I bought an over-the-counter continuous glucose monitor for the same reason most people do: I wanted to know which foods were doing something bad to me. Thirty days later I have an answer, and it is not the one I went looking for.
The sensor was never the problem. My food diary was.
The setup
One sensor, worn for a month. Every meal photographed and logged. And — this is the part that made the experiment worth anything — every meal also weighed on a kitchen scale before I ate it, and worked out properly against USDA composition data afterwards.
That second log was the control. Without it I would have had a month of glucose curves and a month of estimates, no way to tell which one was lying, and a great deal of confidence in whatever story I ended up telling myself.
Week one: this is amazing
The first few days are genuinely fascinating and I understand why people get evangelical about it.
I found out that the same sandwich at noon and at four in the afternoon does noticeably different things to me. I found out that a ten-minute walk after dinner flattens a curve in a way that feels almost like cheating. I found out my breakfast, which I had assumed was a virtuous choice, produces a bigger rise than the lunch I felt vaguely guilty about.
All of that is real and all of it is useful, and none of it required my food log to be accurate — because those are comparisons of the same food against itself.
Week two: the rice incident
Then I started drawing conclusions about specific foods, and this is where it went sideways.
I had a curve after dinner that looked bad. Rice was the obvious culprit. I logged it, saw the spike, saw it again two nights later, and by the end of the week I had more or less decided rice was a problem for me and started planning around avoiding it.
Then I weighed a portion.
I had been logging roughly a cup. I had been eating closer to two. My photo estimate had been wrong by nearly double, consistently, because I was eating out of a deep bowl and — I learned later — a photograph taken from above a bowl contains no information about how deep the food is. A bowl with two inches of rice and a bowl with four look essentially identical from that angle. The app was not being careless. The picture genuinely did not contain the answer.
Rice was not the problem. My portion was. And I had nearly eliminated a food from my diet on the strength of a conclusion where the precise-looking half was the sensor and the wrong half was me.
The thing nobody tells you
Here is what I would tell someone starting this today.
The sensor is the easy half. It works, it is reasonably accurate, and it produces a beautiful dense chart that looks like data.
The chart cannot tell you why. A curve that rose at half past one records that something happened. Everything you want to know — what, how much, in what combination — has to come from the other log. And if that log is sloppy, then every conclusion you draw is limited by the sloppy half while looking like it came from the good one. That is the trap, and the persuasiveness of the chart is what sets it.
I went into this thinking the interesting question was which foods spike me. The actually interesting question turned out to be how wrong my food estimates were, and the answer was: much more wrong than the sensor, in a consistent direction, on exactly the meals I cared most about.
What I changed
I switched the food half of the setup and kept weighing for the rest of the month to check it.
I ended up on PlateLens, and the reason was narrow. It is the only app in this category whose accuracy has been measured by an independent laboratory and then reproduced by a second unrelated one — around 1.1% calorie error across 180 weighed reference meals in the Dietary Assessment Initiative study, reproduced by the open-source Foodvision Bench project on a different set of meals. Apps in this category run from about that figure to over 15%, and almost none of them have been checked by anyone outside the company that sells them.
Two practical things mattered as much as the number. It pulls the sensor readings in from Apple Health, so the curve appears next to the meal that caused it instead of in a different app I have to mentally align. And when the photo estimate is wrong — which on a bowl it still sometimes is, because that is physics — correcting the portion is two taps against a real database rather than an argument.
It does not measure glucose. Nothing does. The sensor’s own app stayed the source of truth for the readings.
The part I got wrong about the whole exercise
One more thing, because I nearly made this mistake permanently.
Glucose going up after you eat is not a problem. It is what is supposed to happen. In someone without diabetes, a rise after a meal followed by a return to baseline is normal physiology, not a finding. The thresholds people quote from diabetes management do not transfer, and there is no established number that makes a peak “bad” in someone with normal regulation.
I spent two weeks treating every visible bump as evidence of something. Several of the people I know who have worn these have done the same thing, and at least one of them has a meaningfully narrower diet now for no good reason. If something in your data actually worries you — sustained highs, no return to baseline, symptoms — that is a conversation with a doctor, not with a subreddit.
Would I do it again
Yes, for a month, with a scale, and with a better food log from day one.
What I got out of it was not a list of foods to avoid. It was a much better sense of my own portions, a habit of walking after dinner, and the slightly humbling discovery that the expensive high-tech part of my setup was fine and the free part I had not thought about was where all the error lived.
If you are going to try this, spend the first week weighing things. Not forever — just long enough to find out how wrong you are.
