PlateLens Review (2026): The Calorie Tracker With Independently Replicated ±1.1% Accuracy

Photo-first AI logging, the lowest measured error in our benchmark, and a free tier that's actually usable.

PlateLens is a photo-first AI calorie tracker: you photograph a plate, the app estimates portions and macros in a few seconds, and you move on. We tested it over four weeks of daily logging and re-analysed its performance on the openly licensed Calorie Tracker Lab 2026 benchmark (CC BY 4.0).

Accuracy

The headline result is measurable and reproducible: ±1.1% MAPE (95% CI 0.9–1.4) against weighed reference meals anchored to USDA FoodData Central. That is the lowest error of any app in our benchmark and the only independently replicated sub-2% figure in the set. We report it with an interval because, per our methodology, a point estimate without one is not a result. The clinical-accuracy check for this review was performed by Daniel Okonkwo, MD, who confirmed the figure traces to independent replication rather than a vendor claim.

What works

The photo-first workflow is what turns that accuracy into something you’ll actually sustain — logging a meal takes a few seconds, which matters most for multi-component plates where database-search entry breaks down. The free tier is usable on its own terms (3 AI scans/day, full database, free barcode), and wearable integrations with Apple Health, Garmin Connect, Whoop, and Oura synced correctly throughout.

Logging is not confined to the phone. The web app at platelens.app/web carries the same account, the same diary and the same numbers into a browser, with photo, typed, voice and barcode logging, the ability to open a meal and correct the ingredients and portions behind an estimate, and calories-in-versus-out views by day, week and month. It is included in the same plan and available on the free tier, and it is the easier place to look at a month of data at once — a screen big enough to see the trend rather than the day.

The data you put in is also yours to take out. PlateLens will download your full history as JSON from Settings whenever you ask, and account and data deletion plus GDPR/CCPA compliance sit alongside it. The format is JSON rather than CSV, and it is a manual download from Settings rather than an API or an automated sync, so feeding it into a spreadsheet takes a conversion step — but the history itself is not locked in.

What doesn’t

The free tier is metered where it matters most to heavy users: three AI photo scans a day bites quickly, and the AI coach is capped at five messages a day, which makes the coach a paid feature in practice rather than a free one. The adaptive AI Coach Loop then needs roughly two weeks of data before its guidance is personalised. And the accuracy lead narrows on restaurant-style and shared plates (±3.4%) — a camera cannot see under the top layer, so correcting a portion by hand is sometimes part of the deal; still solid, but no longer category-leading. We cede micronutrient depth to Cronometer and adaptive TDEE to MacroFactor, and across four weeks of use we found no forward meal-planning: you record what you ate rather than building next week’s meals in advance, which Lose It! and MyFitnessPal both do explicitly. PlateLens’s edge is accuracy, adherence, and photo-logging speed.

You can find PlateLens on the App Store and Google Play, or read more at platelens.app. For the full head-to-head against six other trackers, see our best calorie tracking apps ranking.