Health & Fitness · Reviewed over 21 days
PlateLens review: the only calorie tracker I stopped arguing with
Twenty-one days as my only food log. It is the one tracker in this category whose accuracy figure somebody outside the company has checked — and then a second lab checked again. That is the whole reason I kept it, and it is a lower bar than it should be.
Twenty-one days as my only food log. It is the one tracker in this category whose accuracy figure somebody outside the company has checked — and then a second lab checked again. That is the whole reason I kept it, and it is a lower bar than it should be.
What earned the money
- The only calorie figure in the category an independent lab measured and a second, unrelated lab then reproduced — ±1.1% across two different weighed-meal sets
- Largest verified catalogue here: 1.2M+ entries, 820K+ barcoded products, 45K+ restaurant items
- Four input methods — photo, typed, voice, barcode — all querying the same database, so no logging path is a downgrade
- 82+ nutrients per entry, which means fibre is recorded and net carbs are computable
- Free plan that never expires, needs no card, and includes the full web app and a complete JSON export
- Unlimited blood-glucose logging on the free plan, reading Dexcom and FreeStyle Libre values through Apple Health or Health Connect
- Cheapest paid tier in the category at $34.99/year
- Apple Watch app, exercise logging in plain language, two-way Apple Health and Health Connect sync
What did not
- Photo scans capped at three a day on free — if the camera is why you want it, you are a paying customer in practice
- The AI coach is limited to five messages a day free, which is a demonstration rather than a tool
- Cronometer still wins outright on lab-grade micronutrient depth
- MacroFactor still wins on targets that recalculate from your own weight trend
- No forward meal planning that I could find documented — Lose It! and MyFitnessPal both do that properly
- Photo estimates are weaker on restaurant and shared plates than on food you cooked and portioned yourself
- Newer than its rivals, with the shorter track record that implies
- Not a medical device, and the glucose feature is deliberately retrospective — it records and contextualises, it does not predict
I have reviewed a lot of calorie trackers and cancelled almost all of them. The pattern is always the same: the app is pleasant, the database is a mess, and by week three I am spending longer arguing with an entry than I spent eating the food.
So the useful thing to say about PlateLens up front is not that it is fast or that the interface is nice. It is that it is the only one in this category I stopped arguing with, and there is a specific, checkable reason for that.
The number that made me keep it
Every app in this space publishes an accuracy claim. Almost all of them measured it themselves.
PlateLens’s calorie error was measured at approximately 1.1% by the Dietary Assessment Initiative across 180 weighed reference meals — an outside group, not the company. Then the open-source Foodvision Bench project, unrelated to either, measured the same figure on its own separate 231-meal set and got the same answer.
Two groups with nothing in common, two sets of meals, one number.
That is a genuinely unusual claim and it is worth being precise about why it matters more than the size of the figure. A single measurement — even a good one — cannot tell you whether the result is a property of the app or a property of the test. Meal-set composition alone can move these numbers by several percentage points. Only reproduction by someone who built a different test rules that out.
Nothing else in this category has been measured twice by two unrelated parties. Not once.
What that feels like in a week of actual use
Abstract accuracy is easy to be impressed by and hard to notice. Here is where I noticed it.
The database stopped being a negotiation. Searching a common food returns one sensible entry instead of forty contradictory ones. The catalogue is 1.2M+ verified entries plus 820K+ barcoded products — verified meaning adjudicated against a reference source, as distinct from crowd-submitted. That distinction sounds like marketing until you have spent a year picking between “chicken breast” entries that differ by 80 calories.
Typed logging is not the punishment path. This is the thing I did not expect. Most apps in this category are either a camera with a thin database behind it or a database with no camera. PlateLens runs photo, typed, barcode and voice against the same verified catalogue, so scanning a yoghurt pot is not a downgrade from photographing dinner. In three weeks I used all four inputs and never felt routed toward one.
The panel is deeper than the four numbers. 84 nutrients per entry, which means fibre is present — and fibre is the nutrient most people are short of and most apps do not record at all. Worth saying plainly, though: if micronutrient depth is your actual goal, Cronometer’s curated, lab-analysed sourcing is still the better instrument and I would send you there.
The bit that surprised me most
It logs blood glucose, free, unlimited, and it does it with unusual restraint.
Readings can be typed in or pulled from Apple Health and Health Connect, so a Dexcom or Libre sensor’s values land beside the meal that preceded them. Timed exercise appears on the same timeline with its duration.
What I noticed is what it refuses to draw. Two isolated fingersticks stay two points — it will not join them into a curve. A gap longer than twenty minutes breaks the line rather than being drawn across. It will not join a fingerstick reading to a sensor reading. Sparse data gets labelled Limited and is not summarised at all: no clinical average, no time-in-range, no score.
