CGM for Weight Loss Without Diabetes: What the Trials Show

A continuous glucose monitor has not been shown, in any trial large enough to settle it, to cause weight loss in people without diabetes. The FDA cleared Dexcom Stelo and Abbott Lingo in 2024 to measure glucose without a prescription, not to treat weight.^[1][3] The largest trial of eating to flatten your own glucose spikes lost slightly less weight than a plain low-fat diet over six months, a gap that could be chance.^[11] The trials where CGM users did lose more weight enrolled 30 to 35 people, ran 30 days to 12 weeks, and wrapped the sensor in coaching.^[14][15] For someone on Ozempic or Mounjaro, the case for a CGM depends on diabetes and on what else is being taken, not on weight.

The evidence at a glance

What the FDA actually cleared

On March 5, 2024, the FDA cleared Dexcom’s Stelo as the first continuous glucose monitor sold without a prescription. The agency described it as intended for anyone 18 or older who does not use insulin, including people without diabetes who want to see how diet and exercise affect their glucose. It is not for people with problematic hypoglycemia, because it does not alert the user to a dangerous low. Each sensor lasts up to 15 days, and the app shows a reading every 15 minutes.^[1]

The 510(k) summary shows what changed from the prescription device. Stelo uses the same hardware as the Dexcom G7, but the original app narrowed the display to 70–250 mg/dL and did not provide any glucose or system alerts.^[2] Abbott’s Lingo followed on May 29, 2024, with the same adult, not-on-insulin scope and a display limited to 55–200 mg/dL.^[3] Lingo’s accuracy claim rests on clinical data from the FreeStyle Libre 2 study in people with diabetes. Abbott argued that diabetes, with its wider glucose swings, is the harder test of a sensor.^[3]

The Stelo Glucose Biosensor System may also help the user better understand how lifestyle and behavior modification, including diet and exercise, impact glucose excursion.

— FDA 510(k) K234070, Indications for Use

That sentence is the whole claim: the device helps you understand glucose. Two later clearances widened it. In March 2025 the FDA cleared the Signos Glucose Monitoring System, an app that reads an over-the-counter sensor, with an added line that the information “may be useful in helping users to maintain a healthy weight.” The testing listed in its summary covers software, cybersecurity and usability. No study of weight appears. The company supported the wording by saying studies have shown an association between glucose management and weight management.^[4] In June 2026 Stelo was cleared for anyone age 2 and up who is not on insulin. The new labeling says it is not for people on dialysis, with problematic hypoglycemia, changing medicines without a clinician, or with a history of eating disorders, and it still has no glucose alerts.^[5]

Subscription programs such as Nutrisense and Levels sell sensors bundled with an app and, in some plans, coaching or dietitian access. The FDA clearance covers the sensor and its cleared software. It does not cover a program’s food scores or weight claims. One more boundary: the FDA has warned consumers not to use smartwatches or smart rings that claim to read glucose without piercing the skin, because it has not authorized any.^[6]

What normal glucose looks like on a sensor

The best reference comes from 153 healthy children and adults who wore a blinded Dexcom G6 for up to ten days. Average glucose was 98–99 mg/dL, or 104 mg/dL in people over 60. The typical person spent 96% of the day between 70 and 140 mg/dL, about 30 minutes a day above 140, and about 15 minutes a day below 70.^[8]

Those numbers matter because subscription apps are built around spikes and dips. A Stanford study found that people with normal standard tests still reached the prediabetic range 15% of the time and the diabetic range 2% of the time on a sensor, and argued this showed hidden dysregulation.^[9] It also shows that healthy people go above 140 mg/dL after some meals, so a reading like that after a starchy meal is not, by itself, a diagnosis. The same goes for the other direction: the healthy volunteers in the reference study also logged short stretches below 70 mg/dL most days.^[8] Diabetes is diagnosed with laboratory blood tests, not with sensor peaks.

Does eating for your own glucose curve cause weight loss?

The idea behind the subscriptions comes from a 2015 Israeli study. Researchers monitored 800 people through 46,898 meals and found that the same meal produced very different glucose responses in different people. An algorithm using blood tests, diet, activity and gut bacteria predicted those responses, and in a small blinded trial, diets built on the predictions lowered post-meal glucose.^[10] That trial measured glucose, not weight.

The weight test came later, in New York. The Personal Diet Study randomized 204 adults with overweight or obesity and prediabetes or early type 2 diabetes to six months of either a standard low-fat diet or a personalized diet. The personalized group got app meal scores from an algorithm estimating each person’s glucose response to specific foods. Both groups had 14 counseling sessions. The low-fat group lost 4.31% of body weight and the personalized group 3.26%. The difference was not statistically significant, and the trial concluded that targeting glucose response did not produce more weight loss.^[11]

The most cited positive trial is from ZOE, the company that sells a personalized nutrition program. In 347 US adults over 18 weeks, its app program lowered triglycerides more than standard government diet advice, one of two primary outcomes; LDL cholesterol did not differ. Body weight improved among the secondary outcomes, while fasting glucose and insulin did not differ. The company’s co-founders and employees are among the authors.^[12] The program scored foods on glucose and fat responses, gut bacteria and health history, so the trial cannot say what a sensor alone would do.

The small trials that did show weight loss

A few randomized trials have tested a visible CGM as feedback rather than as the basis of a diet algorithm, and they lean positive. All of them are small.

