Continuous Glucose Monitors and Cravings: What a CGM Can Actually Tell You (2026)
A continuous glucose monitor reads glucose in the fluid under your skin every few minutes. For cravings, the most useful number is the dip 2 to 3 hours after a meal. In human studies, a deeper dip predicts hunger returning sooner and more calories eaten later that day.
Two OTC glucose sensors are now sitting on drugstore shelves in the US. Dexcom's Stelo cleared the FDA in March 2024, Abbott's Lingo followed months later, and neither one needs a prescription. So a lot of people who don't have diabetes are now walking around with a graph of their own blood sugar on their phone.
Most of them are reading it wrong.
The conversation online is almost entirely about spikes. How high did the bagel take you, how flat was your line after the salad. But when researchers actually went looking for the glucose number that predicts eating behavior, the peak barely mattered. The dip did.
Key takeaways
- The 2 to 3 hour dip predicts hunger better than the peak.
- Deeper dips linked to 27% more calories over 24 hours.
- Expert clinicians disagree on reading non-diabetic CGM data.
- Meal order cut glucose response by roughly 40%.
- You can flatten dips without wearing a sensor.
What does a CGM actually measure?
Not your blood. That surprises people.
The sensor sits in the interstitial fluid, the thin layer of fluid between your cells, usually on the back of your upper arm. Glucose moves from your bloodstream into that fluid with a lag of roughly 5 to 15 minutes, and the sensor reads it every 1 to 5 minutes depending on the device.
So the graph on your phone is a slightly delayed, slightly smoothed version of what your blood is doing. That's fine for spotting patterns across days. It's less good for the kind of minute-by-minute detective work people try to do with it.
The other thing worth knowing: consumer sensors are calibrated for a population, not for you. Two sensors on the same arm on the same day can read 10 to 20 mg/dL apart. If you're comparing today's oatmeal to last Tuesday's oatmeal and the difference is small, the difference may be the sensor.
The graph on your phone is a delayed, smoothed guess at what your blood is doing. Treat it like weather, not like a thermometer.
Why does the glucose dip drive cravings?
Here's the study that reframed this for me.
In 2021, a team from King's College London, Massachusetts General Hospital and the University of Leeds published data on 1,070 people who ate 8,624 standardized meals followed by 71,715 free-choice meals, all while wearing CGMs. They wanted to know which part of the glucose curve predicted what people did next.
The answer was the dip at 2 to 3 hours, measured against your own baseline. Not the peak in the first 2 hours, and not the area under the curve, which is the number most glucose content obsesses over.
People with deeper dips reported more hunger at 2 to 3 hours, went a shorter time before their next meal, ate more at 3 to 4 hours, and ate 27% more calories over the following 24 hours. The pattern held in a separate US validation group.
A smaller 2026 study out of Loughborough University tested the same idea under lab conditions. 30 healthy young women ate a fixed breakfast, then had glucose measured both by CGM and by venous blood draw every 15 minutes for 4 hours, followed by an all-you-can-eat pasta lunch.
Same result. A one standard deviation deeper dip corresponded to roughly 16% more calories eaten at lunch, and a higher glucose low point corresponded to about 13% fewer. Both the sensor and the blood draw agreed, which matters, because it means the effect isn't an artifact of the sensor.
What's interesting is what didn't explain it. The researchers ran mediation analysis on insulin, ghrelin, GLP-1 and oxyntomodulin, and none of them accounted for the link. So the mechanism is still open. The association is real; the story we tell about why is still being written.
Is the evidence strong enough to act on?
Partly. Let's be honest about the ceiling.
These are observational associations, and the correlations are modest. In the 1,070-person study the correlation between dip depth and 24-hour energy intake was r = 0.27. That's a real signal at population scale, and it's nowhere near destiny for any individual person.
Nobody has run the trial that would settle it: take people, deliberately flatten their post-meal dips for 12 weeks, and see whether they eat less and lose weight compared to a control group. Until that exists, "flatten your dips" is a reasonable hypothesis with good supporting data, not a proven weight loss protocol.
