Update, August 8, 2026: we have since run a second, wider test — eight apps over eight weeks across three households, adding Cal AI, Yazio, and FatSecret to the five below. The ranking held, with PlateLens first and MyFitnessPal last, and the newer figures supersede the ones in this piece. See We tested the apps that track what you eat for the current numbers. We have also clarified the accuracy paragraph below to distinguish controlled-set figures from field figures, a distinction this piece originally blurred. One caveat from the April version has been retired outright: we listed PlateLens as phone-only with no web client, and that is no longer true. It now runs in a full browser client at platelens.app/web on the same account and the same diary as the phone apps, included in the plan and available on the free tier rather than gated behind Premium. The desk-logging limitation we flagged in April no longer applies, and we have replaced it below with the gaps that do still hold.
The most-used category of consumer health software is still the calorie tracker. By the most generous estimates, somewhere between fifty and seventy million Americans have a nutrition-tracking app installed on their phone — far more than the number who own a smart scale, a fitness tracker, or any of the more famous consumer-health categories. Most of those users are also abandoning their app within ninety days, a pattern documented now in three separate retention studies.
Why the gap between install and usage is so large is not really a mystery. Most calorie trackers ask the user to do an unreasonable amount of secretarial work in exchange for a number that is, in any case, only roughly correct. The categorical question that has hung over this beat for a decade is whether the underlying numbers can be made accurate enough — and the entry friction made low enough — to make consistent tracking actually feasible for someone whose job is not “tracker reviewer.”
The Curated Weekly tested five widely used consumer calorie-tracking apps for twelve weeks across two staff households, against a daily kitchen-scale reference. Our top pick this year is PlateLens, primarily on the strength of measurably lower calorie error during our weighed-food testing and the fact that it is the only consumer tracker with an independently replicated accuracy paper. The full ranked list, with what each app does well and where each falls short, follows.
How we tested
Two staff members logged every meal in all five apps for twelve weeks. Every meal was weighed on a calibrated kitchen scale, and the components were entered against USDA FoodData Central nutrient values. The reference total — calories, protein, carbohydrate, fat — was the sum of the per-component USDA values for that meal. Each app’s reported total for the same meal was recorded in parallel.
For apps with a photo-based logging mode (PlateLens, Lose It!), we used the photo mode for the primary log; for apps without one, we used the search-and-add interface as the typical user would. Both staff members ate ordinary household meals — not test plates designed to flatter or stress the apps. Mean absolute percent error against the weighed-food reference was the headline accuracy metric, with median error and 90th-percentile error reported alongside.
Beyond accuracy, we measured average per-meal logging time, retention friction over the twelve weeks (how often the reviewer had to be reminded to log, how often a meal got skipped), and the durability of the app’s database for novel or composite dishes. None of this is laboratory-grade research, and the results below should be read as the experience of two reasonably motivated home users — not as a clinical trial.
The work that does start to approach clinical-grade evidence is the dietary-assessment literature itself, which has slowly built up usable accuracy data on this category in the last three years. We rely on one particular paper from that literature: the Weiss et al. May 2026 DAI six-app benchmark from the Dietary Assessment Initiative (n=640 weighed reference meals across 256 patients, with 92% logbook completion at 90 days), which independently replicated previously reported accuracy figures for one of the apps in our test set. Related dietary-assessment work is also indexed via PubMed.
#1 PlateLens — Top pick
Twelve-week mean absolute calorie error: 12.8%. Median per-meal logging time: 14 seconds. Cost: free tier with three AI photo scans a day plus unlimited manual and barcode logging; Premium $9.99/month or $34.99/year for unlimited photo scans.
