I used to think of lunch as the easy meal. Breakfast had rules. Dinner had rules. Lunch was just... whatever was in front of me, eaten fast, between two other things.
Then I put a continuous glucose monitor on a client who insisted she was "doing everything right," and watched her glucose climb past 9.5 mmol/L an hour after a lunch she'd have described, confidently, as healthy. Wholegrain wrap. Hummus. A bit of fruit. On paper, sensible. On the graph, a spike followed by a crash that landed her, predictably, face-down in her 3pm slump.
That's the thing a CGM does that no food diary ever will: it stops you arguing with your own biology.
What the data actually shows
This isn't a fringe finding. A landmark 2015 study published in Cell (Zeevi et al.) put continuous monitors on hundreds of people eating identical, standardised meals, and found wildly different glucose responses from person to person. Two people, same sandwich, completely different curves. Glycaemic response isn't a fixed property of food. It's a property of you, your sleep the night before, your stress levels, your menstrual cycle phase, what you ate three hours earlier, even the order you eat things in.
Which is why generic nutrition advice so often fails health-educated, high-performing people. You're not getting it wrong. You're getting generic advice for a deeply individual system.
The other thing the research is consistent on: post-meal glucose patterns aren't really about the spike itself, they're about the shape of the curve. A healthy response looks like gentle rolling hills: a moderate rise, a steady return to baseline. What derails people isn't necessarily eating carbohydrate at lunch, it's the sharp spike-and-crash pattern, where blood sugar rises fast and falls just as fast, leaving you running on fumes by mid-afternoon. In healthy adults without diabetes, research generally puts an "average" post-meal peak somewhere around 5.5 mmol/L (about 99 mg/dL) after a balanced meal, meals lower in fibre and higher in refined carbohydrate push that ceiling considerably higher.
Why this matters more than "what to eat"
The honest answer to "what should I eat for lunch" isn't a meal plan. It's: eat protein and fibre first, keep refined carbohydrate in proportion rather than centre-stage, and then look at your own data before deciding what "works" for you. That's not me being evasive, it's the actual finding. Standardised advice misses the fact that your lunch response today, mid-cycle, under-slept, is not the same as your lunch response three weeks ago.
This is also, not coincidentally, where the post-lunch dip and the glucose curve start to overlap. Researchers studying afternoon alertness have found the post-lunch dip is a real, partly circadian-driven phenomenon, but it's reliably exacerbated by a high-carbohydrate lunch. So the meal doesn't cause the dip outright. It makes it worse, on a day your biology was already heading there.
What I actually changed
Once I started reading my own data rather than guessing, three things shifted permanently in how I build a lunch:
Protein and fat go first on the plate, mentally and literally. Not because carbohydrate is the enemy, because sequencing blunts the spike, and the research on this (alongside the data I've watched in real time on clients) is consistent.
I stopped treating "wholegrain" as a synonym for "safe." A wrap is still a fast-digesting starch in a flexible format. It behaves differently to oats with fat and fibre folded through it, even with similar macros on a label.
I check the curve, not the calorie count. A flat, gentle line in the two hours after lunch tells me far more about whether that meal is working for my body than anything printed on the packet.
This is, in the end, the entire premise of CGM work: not optimisation for its own sake, but replacing assumption with evidence. You don't need to guess what your 3pm will look like. You can see it coming, hours in advance, on a graph, and change the meal that's about to cause it.
This article discusses general research on continuous glucose monitoring and is not personalised medical advice. If you have a diagnosed metabolic condition, work with your GP or a qualified practitioner before making changes based on glucose data.
References
-
Zeevi, D. et al. (2015). Personalized Nutrition by Prediction of Glycemic Responses. Cell, 163(5), 1079–1094.
-
Levels Health (2026). The 2026 Levels Guide to Continuous Glucose Monitoring.
-
Levels Health (2026). What Should Your Glucose Levels Be? The 2026 Levels Guide to Healthy Blood Sugar Ranges.
-
Monk, T.H. (2005). The post-lunch dip in performance. Clinics in Occupational and Environmental Medicine, 4(2), 105–112.