"Where should we eat?" is now a question a growing share of diners ask an AI assistant instead of a friend. 22% of US diners have used ChatGPT or Gemini to choose a restaurant, and 45% now use AI for local business recommendations generally — up from just 6% a year ago. AI has effectively caught up to Yelp as a discovery channel: 26% TikTok/Instagram, 24% Yelp, 20% AI search.
There's no page two
The data on AI restaurant recommendations describes a winner-take-most market. The top three brands per category hold 53.4% of AI share of voice. 80% of restaurants get cited at least once — but only around 15% are ever the actual top recommendation. Being mentioned and being the answer are different games, and 83% of restaurant locations never appear in AI-generated recommendations at all, despite 86% maintaining a Google presence. Google visibility does not transfer to AI search visibility.
Your reviews aren't for humans anymore — they're training data
AI-recommended restaurants carry roughly 3.6× more Google reviews than comparable non-recommended restaurants, and average a 4.3 star rating on ChatGPT specifically (Perplexity ~4.1, Gemini ~3.9). But volume alone doesn't win: a review that says "great place, 5 stars!" teaches an AI model nothing. A review that says "exceptional truffle risotto, perfect for anniversaries, attentive staff" teaches it dish, occasion, and service — content it can actually extract and repeat. Recency matters more than volume: 40 reviews in the last 60 days can outrank 400 reviews collected over five years.
It's reading your listings, not your homepage
Over 41% of the sources AI cites when recommending restaurants come from listing platforms like DoorDash — not the restaurant's own website. Restaurants with proper Restaurant, Menu, and LocalBusiness schema markup have a 2.5× higher chance of appearing in AI answers. The economics were already real before AI (a one-star Yelp increase moves revenue 5–9%), but AI has made reviews and structured listings the whole interface between a restaurant and a hungry diner.
These aren't browsers. They're buyers.
AI search traffic converts at roughly 14.2%, versus 2.8% for Google organic — nearly five times higher. 61% of diners say eating out now feels like a special occasion, and 54% will pay a premium for a one-of-a-kind dining experience — exactly the high-intent decisions people research with AI first. And the window is closing: only 39% of operators have done anything at all to optimize for AI, and what they're doing is basics — menus, reviews, photos. Almost nobody is doing the citation-layer work yet.
The mechanics that fix this — crawler access, schema, extractable content — are the same ones covered in the hospitality GEO checklist, and the same ones that moved Arros QD from invisible to cited by Perplexity directly.