How I test, and why a home cook is the right person to do it

Who's writing this

I'm Marisol Vega. I'm a home cook and food writer — the person at the table photographing her plate before anyone's allowed to eat. I grew up in a kitchen where dinner was pozole, mole, a pot of beans that simmered all afternoon, and exactly zero of it came in a package with a barcode. So when calorie apps started promising they could "just read your food from a photo," I was curious and a little skeptical. Most of what I cook doesn't exist in any food database.

I am not a registered dietitian, a nutritionist, or a doctor, and I won't pretend to be. What I'm good at is cooking real food from a lot of different cuisines, then pointing a phone at it and noticing — honestly — when an app gets the dish right, when it guesses, and when it quietly gives up and asks me to scan a barcode that was never going to exist.

The Dish Decoder exists because most "best calorie app" lists are written by people who tested the apps on a granola bar and a can of soda. That's the easy case. The hard case — the one that actually matters if you eat like most of the planet — is a home-cooked, mixed, sauce-covered, gloriously un-packaged plate of food. That's the only thing I test.

How I test, with real food

Every app on this site goes through the same kitchen-table gauntlet:

• Real, home-cooked meals. I log the food I actually make and eat: Mexican, Indian, several Asian cuisines, plus weeknight leftovers and ambiguous bowls. No demo food, no staged "perfect" plate, no single-ingredient cop-outs.
• No barcode allowed. The whole point is food that isn't packaged. I deliberately avoid the barcode scanner so I can see how an app copes when there's nothing to scan — which is most of my fridge.
• Dishes that aren't in any database. I test things a US-centric food database has likely never heard of, and watch whether the app reasons about what the dish is or just throws up the nearest American fast-food match.
• Every logging path. Photo / AI estimate, manual search, and the editing flow afterward — because the real question is how much fighting it takes to get a sane number into the log.
• Honest, qualitative accuracy. I describe how close estimates land and where they drift, in plain language. I don't publish invented precision — no made-up accuracy percentages, no fabricated study counts, no fake benchmark scores.

What I'm looking for

The apps I end up recommending all share one trait: instead of demanding a database entry, they reason about what the dish is from the photo and the context, give a sensible estimate even for food they've never seen, and — crucially — ask me to confirm when they're unsure rather than pretending to be certain. That confirm-when-in-doubt behavior is the difference between an app you can trust with real cooking and one that only works on a vending machine.

How I make money

Some links may be affiliate links, and they never change a rating or a ranking. If an app mangles my dinner, I'll say so whether or not there's a referral attached. Ratings are my own judgment from cooking real food and logging it, full stop.