Your restaurant profiles disagree. AI and search have to choose one.

Why does an AI assistant or search result get basic facts about a restaurant wrong? Usually because the restaurant's own public record disagrees with itself. Your name, hours, location, menu, prices, cuisine, reservation path and dietary answers exist on many surfaces at once — your website, Google Business Profile, Maps, delivery apps, reservation platforms, review sites, old press. When those surfaces conflict, every answer engine has to pick one version, and it will not necessarily pick yours. The wrong Tuesday hours on one listing become somebody's wasted trip and a one-star review about a fact, not the food.

This is an ownership problem wearing a technology costume, and it is fixable with a quarterly audit and a named owner.

The entity, in plain terms

To search and AI systems, your restaurant is an entity: a bundle of facts assembled from every source that describes you. The stronger and more consistent the sources, the more confidently a system answers; where sources conflict, it guesses, hedges, or quietly prefers the source it trusts most — which is frequently a large platform's listing rather than your website. That has a blunt commercial corollary: third parties are describing your business to your customers, and the description is only as current as your least-maintained listing. In DoorDash and SevenRooms' 2026 US research, 41% of AI restaurant recommendations were sourced from listing platforms (US, vendor data) — the answer is built substantially from surfaces you may not have opened in a year.

The eight facts that drift

1. Official name — "Náme Restaurant", "Name — Downtown", "Name by Chef X": three strings, one venue, machines unsure whether it's one entity or three.
2. Address and location pin — especially in malls, hotels and districts where the pin was set once, by someone, in a hurry.
3. Opening hours — the classic. Ramadan hours, seasonal terraces, a closed Monday added operationally but never digitally. "Google says open, the door says closed" is the single fastest way to convert marketing spend into resentment.
4. The menu and prices — a conversion document in its own right; an old copy on an aggregator can outrank the current one on your site, so the "wrong" price a guest saw was really your unmaintained listing.
5. Cuisine and category — the labels systems use to match you to "best Italian near me." A miscategorised profile is invisible for exactly the queries you want.
6. Reservations — does the profile's booking action point at your current provider, the old one, or nothing? A booking button pointing at a dead widget is a silent leak at the highest-intent moment.
7. Dietary and policy answers — vegan options, alcohol service, shisha, kids, terrace smoking: the questions people genuinely ask assistants, answerable only from whatever text exists somewhere.
8. Delivery and ordering links — which platforms, which menu version, and whether "order direct" exists at all.

What structured data can and cannot do

Restaurant structured data on your own site — cuisine, price range, hours, menu and reservation properties — is worth maintaining: it labels your facts unambiguously on the pages you control. What it cannot do is override a contradiction. Markup describes; it does not adjudicate. If your site says one thing and three high-authority listings say another, no schema resolves that in your favour — and Google's own guidance is that its AI features require no special schema beyond ordinary search fundamentals (Google Search Central; re-verify wording before relying on it). Anyone selling "AI schema" as the fix for wrong answers is selling markup for an agreement problem. The fix for disagreement is agreement.

The Katalyst Restaurant Entity Accuracy Audit

A Katalyst method — run it quarterly, and after every menu change, seasonal shift or platform migration:

1. List the surfaces. Website, Google Business Profile, Maps, each delivery app, each reservation platform, the top review sites, social bios. Most operators find more surfaces than they expected, including ones nobody remembers creating.
2. Pull the eight facts from each into one sheet. The audit is literally a table: surfaces down, facts across.
3. Mark every disagreement. No judgment yet — just the conflicts.
4. Fix in priority order: hours and location first (wasted-trip risk), then reservation path (revenue risk), then menu and prices (expectation risk), then category and dietary answers (visibility risk).
5. Ask an assistant. Put five real guest questions to an AI assistant — "is X open Monday?", "does X take reservations?", "does X have vegan mains?" — and note what it says and, where shown, which sources it leaned on. One answer is an observation, not a verdict; a wrong fact repeating across surfaces points at a source you can now find in your sheet.
6. Name the owner. One person owns the sheet and every future change propagates through it. Facts drift at the speed of operations; the audit is only as good as its maintenance.

What this is not

It is not a promise about rankings, AI mentions or "AI visibility scores." Answer engines are not deterministic and nobody controls their output. Entity accuracy is narrower and more honest: it makes the correct facts the easiest facts to find, everywhere, so that whichever source gets chosen, it agrees with your door. That is also the precondition for everything upstream of it — an accurate entity cannot rescue a weak concept, but an inaccurate one can quietly tax a strong one, which is why this audit sits inside the Restaurant Discovery Chain at the validation link.

Sources and limitations

DoorDash × SevenRooms 2026 Restaurant Industry Trends (Dynata, US consumers — vendor data) for the listing-platform sourcing figure, labelled US. Google Search Central for the no-special-schema position and structured-data role; Google Business Profile documentation for profile menu and booking capabilities — platform documentation changes and must be re-verified before publication-day use. The audit is a Katalyst method; the failure patterns are operator-side observations, not statistics. No claim is made about how often any platform's data is wrong — the audit exists precisely because the only rate that matters is yours.

The F&B Growth & Margin Diagnostic runs this audit as standard, alongside menu, discovery and booking-path checks — part of the F&B Growth & Revenue practice. Hotel readers: the same discipline for properties is in hotel AEO is not FAQ schema.