Anatomy of a Restaurant Search in ChatGPT
How ChatGPT ranks, cites, and retrieves restaurants: from the pre-classifier to query fan-out and source fusion.
When a diner types "best Italian restaurant in Bordeaux" into ChatGPT, a lot happens in a few seconds. If you know the pipeline, you know where your restaurant gets picked or skipped.
1. The pre-classifier
Before the big model answers, a small classifier makes the call: is this a "local" query? Does it need a web search? Does it need images? For a restaurant query, the web search fires almost every time.
2. Query fan-out
ChatGPT does not send your question as-is. It breaks it into 5 to 7 variants sent in parallel to different providers: "best Italian Bordeaux", "trattoria Bordeaux reviews", "Italian restaurant Bordeaux reservation"… Your visibility depends on being present across all of those variants.
3. Data providers
For restaurants, most of the data comes from Google (Web Search + Places/Business Profile), plus TripAdvisor, TheFork, guides, and local press. The Google Business Profile acts as the entity anchor: it ties "your name" to a listing the model can identify.
4. Entity recognition
The model tries to match a cited name to a unique identifier (Google Place ID). A distinctive name ("Chez Marguerite") matches far more cleanly than a generic one ("Le Bistrot"), which gets lost among dozens of namesakes. Your name is a visibility factor.
5. Source fusion (RRF)
Results from the different variants get merged by a Reciprocal Rank Fusion algorithm. The result is counter-intuitive: being present everywhere, ranked decently, beats being first on a single source. A restaurant that places 3rd on Google, 2nd on TheFork, and 4th on TripAdvisor will rank above one that places 1st on Google and nowhere else.
6. Response generation
The model then writes a natural-language answer, usually with entity cards, ratings, and links. This is where perception gets decided: which attributes it pulls forward ("romantic setting", "generous cuisine").
What to take away
To show up in ChatGPT, a restaurant needs a Google listing kept tight, a coherent presence across several sources, a name nobody confuses with another, and clean structured data. RestoRank measures each of those signals and tells you which one to fix first.
Is your establishment recommended by AIs?
Run the free audit and discover your Visibility Index across the five major AI search engines.
Run the audit