Restaurant recommendations have a new source.
Restaurant recommendations have always depended on who was asked.
A critic, a concierge and a friend who never seems to have a bad meal might each send you somewhere different. Their answers reflect what they know, what they value and which version of the city they have experienced.
Now diners are asking ChatGPT, Claude, Gemini and Perplexity.
These systems can produce a confident list of restaurants in seconds. But they have never tasted the food, sat in the dining room or watched the way a restaurant handles an important occasion.
They have to reconstruct the experience from what has been documented about it.
The AI Scene studies the recommendations that emerge from that process.
We do not review restaurants. We review how AI understands them.
Before I began researching AI recommendations, I spent more than a decade working in food, wine and hospitality.
I sold Italian wine to sommeliers in San Francisco, represented a Napa Valley estate winery and worked with Nespresso at James Beard dinners and Michelin events. For eight months, I joined the wine team at a Michelin-starred restaurant.
Those years taught me that the qualities that make a restaurant worth choosing are rarely captured by a category or accolade alone.
The reason one place belongs in an anniversary recommendation but not a business dinner may come down to the pace of service, the energy of the room or how personally the restaurant makes guests feel. These distinctions are obvious once someone has experienced them. They are much harder for an AI system to infer from the available digital record.
When I began collecting restaurant recommendations across multiple AI platforms, I saw how unevenly those distinctions traveled.
A restaurant could appear constantly for one dining occasion and disappear from another it was well equipped to serve. Some names were recommended across every platform. Others were heavily favored by one and nearly absent from the rest.
One isolated AI response could be dismissed as an anecdote. Hundreds of responses began to reveal patterns.
I created The AI Scene to document them.
Each edition focuses on one city and dining category during a defined collection period.
A consistent set of questions, modeled on actual dining decisions, is tested repeatedly across ChatGPT, Claude, Gemini and Perplexity. The questions cover different occasions and preferences because "Where should I eat?" is rarely the full question a diner is trying to answer.
The results are published in two parts.
The Restaurant Index records which restaurants appeared most frequently and shows how their visibility differed by platform.
What the Machines Said examines the patterns behind those numbers: where the platforms agreed, where they diverged and which kinds of restaurants each appeared to understand particularly well or poorly.
The objective is not to produce another list of the city's best restaurants. It is to observe how a new recommendation system represents the dining landscape.
A restaurant's position in the Index is not a score for food, service or overall quality.
The restaurant ranked first was recommended most frequently within that edition's research. It was not declared the best restaurant in the city.
A lower position does not mean a restaurant is less deserving. It means the restaurant appeared less often under the questions, platforms and collection conditions used for that edition.
AI responses also change. Platforms update their models, retrieval methods and available sources. The same question can produce different recommendations from one run to the next.
That is why every edition is dated and every question is tested more than once. The results are a documented snapshot, not a permanent hierarchy.
The research does not tell us whether a diner made a reservation, changed an existing preference or spent money because of an AI answer. It establishes something narrower: which restaurants the platforms selected when asked to respond to particular dining needs.
The AI Scene is an independent research publication created by Ally Kiel, founder of Ally Kiel Consulting.
Ally works with food, beverage and hospitality businesses to understand how AI platforms recognize and recommend them. Her background in hospitality shapes the editorial analysis behind each edition.
The numbers establish what happened in the research. Industry experience helps determine which patterns are meaningful, where an AI description reflects the restaurant accurately and where the experience has been flattened into something incomplete.
For research inquiries or a private investigation into how AI platforms currently understand your restaurant, visit allykielconsulting.com.