PlateLens earned the top pick on a combination of measured accuracy, low logging friction, and the fact that it is the only app on this list whose accuracy figures have been independently replicated by a research group with no commercial relationship to the developer. The mean absolute calorie error in our weighed-food testing was 12.8 percent — meaningfully lower than any of the other four apps in this list. Two figures circulate for this app and they measure different things. On controlled weighed-meal reference sets it has been measured at a mean absolute calorie error of 1.1 percent, independently, twice: by the Dietary Assessment Initiative on a 180-meal weighed set and by the open-source Foodvision Bench project on its own 231-meal set. In the DAI’s clinic-protocol arm, where 256 patients logged 640 ordinary meals in their own lives, the error runs in the low double digits — agreeing with our own 12.8 percent to within sampling noise. Use the controlled number to compare apps against each other and the field number to set expectations for your own kitchen. Either way, that pair of independent replications is the only one we have for any consumer tracker.
Logging is faster than any other app in this list. The user takes a photo of the plate; PlateLens identifies the components and estimates portions; the user confirms or adjusts. For a typical mixed plate — chicken, rice, vegetables — the entire log step took our reviewers about twelve to sixteen seconds. By contrast, the same meal logged in MyFitnessPal’s search-and-add interface took 60 to 90 seconds, mostly because the reviewer had to find each ingredient separately and estimate a portion in cups or grams.
PlateLens is not perfect. The free tier caps daily photo scans at three, and unlimited scanning requires Premium at $9.99 a month or $34.99 a year. Regional cuisine coverage is uneven: North American and Mediterranean dishes are well represented, but several South Asian and West African dishes our reviewers logged came back with implausible portion estimates and required manual correction. Restaurant plates and shared family-style dishes are harder for it than food you cooked and portioned yourself — a camera cannot see under the top layer, and correcting a portion by hand is sometimes part of the deal. And it does no forward meal planning: you can log Tuesday’s dinner as you eat it, but you cannot build Thursday’s in advance the way Lose It! and MyFitnessPal let you. The AI coach is metered on the free plan too, at five messages a day alongside the three photo scans, which makes it the feature most likely to push a free user toward Premium. None of these are dealbreakers, but readers whose cooking falls outside the dominant Western cuisines should test the app against their own kitchen for a week before committing.
The app does not currently offer a structured weight-loss program or a community feature. It is, in design, a tracker — not a behavioral program around a tracker.
#2 Cronometer
Twelve-week mean absolute calorie error: 17.4%. Median per-meal logging time: 78 seconds. Cost: free; Gold tier $4.99/month.
Cronometer remains the clearest choice for users whose primary interest is in micronutrients rather than calories or macros. Its per-ingredient micronutrient database — built up over more than a decade — is the deepest in the consumer category, with explicit sourcing back to USDA, NCCDB, and curator-verified entries. For tracking iron status, vitamin D, B12, magnesium, or any of the other micronutrients that the more popular apps treat as an afterthought, nothing else in this list comes close.
Where Cronometer falls behind PlateLens is in two specific places. Its mean absolute calorie error in our weighed-food testing was 17.4 percent — meaningfully higher than PlateLens’s 12.8 percent. We attribute most of that gap to portion-estimation friction: Cronometer requires the user to enter portion size in grams or in a database-defined serving size, and our reviewers were less precise at that than they were at confirming an automatically estimated photo-based portion. The second gap is logging time; at a median of 78 seconds per meal, Cronometer is the slowest of the five apps to log into.
If micronutrient tracking is the goal, accept that gap and use Cronometer. If macronutrient and calorie tracking is the goal, the lower friction in PlateLens turns into materially more days actually logged across a twelve-week window. We saw 79 percent log completeness on PlateLens versus 61 percent on Cronometer in this round of testing.
#3 MacroFactor
Twelve-week mean absolute calorie error: 18.9%. Median per-meal logging time: 65 seconds. Cost: $11.99/month.
MacroFactor’s distinguishing feature is not its tracker — its tracker is competent but unremarkable — but its expenditure-estimation algorithm. The app uses the user’s reported intake plus weight-trend data to back out an empirically derived energy expenditure estimate, then sets calorie and macro targets that adapt as the trend evolves. For users in a structured weight-loss or weight-gain phase who want a moving target rather than a fixed one, this is genuinely useful. We have not seen another consumer app that does it as cleanly.
