Glycemic index vs glycemic load: why GI alone misleads

The glycemic index (GI) has been one of the most influential concepts in popular nutrition since its introduction in the 1980s by David Jenkins at the University of Toronto. It's also one of the most-misused metrics, mostly because it answers a narrower question than people assume.

What GI actually measures#

GI is a ranking of how much a fixed dose (50 g of available carbohydrate) of a specific food raises blood glucose over two hours, compared to a reference (50 g glucose = 100, or 50 g white bread = 100, depending on the scale used).

It's a measure of carbohydrate quality, normalised to dose. It tells you "if I ate 50 g of carbohydrate from this food, how much would it raise my blood glucose compared to 50 g of pure glucose?"

That's a useful question, but it's not the one most people are actually asking.

Why GI alone misleads#

The famous example is watermelon. Watermelon has a high GI (~76). A typical serving has very little actual carbohydrate (about 12 g per 200 g serving). The actual glucose impact of a normal portion of watermelon is small. GI alone makes watermelon look problematic when it isn't.

Conversely, a moderate-GI food eaten in a large portion can produce a large glucose response.

Glycemic load (GL)#

Glycemic load corrects for portion size:

GL = GI × (grams of carbohydrate per serving) / 100

• Low GL: <10
• Medium GL: 11–19
• High GL: ≥20

Watermelon, 200 g serving: GI 76 × 12 g / 100 = GL ~9 (low). A baked potato: GI 85 × 30 g / 100 = GL ~25 (high). These numbers reflect what's actually going to happen in your blood.

What both metrics miss#

GI and GL are population averages. The same food produces widely different responses in different people (Zeevi et al. 2015, PREDICT 1).

Both metrics are also measured in isolation. Real meals combine foods. Pairing high-GI carbohydrate with protein, fat, or fibre flattens the curve substantially: see fibre-first-meal-order.

And both metrics are limited to glucose. Insulin response can rise without a corresponding glucose rise (the insulin index). Some low-GI foods produce surprisingly large insulin responses.

What the long-term outcome evidence shows#

The dietary-pattern epidemiology is consistent: diets dominated by low-GL foods (whole grains, legumes, non-starchy vegetables, intact fruit) are associated with reduced type 2 diabetes incidence and modestly reduced cardiovascular mortality. This is a real signal across many cohorts.

But the effect is plausibly more about whole-food vs ultra-processed-food patterns than about a precise GL number. Low-GL whole foods come with fibre, micronutrients, satiety, and less calorically dense packaging.

What we'd actually say#

GL is the more useful metric than GI. Both are useful frames for thinking about meals, but neither captures the reality of how your individual body responds to a specific meal in a specific context.

If you're using a CGM, your own curves are more informative than any published GI/GL table. If you're not, a simple rule of thumb that captures most of the signal: eat mostly whole foods, mostly plants, with protein and fat at most meals. The composition will land in the low-GL zone naturally.

When to talk to a clinician#

• Persistent post-meal glucose >11 mmol/L or fasting >7 mmol/L
• Significant unexplained weight loss with high carbohydrate intake (possible new-onset diabetes)
• On insulin or sulfonylureas: match medication timing to your actual meal patterns, not population GL tables