What I take from it is narrower and more useful. If you get hungry 2 to 3 hours after eating and you'd assumed that was a character flaw, there's a measurable physiological pattern that lines up with the experience. That reframe is worth something on its own. We wrote more about the mechanism in blood sugar spikes and hunger.
What the research has actually studied in people with diabetes
A note on where a chunk of the CGM literature comes from, because it's easy to read across in ways you shouldn't.
Most CGM research was built for people managing diabetes under medical supervision. A 2026 crossover trial in Diabetologia put 25 adults with type 2 diabetes on two 4-week diets and reported 15% to 20% lower hunger and sweet craving scores on the dairy-enriched, protein-forward-breakfast arm. A 2023 observational study followed 21 adults with non-insulin-treated type 2 diabetes and found the dawn phenomenon raised morning glucose on about half of days.
Both are interesting. Neither tells you what your sensor means, because those participants were patients receiving care for a diagnosed condition, and their glucose physiology is not yours. If you have diabetes or prediabetes, your CGM decisions belong with your doctor, not with a blog. Our prediabetes overview covers where that line sits.
Should you buy a CGM if you don't have diabetes?
Here's the finding that should slow everyone down.
In 2025, researchers sent 20 genuinely ambiguous CGM reports from people without diabetes to 18 expert clinicians, including some of the biggest names in diabetes technology, and asked a simple question: would you recommend follow-up?
Agreement between them was poor. The Fleiss Kappa came out at 0.36, which in plain terms means the experts were closer to disagreeing than agreeing. Their own conclusion was that normative CGM data for people without diabetes doesn't really exist yet.
If 18 world experts can't agree on what a healthy person's glucose report means, your app's green-and-red score is a guess wearing a lab coat.
| A CGM is genuinely good at | A CGM is bad at |
|---|---|
| Showing your own repeated patterns across weeks | Comparing you to other people |
| Revealing that a "healthy" food does something odd to you | Telling you a food is good or bad in general |
| Making the timing of your hunger visible | Diagnosing anything |
| Testing one specific change (meal order, a walk) | Small differences inside sensor error |
| A 2 to 4 week experiment | Being worn forever without anxiety creeping in |
My take, for whatever a founder's opinion is worth: a CGM is a great 2 to 4 week experiment and a mediocre permanent accessory. Wear it, learn your own three or four patterns, take the sensor off, keep the habits.
There's also a real downside nobody selling these mentions. Some people get anxious watching a number all day, and food gets moralized fast. If you have any history of disordered eating, this is a tool to discuss with a professional before you buy it, not after.
What should you look for in your own data?
Four things, and none of them are the peak number.
1. The 3-hour reading versus your morning baseline. Check your fasting number when you wake up. Then, 2 to 3 hours after each meal, check again. If you're regularly landing below your morning baseline, that's the dip pattern the research describes.
2. Which specific meals do it. This is the part that's actually personal. People respond differently to identical food, and your sensor is one of the few ways to see your own version. A breakfast that flattens one person can dump another.
3. The gap between the dip and your snacking. Line up your dips with the times you actually go looking for food. If your 3pm slump and your 3pm dip land in the same window, you've learned something concrete.
4. What changes when you change one thing. Eat the same meal twice, change one variable, compare. Add protein first. Take a 10 minute walk after. Move the meal an hour earlier. One variable, two trials minimum, because sensor noise is real.
What isn't worth your attention: a single alarming spike, a number your friend's app called red, or any comparison to a stranger's screenshot. This connects to the wider pattern we cover in the insulin resistance and cravings cycle.
How do you flatten the dip without a sensor?
Most of the useful moves were established in controlled trials long before consumer CGMs existed. You can run them today.
Meal order does more than it has any right to. In a randomized crossover trial with 16 healthy adults, eating vegetables first, then protein, then carbohydrate blunted the glucose response substantially compared to eating the rice first, and the vegetable-protein-carb order produced higher GLP-1 release without demanding more insulin. A separate 18-person crossover trial in the UAE found the same sequence cut the 2-hour glucose response by 40.9% and left people fuller at 60 and 120 minutes.
