What Menu Decoder actually does
One profile, three ways to scan a menu, and verdicts that never overclaim. Here's the full catalog.
Scan & verdicts
Photo, paste, or link
Snap a menu photo, paste menu text, or drop in a link — including QR-code menus and restaurant websites (HTML or PDF).
Dish-by-dish verdicts
Every dish gets exactly one of AVOID, CAUTION, UNKNOWN — ask staff, or NO FLAG, with the specific reasons behind it.
Hidden-ingredient knowledge base
A curated dish database catches what a dish's name doesn't say — satay's peanut sauce, hollandaise's egg and dairy, red curry's shrimp paste.
Verified vs. unverified
Curated database matches are labeled verified. AI inferences are labeled unverified, shown only above a strict confidence threshold, and can only add caution — never remove a flag.
UNKNOWN, shown loudly
When there isn't enough to judge confidently, the dish is marked UNKNOWN — ask staff, and the count is displayed prominently, never buried.
Ask-staff questions
Every CAUTION and UNKNOWN dish comes with specific questions worth asking your server, that you can check off as you go.
Diet programs
Ten programs, scored separately
Keto, low-carb, paleo, Whole30, Mediterranean, DASH, carnivore, high-protein, low-fat, and calorie-conscious — scored on their own fit scale, never the safety colors or words.
Top picks for your plan
The best-fitting dishes, ranked automatically — pulled only from dishes with nothing flagged. Multiple programs combine, so a pick has to fit all of them.
Pick a subset per scan
The Today bar lets you choose which of your saved programs apply to this one scan, without touching your saved profile.
Medications
Drug-food interactions as safety flags
Add medications in Settings and interactions come back as AVOID or CAUTION just like an allergy, with the reason spelled out — not buried in a separate list.
Alongside conditions, not instead of them
Health conditions (celiac, diabetes, low-sodium, and more) and medications live in their own Settings sections and combine with your allergies in a single set of verdicts.
Chef cards & 19 languages
Chef card, built from your profile
Your saved allergies and avoid list turn into a large-type card the kitchen can read at a glance — nothing new to fill in.
Country picker
Tell it where you're dining and it suggests the right language automatically, or pick a language yourself from the full list.
19 app languages
The entire app interface — not only the chef card — is available in 19 languages, switchable anytime from Settings.
Recipes & saved
Typical recipes
Look up how a flagged dish is usually made — ingredients and steps — to understand exactly why it was flagged and what a kitchen would need to change.
Saved cards & recipes
Save a menu's summary card or an individual recipe, filed under Saved with filters for All, Cards, and Recipes.
Stays on your device
Everything you save lives on your device, ready the next time you're back at the same table.
Speed & honesty
Instant results, then a double-check
You get a deterministic answer from the knowledge base first, in seconds — an AI double-check refines it quietly in the background and can only add caution, never take one away.
Built to a 3-5 second budget
Every tap responds within about a second, with a full answer inside three to five seconds — nobody should wait on a model with a table full of people.
The honesty vocabulary
The whole product is built around never overclaiming. See exactly what each verdict means below.
Four words, and only four words
Menu Decoder never tells you a dish is free of risk. This is the exact vocabulary it uses instead, everywhere in the app.
AVOID
A direct match to your profile, or a known trap dish. Treat it as a hard no unless staff confirms otherwise.
CAUTION
A likely conflict — a cuisine, prep style, or cross-contact risk worth a question before you order.
UNKNOWN — ask staff
Not enough information to judge confidently. Shown loudly, every time — silence is never a safety claim.
NO FLAG
Nothing in the menu text matched your profile — a statement about our reading of the menu, not about the food.
Verified and unverified mean different things — say so, every time
A verified flag comes from our hand-curated, human-reviewed database. An unverified flag comes from an AI reading the menu text, shown only above a strict confidence threshold, and only ever allowed to add a caution — never to remove or downgrade a verified one. You always know which kind you're looking at.
Every result carries this line, and always will:
"Menus change and kitchens share equipment. Always confirm with staff. This app does not provide medical advice."