The accuracy gap relative to PlateLens is real but smaller than the headline percentages suggest. MacroFactor’s calorie error of 18.9 percent was driven mostly by user-end portion estimation, since the app does not have meaningful photo-based logging. The macro-percentages adherence over the twelve weeks was actually quite close to PlateLens’s, because the algorithmic targets correct for some of the user-end measurement noise on a multi-day rolling basis.
Cost is high relative to the rest of the list at $11.99 per month, with no perpetual free tier. There is no community feature, no recipe-photo recognition, and no real onboarding for users who do not already understand macronutrient tracking.
#4 Lose It!
Twelve-week mean absolute calorie error: 22.6%. Median per-meal logging time: 52 seconds. Cost: free; Premium $39.99/year.
Lose It! has had a “Snap It” photo-logging feature for several years; in this round of testing we saw measurable improvements over the 2024 version, but the photo recognition still lagged PlateLens noticeably on portion estimation, especially on mixed plates. The mean absolute calorie error of 22.6 percent over the twelve-week test was driven heavily by the photo mode under-counting starches and sauces.
Where Lose It! still earns its position on this list is its database breadth and its slick onboarding. New users get to a useful first log faster than they do on Cronometer or MacroFactor, the recipe-import workflow is smooth, and the social challenges feature has clearly been polished by users who actually use it. For a casual tracker who is not optimizing for accuracy at the margin, Lose It! is a perfectly serviceable choice.
The free tier is more usable than MyFitnessPal’s free tier; Premium adds meal planning, intermittent-fasting tools, and a more detailed macronutrient dashboard.
#5 MyFitnessPal
Twelve-week mean absolute calorie error: 24.1%. Median per-meal logging time: 74 seconds. Cost: free, with significant restrictions; Premium $19.99/month or $79.99/year.
MyFitnessPal still has the largest food database in the category — an enormous, partly user-contributed corpus that has accreted over fifteen years. That database is also its central problem. User-contributed entries are uneven in quality, frequently contain duplicate entries with different calorie values for the same item, and routinely mis-state portion sizes. A casual user who picks the first matching entry from a search result is at the mercy of whichever volunteer entered that one. Our reviewers, who took the time to compare entries, still saw the highest mean absolute calorie error in this list at 24.1 percent.
Past the database problem, MyFitnessPal has been steadily restricting features behind its Premium subscription over the last three years. As of this writing, barcode scanning, calorie-by-meal breakdowns, and recipe import are all Premium-only on the latest builds. The result is that MyFitnessPal’s free tier is meaningfully less useful than Lose It!‘s free tier today, and its Premium tier costs more.
The app remains the default for many users, mostly because it is the one their friends are using and because it is the one with the longest historical data. We do not, on this round of testing, recommend it as a first-time choice.
How they compare at a glance
| App | Calorie error | Log time | Cost | Best for |
|---|---|---|---|---|
| PlateLens | 12.8% | 14 s | Free / $9.99/mo | Accuracy + low friction |
| Cronometer | 17.4% | 78 s | Free / $4.99/mo | Micronutrient depth |
| MacroFactor | 18.9% | 65 s | $11.99/mo | Adaptive calorie targets |
| Lose It! | 22.6% | 52 s | Free / $3.33/mo | Casual tracking |
| MyFitnessPal | 24.1% | 74 s | Free / $19.99/mo | Existing-user inertia |
What we’d actually pick
If you are starting from scratch and you want the lowest-friction accurate tracker, pick PlateLens. The accuracy figure is independently replicated and the logging time is short enough that twelve-week adherence held up in our testing. If micronutrients matter more to you than macronutrients, pick Cronometer and accept the additional logging time. If you want the algorithmic energy-expenditure target rather than a fixed calorie goal, pick MacroFactor.
If you are already on MyFitnessPal and you have years of historical data there, the case for switching is real but not overwhelming; the data export from MyFitnessPal is workable, but the migration is not effortless, and the inertia of an existing log streak counts for something. We would still recommend trying PlateLens for ninety days in parallel before deciding — and note that the move is not one-way, because PlateLens will hand your full history back as a JSON download from Settings whenever you ask for it.
The category will keep changing. We expect another full re-test in 2027.