Continuous glucose monitors went from diabetic medical devices to fitness trend in about 18 months. Levels, Abbott Libre, Dexcom Stelo. You can buy one over the counter now. Strap it on your arm, stream the data to your phone, watch your blood sugar rise and fall throughout the day.
The marketing is seductive. Eat this to optimise your glucose. Avoid these foods. Stay in range. Maximise your metabolic health.
I’ve worn one for six weeks. I’ve talked to triathletes, Ironman athletes, and strength athletes who’ve worn them for months. I’ve read the nutrition research on non diabetic populations. Here’s what the honest picture looks like.
What a CGM Actually Tells You
A CGM measures interstitial fluid glucose every few minutes. Not blood glucose, but a close proxy, with a 10 to 15 minute lag from the actual bloodstream value.
For diabetics, this is life changing technology. For athletes and fitness enthusiasts without metabolic disease, the data is less actionable than the marketing suggests.
A healthy, non diabetic athlete will see glucose readings that look like a gently rolling hill. Baseline around 80 to 90 mg/dL. Rises to 130 or 140 after meals. Back to baseline in 90 to 120 minutes. Mild overnight drift down during sleep. Occasional spikes to 160 or 170 after particularly carb heavy meals.
Everything in this range is normal. It’s what your body is supposed to do.
The CGM also shows you artifacts. Sensor compression when you sleep on it. Sudden drops to 55 that are measurement noise, not real hypoglycemia. Spike patterns that make you think a banana ruined your morning when in fact your insulin response handled it cleanly and appropriately.
The signal is there. The noise is substantial. The interpretation skill required to separate them is real.
The Questions Athletes Think CGMs Answer
Here are the questions I hear athletes say they want a CGM for.
“Does eating X spike my glucose?” Almost everything spikes glucose. Rice spikes glucose. Bananas spike glucose. Oats spike glucose. The spike is not the problem. The question is whether the spike is managed appropriately and returns to baseline at a normal rate. The CGM shows you spikes but doesn’t interpret them for you.
“Am I in the right fuelling zone for training?” Glucose during exercise follows its own dynamics. Starting glucose doesn’t predict training quality. Mid workout glucose is often low in trained endurance athletes because fat oxidation is dominant. A reading of 75 during a 2 hour ride is not fuelling failure. It’s physiology working as designed.
“Am I metabolically flexible?” Metabolic flexibility is a real concept with specific definitions. You cannot infer it from CGM data alone. You need to look at glucose response alongside insulin, alongside fat oxidation rates, alongside substrate utilisation under load. A CGM gives you one of those inputs.
“Should I eat this?” The CGM cannot tell you this. The CGM tells you what happens to your glucose after you eat. Whether you should eat something depends on your training load, body composition goals, recovery status, and a dozen other variables the sensor does not measure.
The device answers none of these questions cleanly. It provides one dimensional data that requires context to interpret usefully. Most users don’t have that context.
What CGMs Are Genuinely Useful For
There are specific, real use cases where CGM data earns its price tag for athletes.
Reactive hypoglycemia detection. Some athletes, especially after hard morning training, experience a significant glucose crash 90 to 120 minutes after refuelling. Not a normal post-meal return to baseline. An actual undershoot into the 60s or 55s, accompanied by shakiness and cognitive fog. A CGM catches this pattern reliably. If you’re experiencing crashes a couple of hours after post training meals, two weeks of CGM data can confirm or rule out reactive hypoglycemia and guide a specific nutritional adjustment.
Pre race fuelling strategy validation. Testing race day nutrition in training is difficult because you can’t easily quantify how well a fuelling approach is working. CGM data during training sessions with different fuelling protocols gives you objective evidence about whether your carbohydrate intake is matching oxidation. If your glucose is spiking above 160 during exercise or dropping below 75, the fuelling strategy needs adjustment.
Detecting significant metabolic dysfunction. Occasionally, a CGM catches something the athlete genuinely did not know. Glucose never dropping below 100, even fasted. Consistent post meal spikes above 180 with slow return. Overnight glucose rising instead of stable. These patterns suggest real metabolic issues that warrant a fasting insulin test and a conversation with a doctor. Rare, but genuinely actionable when found.
Understanding individual meal response variability. Your carbohydrate tolerance is different from mine. A person with excellent metabolic health can handle 80 grams of rice without a meaningful spike. Another person responds to the same meal with a 70 point rise. If you want to understand your specific response to specific meals, two to three weeks of CGM data is enough to map your own patterns.
The Mistakes Athletes Make with CGM Data
These are the common misuse patterns I see.
Treating spikes as bad. Spikes are normal. A glucose rise to 140 after an oats and honey breakfast is not a problem. The quality metric is how fast it returns to baseline, not whether it went up at all. Athletes who start avoiding any food that causes a spike end up undereating carbohydrates they actually need for training.
Chasing flat line glucose. The healthiest looking CGM trace is not actually flat. A healthy metabolism has meaningful glucose variability. Rolling hills, not a straight line. Athletes trying to engineer a perfectly flat glucose curve often end up eating in ways that are suboptimal for training adaptation.
Ignoring exercise glucose drops. Glucose often falls during endurance exercise, especially after 90 minutes. This is normal and does not require immediate fuelling for most athletes. The body is drawing on stored glycogen and fat. Responding to a normal exercise glucose drop with aggressive mid workout carbs can actually blunt the metabolic adaptations you’re trying to train.
Making dietary decisions off noise. Short term CGM data is noisy. Sensor placement, sensor age, sleep compression, hydration, and ambient temperature all affect readings. Decisions based on one meal on one day with one sensor are often wrong. Patterns across two weeks of consistent sensor data are what has signal.
What CGMs Don’t Help You With
Training load. Recovery. Sleep quality. HRV. Hydration. Protein intake. Micronutrient status. Stress. None of these are captured in CGM data. All of these affect your performance more than your glucose trace on a normal day.
A CGM is one stream of data about one system (glucose homeostasis). The things that actually determine whether you have a good race are largely outside of what that sensor can see.
The Honest Recommendation
If you’ve got a specific question a CGM can answer (fuelling strategy testing, suspected reactive hypoglycemia, curiosity about individual response to specific foods), use one for 2 to 4 weeks, gather the data, act on what you learn, and then take it off.
If you’re considering wearing one chronically to optimise your metabolism in general, the evidence doesn’t support it for healthy athletes. You will spend $50 to $80 a month generating data you mostly cannot act on, and you will probably accumulate enough noise and anxiety about meals to damage your training nutrition.
The fitness industry moved from heart rate monitors to GPS watches to power meters to recovery scores to HRV to CGMs. Each one promised to be the missing piece. Each one turned out to be useful in specific contexts and overhyped in general. CGMs are following the same curve.
The harder work, connecting the data you already have across multiple sources into a coherent picture of what’s happening in your training, is what actually moves performance.
Green score. Destroyed legs. There are 6 blind spots in your wearable data. We wrote a free guide covering every one of them.
Green score. Destroyed legs. There are 6 blind spots in your wearable data. We wrote a free guide covering every one of them.
Download the Free Guide