That's a free intervention. Same food, same calories, different order.
| Tactic | What the human evidence shows | Effort |
|---|---|---|
| Veg and protein before carbs | 40.9% lower 2-hour glucose response, more fullness (n=18 crossover) | Free |
| Allulose with a carb load | Dose-dependent lower glucose and insulin excursion (n=30, randomized); 10% lower glucose iAUC in meta-analysis | Low |
| Protein-forward breakfast | 15-20% lower hunger and sweet craving scores (type 2 diabetes trial, n=25) | Low |
| Not eating carbs alone | Mixed meals blunt the curve versus carbohydrate first | Free |
Allulose is the one sweetener with real postprandial data. In a randomized double-blind crossover trial, 30 adults without diabetes took escalating doses of allulose alongside a 50g sucrose load. Glucose at 30 minutes came down in a dose-dependent way, significantly at 7.5g and 10g, and the 10g dose cut the overall glucose excursion. A 30-person Thai trial reproduced the dose-dependent pattern, and a meta-analysis of 40 trial comparisons put allulose at a 10% reduction in postprandial glucose iAUC with moderate certainty of evidence.
We went deeper on this in does allulose spike blood sugar and the allulose and GLP-1 research.
Chromium is the quiet one. It's an essential trace mineral involved in insulin signaling, which is why it shows up in so many blood sugar formulas. It supports blood sugar control rather than dramatically changing a curve. More in chromium for cravings and blood sugar.
Where does Ozzi fit into this?
Honestly? Ozzi is not a glucose product, and I'm not going to pretend it is.
Ozzi is a drink stick built for cravings and food noise. Three of its ingredients happen to sit in the same territory this article covers, so here's the accurate version.
8g of allulose is the largest ingredient by weight, and it's the ingredient with the human postprandial data described above. 11mg of chromium as Metabolex supports blood sugar control through insulin signaling. 500mg of glucomannan is a viscous fiber that absorbs water and expands in the stomach, which is a satiety mechanism, not a glucose one.
The rest of the formula (short-chain fatty acids from 500mg of L-Lysine Butyrate, plus 500mg of chicory root inulin) is aimed at gut and GLP-1 support, and we've written about that elsewhere in what GLP-1 actually is.
If your cravings are landing 2 to 3 hours after meals, the moves that pay off hardest are still the free ones: order your food differently, get protein in early, walk after dinner. Ozzi is what I built for the part that those don't fully solve, which is covered in how to stop food noise naturally.
Give the 3pm dip something to do
One stick in 16oz of water. 8g allulose, 500mg butyrate, 500mg glucomannan, 11mg chromium. Try it for 14 days straight. If it doesn't work, we'll refund your first bag.
Frequently asked questions
Can a CGM tell me why I crave sugar at night?
Not directly. It can show you whether your glucose is dipping below baseline in the hours before your cravings hit, which is one contributing pattern. Night cravings also involve sleep, stress, habit and how much you ate earlier in the day, and none of those show up on the graph.
Are OTC glucose monitors accurate for people without diabetes?
They're accurate enough to show trends and not accurate enough for precision comparisons. Sensors can read 10 to 20 mg/dL apart from each other, so treat small differences between meals with suspicion and look for patterns that repeat.
What's a normal glucose dip after eating?
There isn't an agreed answer, which is the honest problem. A 2025 survey of 18 expert clinicians found poor agreement on interpreting CGM reports from people without diabetes, and the authors called for normative data to be developed. Use your own baseline as the reference, not a published range.
Does a big spike mean I ate something bad?
Not on its own. Glycemic responses to identical meals vary a lot between people, and the research linking glucose to appetite pointed at the dip rather than the peak. A single reading tells you very little.
How long should I wear a CGM?
Two to four weeks is enough to identify your repeating patterns. Beyond that you're mostly paying for confirmation. If you find yourself checking it anxiously or scoring foods as good and bad, that's a reason to stop.
Will eating in a different order really change my glucose?
In controlled trials, yes. An 18-person crossover trial found vegetables and protein before carbohydrate cut the 2-hour glucose response by 40.9% versus the same food eaten as a standard mixed meal, and people reported more fullness at 60 and 120 minutes.
Does allulose raise blood sugar?