• Croatia, 2026. 35 women with overweight or obesity, 12 weeks, with the same supervised exercise and diet advice in both groups. Women who wore a sensor three times lost 5.5 kg; women who wore it only at the start and end lost 0.2 kg.^[14] The control group losing almost nothing despite three exercise sessions a week is unusual, and the abstract does not report diabetes status.
• Prediabetes, 2024. 30 adults with overweight or obesity and prediabetes, 30 days. People who could see their readings and had personalized nutrition goals lost about twice as much weight and fat as those whose sensors were blinded.^[15]
• Type 2 diabetes, 2026. 151 adults, six months. A digital weight program with a CGM beat usual care by 3.2 percentage points of body weight and 0.46 points of HbA1c.^[16] The comparison was program plus sensor against a single dietitian visit, so the share due to the sensor cannot be separated.

A 2026 systematic review that pooled the evidence in adults without diabetes found 12 studies. A visible CGM may help with some diet changes when it is part of structured education; in one eight-week trial, people ate about 15 g less carbohydrate a day. It found no clear change in physical activity, too little evidence to judge psychological benefit or harm, and low to very low certainty overall. Its conclusion: large, lasting benefits and stand-alone effects are not established.^[13]

How accurate the readings are when you do not have diabetes

Sensors measure glucose in the fluid under the skin, not in blood, and the numbers lag and drift. The first independent test of over-the-counter sensors in adults without diabetes, published in September 2026, compared them with fingerstick meters in 39 young adults. Average error was 12.9%, and only 66.5% of readings fell within 15% or 15 mg/dL of the meter. Error was worse after a sugary drink than after placebo (15.3% vs 10.6%), readings trailed by about six minutes, and 13.8% of the trend arrows were rated moderate to extreme risk if acted on. The authors judged the devices reasonable for trends and unreliable for exact values or thresholds.^[17]

A British crossover trial in 15 healthy adults found a sensor read about 0.9 mmol/L (around 16 mg/dL) higher than fingerstick blood, both fasting and after meals. It overstated time above 140 mg/dL about four-fold, and it scored the glycemic index of the same fruit smoothie at 69 against 53 by fingerstick.^[18] Two sensors of the same brand worn at once agree closely, but different brands differ more: in ZOE’s PREDICT study, two sensors measuring the same meal varied by 12.5% between brands, against 3.7% within a brand.^[19] The ranking of meals from best to worst was more stable than the raw numbers.

The risk side: anxiety and food restriction

In a survey of 56 adults not on insulin who had used a CGM in the past year, 68% said high readings made them afraid of getting type 2 diabetes, whether or not they had diabetes. Younger people and those with obesity reported more distress, and eating-disorder symptoms were linked to distress about the sensor itself. Most also reported healthier eating or activity because of it.^[20] The 2026 Stelo labeling now excludes people with a history of eating disorders.^[5] Anyone who already counts, restricts or fears food should weigh that before buying; see GLP-1 drugs and eating disorders.

On a GLP-1: when a CGM matters

If a GLP‑1 is the only glucose-lowering drug you take and you do not have diabetes, the drug alone rarely pushes glucose too low, and no study has tested an over-the-counter CGM as a tool on Wegovy or Zepbound. A sensor will show flatter readings after meals, but whether that helps anyone lose more weight is unknown. The same is true after stopping: a 2026 review found no prospective study showing that sensor readings predict weight regain once the drug ends.^[21] The broader picture of what low blood sugar risk looks like on these drugs is in GLP-1s and low blood sugar.

Diabetes changes the answer. The Ozempic label warns that people taking it with insulin or a sulfonylurea may have more hypoglycemia, including severe hypoglycemia, and that cutting the dose of those drugs may lower the risk.^[22] The American Diabetes Association’s 2026 standards recommend a CGM for people with diabetes on insulin and for those on non-insulin drugs that can cause hypoglycemia.^[7] Those are the people the over-the-counter sensors were not built for. The ADA notes that OTC CGMs lack alarms and alerts and are meant for people not taking medicines that raise hypoglycemia risk; it lists both Stelo and Lingo as having no alarms.^[7] Someone on Ozempic plus insulin or a sulfonylurea needs a prescription CGM that can sound an alarm when glucose drops, and a plan from the clinician managing the diabetes.

Type 1 diabetes on a GLP‑1 is a separate case covered in GLP-1 drugs in type 1 diabetes, and how metformin fits in is in metformin and a GLP-1 together. A normal sensor trace on tirzepatide is also not the same as diabetes going away; see normal blood sugar is not remission. CGM data have been most useful in GLP‑1 research in insulin-treated diabetes, such as the study of GLP-1 drugs during Ramadan.

The verdict

Over-the-counter CGMs are real, FDA-cleared measuring devices, and they can teach you which foods raise your glucose most. That is a different claim from helping you lose weight, and the same gap applies to breath devices such as Lumen, which have even less behind them. The one trial built to test glucose-personalized eating for weight found no advantage over a low-fat diet,^[11] the positive trials are small and bundle the sensor with coaching,^[13] and the readings are least reliable at the after-meal peaks the apps grade.^[17][18] On a GLP‑1, the device matters most for people with diabetes who also take insulin or a sulfonylurea, and they need a prescription sensor with alarms rather than an over-the-counter one.