Your CGM is mostly noise.

A continuous glucose monitor, or CGM, is a small sensor that sits in the back of your arm and reads your blood sugar every five minutes for two weeks. It is the device behind a multi-billion-dollar wellness category. Levels, Lingo, Stelo, Oura (which absorbed Veri's CGM app when it bought the company in 2024). All selling roughly the same promise: wear one of these and you will learn things about your metabolism you cannot learn any other way.

The largest study to date says the device, worn by a healthy person, mostly measures itself.

Here is the simplest way to see what the data showed.

Imagine two people sit down and eat the same banana, at the same time, in the same room. Both are wearing CGMs. The sensors record very different blood sugar curves. One reads a small bump. The other reads a tall spike. The marketing reading: bananas are different for different people, this is why you need a CGM, your metabolism is unique. The actual reading, after thousands of healthy adults have done versions of this experiment, is that the sensors are noisy. Two of the same sensor on two of the same arms will not give you the same answer either.

Now run the experiment differently. The same person eats the same banana on Monday and on Wednesday. Same person. Same banana. The two traces look like they came from two different bodies. The variation inside one person, day to day, is roughly the size of the variation between two different people. The two-week average can smooth some of this out. The minute-by-minute panic that the apps are built around does not.

Save this. Three numbers the CGM marketing does not lead with.

The CGMap paper, by Keshet, Shilo and Segal in Cell Metabolism in 2023, is the largest single study of CGM data in healthy non-diabetic adults. The dataset covered 7,104 people aged 40 to 70, across two weeks of monitoring, with 49 different summary measures calculated from the traces. The headline finding that does not appear in any consumer CGM marketing: the variability between two different people, and the variability in the same person from one day to the next, are both larger than the cutoffs medical care uses to separate diabetic from non-diabetic. The same person, measured on Monday and on Wednesday, could land in two different metabolic categories depending on which day's trace you handed the algorithm.

Wyatt and colleagues, Nature Metabolism 2021, did find one real signal in CGM data on healthy adults. The dips after eating predict appetite and what people end up eating next. That is a useful finding. Less useful, in the same paper: the response to the same food in the same person on different days could move the verdict on whether that food was good for them in opposite directions. The dips predict appetite. The food-by-food readout most consumer apps build around does not survive a second test on a different day.

The 2026 systematic review of CGM use in non-diabetic populations is the cleanest summary of where the literature has landed. The pooled finding across the trials: in healthy people with normal glucose, the CGM did not produce a measurable improvement in any glucose number worth tracking. The studies that did show real benefit were almost all in people with pre-diabetes or actual diabetes.

The strongest case for wearing a CGM as a healthy person is not about the data. It is about behaviour. Wearing the device makes you think about what you eat. Users report changing their meals, cutting late snacks, taking the post-meal walk that comes up later in this week. The behaviour-change effect is real. The question is whether you need a $200-a-month subscription to make those changes, or whether the same behaviours, recommended by every clinical guideline for two decades, will land just as well without a sensor on your arm. For most people the second answer is the right one. At its best the consumer CGM is a $2,400-a-year nudge.

The CGM is a tool. In the wrong population it is a $200-a-month subscription to your own day-to-day bounce. Levels Classic, the bundled tier with two Dexcom or Libre sensors, runs about $199 USD a month. Lingo and Stelo run cheaper but the same logic applies. In the right population it is a life-changing device the brand fully endorses.

If you are worried about your metabolic health and you are not diabetic, the better first move is the test we covered Monday. Fasting insulin paired with fasting glucose, drawn once, gives you a sharper picture of where you sit than two weeks of CGM noise. In Australia, i-Screen sells the Insulin Resistance and Diabetes Risk Test direct for roughly $80 AUD. In the US, Function Health bundles it into a 100-plus-marker annual panel.

If you have type 1, type 2 or gestational diabetes and your clinician has not yet recommended a CGM, that conversation is worth having. The American Diabetes Association Standards of Care document the evidence and the use cases.

If you want one book on why glucose physiology matters in the people who actually need it, The Diabetes Code by Jason Fung is the most pragmatic single book on the subject. The first three chapters are the highest-leverage hour you can spend on the question.

Three behaviours that flatten the post-meal blood sugar curve, none of them requiring a sensor. Protein before carbohydrates. Vinegar before the meal. A ten-minute walk after. The research behind each is decades old. The brand has named all three across the year. Friday consolidates them into one weekend playbook.

Longevity Daily · The Building Decades

Share this edition with someone you love.

P.S. If you have not yet, the Healthspan Score is the five-minute assessment that surfaces which of your metabolic markers is most likely the next lever. Free, no blood draw.

Sources: Keshet et al, Cell Metabolism 2023 · Wyatt et al, Nature Metabolism 2021 · CGM non-diabetic meta-analysis, 2026

Follow @longevitydaily_  |  Subscribe