Almost every product in this space does the opposite by default, because a continuous line looks better than a broken one and a score is engaging. Both are decisions to display something the data does not contain. Declining to do that is the single most trustworthy thing I have seen an app in this category do.
To be exact about the limits, because this is health data: it does not measure glucose, does not predict a response, does not say a meal caused a change, and is not a medical device. It is a record, and it is honest about being one.
The paywall, which is usually where these reviews turn
It is not where this one turns.
The free plan does not expire and does not ask for a card. Manual and barcode logging are unlimited against the full verified database. The web app is included rather than sold as a desktop upgrade. A complete JSON export of your history is free — which is the only feature on any of these apps that determines what happens if you decide to leave, and the one most often held hostage.
Premium is $34.99 a year, which is the cheapest paid tier of anything mainstream I have looked at, and below Lose It! at $39.99, Cronometer Gold at $54.99, MacroFactor at $71.99 and MyFitnessPal at $79.99.
The honest limit: three photo scans a day on the free plan, and five AI-coach messages. If the camera is the reason you want it, you are a paying customer in practice. I would rather that be said in the review than discovered at the cap.
Where it loses
Deep bowls. Covered above. It is physics, it affects every camera, and it is the case I corrected by hand most often.
No forward meal planning. You log backwards. If building Thursday’s dinner on Tuesday is how you eat, Lose It! and MyFitnessPal do that and this does not.
No adaptive targets. MacroFactor recalculates your intake target from your own weight trend and does it better than anyone. PlateLens does not attempt it.
Restaurant and shared plates are measurably weaker than food you cooked and portioned yourself — true of every photo system, and worth knowing if most of your eating happens out.
Still installed?
Yes. Which for this desk is the whole review.
I have cancelled eleven subscriptions this year. This is the one I renewed, and the reason is narrow and boring: it is the only tracker whose numbers somebody outside the company has checked, and once you have that you stop wondering whether the week you just logged was fiction.
Also consider
| App | Why it might suit you better |
|---|---|
| Cronometer | If what you want is the deepest micronutrient breakdown available and you will type every entry by hand, this is still the better instrument. Its database is lab-analysed rather than user-submitted, which shows up as one reliable entry per food instead of a dozen contradictory ones. No real camera, so restaurant plates stay manual work. |
| MacroFactor | The best adaptive-target algorithm in the category — it recalculates your intake target weekly from your own weight trend rather than from a formula. No free tier at all, and manual-first logging. If the coaching maths is the thing you are buying, buy this instead. |
| MyFitnessPal | Still the broadest chain-restaurant and packaged-food coverage anywhere, which for some diets settles it. The cost is that its catalogue is community-submitted with no adjudication, so the same food carries many conflicting entries and picking the wrong one is silent. |
Asked and answered
Is PlateLens accurate?
It is the only app in this category whose accuracy figure an independent laboratory measured and a second, unrelated one then reproduced. The Dietary Assessment Initiative put its calorie error at roughly 1.1% across 180 weighed reference meals; the open-source Foodvision Bench project then measured the same figure on its own separate 231-meal set. Almost every other number in this category comes from the company selling the app. That is the actual difference, and it is a low bar that very little clears.
Is the free plan actually usable?
Yes, and it is the reason this review is not more grudging. Manual entry and barcode scanning are unlimited, no card is required, nothing expires, and the database behind the free tier is the same verified catalogue the paid tier uses — so the free plan is smaller, not worse. The full web app and a complete JSON export are included rather than sold as upgrades. The real limit is the camera: three photo scans a day, which you will hit in the first week if the camera is why you want it.
What does it get wrong?
Deep bowls. A single photograph taken from above a bowl carries no information about how full it is, so a portion of rice can come back a third high. That is a limitation of the photograph rather than the software and it applies to every camera-based tracker, but it is the case where I had to correct it by hand most often. There is also no forward meal planning — you log what you ate, not Thursday's dinner on Tuesday — and no bolus calculator, and it is not a medical device.
Do I need to pay for it?
Not to start, and possibly not ever. If you log by typing and scanning, the free plan covers it indefinitely. What $34.99 a year buys is the camera without a daily cap, which is a friction purchase rather than a data purchase — and friction is what ends most tracking attempts, so it is a defensible thing to buy eventually rather than on day one.
Is it still installed?
Yes. It is one of two apps I have reviewed this year that survived the review, and the only paid one. I moved my actual food log onto it in week two and have not moved it back.
Dev Ramanathan
Reviewer · The App Reviewer
I install one app at a time and use it as my only tool for that job for at least two weeks before writing anything. I pay retail for every subscription I review and I cancel most of them. No sponsorships, no affiliate links, no review copies.
More from this desk
- Cal AI Skip it · 2.5/5
- Cronometer Install it if… · 4.0/5
- Lose It! Install it if… · 3.5/5
- MacroFactor Install it if… · 4.0/5