2026 Study: The Footprint of French Restaurants in AI

We ran hundreds of diner queries through ChatGPT, Gemini, Perplexity, Claude, and Google AI Overviews across several French cities, then parsed the answers. One pattern stands out: AI visibility is extremely concentrated.

Methodology: location × persona × topic prompts, multiple runs per model, structured parsing of mentions, ranks, competitors, and sources. Figures are indicative and will be updated as our measurements accumulate.

1. A handful of names captures almost everything

On a general query ("best restaurant in [city]"), AIs return on average 3 to 4 names. Across an entire city, fewer than 10% of establishments capture almost all the mentions. For the other 90%, the AI is a wall.

2. Aggregators dominate the sources

In most responses that cite a source, the link points to TheFork, TripAdvisor, or Google rather than the restaurant's own website. By default, "AI distribution" passes through paid intermediaries.

3. The name matters more than people think

Restaurants with a distinctive name get linked as entities far more reliably than generic ones. A good share of the "invisible" ones are invisible because of a disambiguation problem, not a quality problem.

4. High variability between models

Overlap among the "top 5" across platforms is low. A restaurant can be first on Perplexity and absent from Gemini. Tracking a single AI gives a misleading picture.

5. Review freshness makes the difference

Restaurants with a steady flow of recent reviews and owner responses show up more often. Models weight that signal heavily.

What this means for you

If you do nothing, your odds of landing in the visible 10% stay low, especially without a distinctive name, clean structured data, and presence on several sources. The good news: these levers are actionable, and the gap is widening in favour of restaurants that act now.

Start by measuring where you stand with the free RestoRank audit.