Lifesum: AI Calorie Tracker reviews

Buried in settings is a toggle that disables AI and restores the old search-with-database flow; those who find it instantly get the fast input and accuracy back. It's an escape hatch that retains a slice of would-be churn.

The real retention engine isn't AI but the catalog: almost any brand is in the database, the barcode pulls a product instantly, and it works across many languages with no wall on the basic flow. This is exactly what brought people over from MyFitnessPal and kept them for 5-10 years.

Even scathing reviews almost always note the interface is pleasant, the main-screen navigation tidy, the colors and visuals enjoyable. That aesthetic is an emotional anchor: it pulled people to premium and makes them give the app «one more chance» instead of deleting instantly.

The strongest value signal is people who log hundreds of days straight and drop 21-150 pounds, crediting the result to daily logging in Lifesum specifically. These streaks aren't a metric, they're the product: as long as the app made repeat logging frictionless, it was effectively selling discipline. Any change that lengthens the daily gesture hits this asset directly.

The «+» used to open a list and logging took seconds; now the same button routes into an AI chat where you describe food in words. The muscle-memory reflex «open — pick yesterday's — log» is replaced by a dialogue that types one letter at a time and gets things wrong. For logging meant to repeat 5x a day forever, the extra steps kill the habit itself — people quit not over AI quality but because the fast path vanished.

A calorie tracker sells exactly one thing: you can trust the number. The same food gets different values run to run (carrot — protein from 4 to 12g, coffee — 7000 cal), and the user loses the reason to keep a diary at all: if numbers are invented, the diet is built on noise. This hits the product's core — not «inconvenient» but «pointless».

For a whole paying segment the IF timer was the core of usage — not calories, the fasting window. It vanished with no warning and no way to bring it back, and these people aren't «disappointed in a feature» — their reason to subscribe disappeared mid paid-term. Deleting an anchor feature from a paying segment turns manageable churn into resentment.

Logging meals jointly with a partner wasn't a nicety but a social lock-in: two accounts hold each other and neither leaves. Removing sharing turned the couples scenario into manual double-entry of the same food — and now both look for a new app at once. The stickiest cohort became the angriest.

Fusing photo recognition and the barcode scanner into one button sounds like simplification, but in practice the camera stopped focusing on barcodes: people describe a workaround of «add new food → scan → confirm item exists». The single most frequent daily gesture — barcode scan — became a quest because it got subordinated to the AI flow.

People eat the same things day to day — the base pattern that makes logging effortless. The AI version won't let you favorite a meal, doesn't remember ingredients individually, and suggests not «lunch for lunch» but «yesterday's dinner». Every repeat is forced through a fresh AI query — the product fights the very repetition that makes tracking sustainable.

The app auto-sets dangerously low daily goals, and a newcomer can't change them. For someone who doesn't know nutrition, this isn't just a limit — it's a risk: a health product nudges toward an unhealthy deficit. The disordered-eating segment is hit by this mechanism directly — trust in a «healthy» app turns into a hazard.

The keto/diabetic segment came for one thing: accurate net-carb accounting and sugar tracking. Here the AI errs critically (chia as 10 net carbs instead of zero), and you can neither exclude fructose recipes nor reliably fix macros. For a user whose blood sugar spikes, a wrong number isn't an inconvenience but a medical risk — and the product loses its most motivated segment.

AI recognition routes every entry to the server, so the simplest food log hits 10-30 second delays and timeouts, and on mobile data it doesn't work at all. A food tracker by definition is needed at the moment of eating — in a cafe, on the road, off Wi-Fi; by tying the basic log to an online call, the product broke the scenario exactly where it matters.

A food tracker's audience is people who want verifiable numbers and control over their data, not a «guessing game». For them «AI» in the name and the forced upload of their diet to the cloud read as a negative, not a feature: they explicitly refuse to give AI access to their data and leave to find an AI-free alternative. A brand built on trust in the number slapped on a label that repels its own core.