Get Your Restaurant Recommended by ChatGPT
Updated on 5 September 2026
In short
- Google's page ranking and generative AI's (ChatGPT, Perplexity, AI Overviews) source selection work differently — doing well in one doesn't guarantee the other.
- An AirOps study (548,534 pages, 15,000 prompts) found ChatGPT cites only ~15% of the pages it retrieves.
- Only 12% of URLs cited by ChatGPT rank in Google's top 10 — classic SEO is a necessary base, not a sufficient one.
- Most AI visibility happens as brand mentions with no clickable link, not as measurable traffic.
Why this is a distinct question from classic SEO
Google ranks a list of pages. A generative AI reads several sources, keeps a fraction of them, and synthesizes a single answer. These are two different mechanics, optimized differently. A restaurant can rank well on Google and be absent from ChatGPT's answers, or the reverse.
This distinction matters increasingly: Malou, the leading French review-management software vendor, is now introducing "share of traffic from generative AI" as a 2026 metric, citing that one in five consumers already chooses a restaurant via generative AI.
How AI decides who to recommend
An Ahrefs study covering 75,000 brands (December 2025) identifies the factors most correlated with AI citation:
| Factor | Correlation strength |
|---|---|
| YouTube mentions | Strong |
| Brand mentions on the web, even without a link | Strong |
| Classic backlinks | Weak to negligible |
The shift in logic matters. In classic SEO, an inbound link counts. In AI visibility, what counts is being mentioned by name, even without a link — the model memorizes the association between an entity and its attributes ("SEOresto = review management software for restaurants in France, €39/month").
The three levels of AI visibility
- Citation — a link to your page appears in the answer.
- Mention — your name appears in the answer's text, without a link.
- Recommendation — the model actively proposes you as a solution.
Most visibility happens at level 2: mentions without a link, so no measurable traffic in classic analytics tools, but a real effect on the customer's decision.
What makes a page citable
An extractable paragraph. The model looks for a passage it can insert into its answer almost without rewriting it: name, who it's for, what it does (3-5 points), price in numbers, one honest caveat.
The answer before the explanation. The first 2-3 sentences after a heading should contain the complete answer — the model often takes only the start of a section.
Dated, attributed numbers. The format most reused: "According to public pricing on [source], as of [month year]: [figure]." Source + date + figure.
Comparison tables. Tabular data parses almost losslessly, unlike the equivalent paragraph in prose.
Freshness visible in the text, not just in metadata — an explicit date, a year in the title.
Entity consistency: name, description, price, category identical everywhere your business is mentioned (website, social media, directories, reviews).
What not to do
Don't hide your price behind a mandatory sales demo: a model can't cite a figure it can't find anywhere. Don't rely on llms.txt or "AI-special" schemas: Google has confirmed they're not used for generative features. Don't measure your AI visibility by traffic: most of the effect happens through mentions with no click.
How to check where you stand today
Ask ChatGPT, Perplexity, and Google's AI Overviews a series of questions a real customer would ask ("what Italian restaurant in [city]", "what software to manage my Google reviews"), around twenty prompts, once a month. Note who gets cited, mentioned, or recommended. It's the most reliable method on a zero budget, before considering paid tracking tools.
For the full technical detail (schema.org, robots, AI bot access), see structured data for restaurants: 2026 guide and should you block AI crawlers?. For the underlying difference between GEO and SEO, see GEO vs SEO for restaurants.
SEOresto structures your review responses so they also work for your visibility in generative AI. From €39/month. See pricing.
Sources: Ahrefs research on 75,000 brands (Dec. 2025) via rzlt.io; AirOps research cited by erlin.ai; Google's public statements on llms.txt.
Author: Dmitrii Portnov, founder of SEOresto.
