clin Nutr 2026.^[9]