Human trials show the opposite direction. In randomized crossover studies, allulose taken with a sucrose load lowered the 30-minute glucose and reduced overall glucose excursion in a dose-dependent way. A meta-analysis of 40 comparisons put the effect at about 10% lower postprandial glucose iAUC.
Do I need a CGM to fix my cravings?
No. Every tactic in this article works whether or not you're wearing one. A sensor makes the feedback faster and more personal, which some people find motivating and others find stressful.
Can I use a CGM if I have prediabetes?
Talk to your doctor first. Most CGM research in that territory comes from supervised clinical settings, and interpretation genuinely changes when there's a diagnosis involved.
Does Ozzi affect what a CGM shows?
Ozzi isn't designed or tested as a glucose intervention and we don't make that claim. Its allulose and chromium sit in the blood sugar support category, and its glucomannan works through fullness rather than glucose.
References
- Wyatt P, Berry SE, Finlayson G, et al. Postprandial glycaemic dips predict appetite and energy intake in healthy individuals. Nature Metabolism. 2021;3(4):523-529. https://doi.org/10.1038/s42255-021-00383-x
- Jin Z, Thackray AE, Willis SA, et al. The association between glucose dynamics and energy intake in young, healthy women. European Journal of Clinical Nutrition. 2026;80(7):725-733. https://doi.org/10.1038/s41430-026-01727-0
- Spartano NL, Prescott B, Walker ME, et al. Expert Clinical Interpretation of Continuous Glucose Monitor Reports From Individuals Without Diabetes. Journal of Diabetes Science and Technology. 2025;20(3):727-735. https://doi.org/10.1177/19322968251315171
- Sun L, Goh HJ, Govindharajulu P, et al. Postprandial glucose, insulin and incretin responses differ by test meal macronutrient ingestion sequence (PATTERN study). Clinical Nutrition. 2019;39(3):950-957. https://doi.org/10.1016/j.clnu.2019.04.001
- Shaheen A, Sadiya A, Mussa BM, Abusnana S. Postprandial Glucose and Insulin Response to Meal Sequence Among Healthy UAE Adults: A Randomized Controlled Crossover Trial. Diabetes, Metabolic Syndrome and Obesity. 2024;17:4257-4265. https://doi.org/10.2147/DMSO.S468628
- Franchi F, Yaranov DM, Rollini F, et al. Effects of D-allulose on glucose tolerance and insulin response to a standard oral sucrose load: results of a prospective, randomized, crossover study. BMJ Open Diabetes Research & Care. 2021;9(1):e001939. https://doi.org/10.1136/bmjdrc-2020-001939
- Braunstein CR, Noronha JC, Khan TA, et al. Effect of fructose and its epimers on postprandial carbohydrate metabolism: A systematic review and meta-analysis. Clinical Nutrition. 2020;39(11):3308-3318. https://doi.org/10.1016/j.clnu.2020.03.002
- Buranapin S, Kosachunhanan N, Waisayanand N, Yokoi H, Tokuda M. Effects of D-Allulose with Sucrose Beverage on Glucose Tolerance and Insulin Levels among Thai Healthy Volunteers. Journal of Nutritional Science and Vitaminology. 2024;70(3):203-209. https://doi.org/10.3177/jnsv.70.203
- Tsameret S, Froy O, Matz Y, et al. Glycaemic, appetite and circadian benefits of a dairy-enriched diet with high-protein breakfast and early daytime-restricted carbohydrate intake in type 2 diabetes: a randomised crossover trial. Diabetologia. 2026;69(4):1021-1034. https://doi.org/10.1007/s00125-025-06658-2
- Jospe MR, Marano KM, Bedoya AR, et al. Exploring the Impact of Dawn Phenomenon on Glucose-Guided Eating Thresholds in Individuals With Type 2 Diabetes Using Continuous Glucose Monitoring: Observational Study. JMIR Formative Research. 2023;7:e46034. https://doi.org/10.2196/46034
Research sourced via PubMed. This article is educational and is not medical advice. Ozzi is a dietary supplement and is not intended to diagnose, treat, cure or prevent any disease. Talk to your doctor before starting a continuous glucose monitor if you have a diagnosed metabolic condition.
