Most numbers about restaurants and AI search get passed around with no source, no date and no country attached, which makes them useless the moment anyone checks. This is the sourced version. Every figure below names who measured it, when, and in which market, so you can quote it with confidence and know exactly where it holds and where it is only directional. The short story: the invisibility numbers are large and consistent across independent studies, almost all the hard data so far is from the United States, and the demand-side shift is already measurable in the UK and Germany.
TL;DR
- Independent US audits put restaurant invisibility in AI answers at roughly 75% for Google AI Overviews and around 83% for ChatGPT.
- A higher star rating does not protect you: 4.8-star venues were hidden more often than 4.5 to 4.7 star ones in the largest audit.
- Demand is real and UK-specific: 26% of UK consumers already use AI apps to help decide where to eat or drink.
- Almost all invisibility data is US-sourced. Treat the percentages as directional for the UK and Germany until local studies exist, which is a gap worth naming rather than papering over.
How many restaurants are invisible in AI search?
Across every independent audit run so far, most of them, with remarkably consistent numbers given the different methods. Here is the sourced picture.
| Statistic | Figure | Source | Date | Market |
|---|---|---|---|---|
| Restaurants never appearing in Google AI Overviews | 74.9% | Local Falcon (10,000 restaurants) | May 2026 data | US |
| Restaurants invisible on ChatGPT | 83% | Local Falcon (approx 190,000 ChatGPT results vs 16.4M Google) | March 2026 | US |
| Restaurants invisible on Google (classic) in the same study | 14% | Local Falcon | March 2026 | US |
| Multi-location restaurant locations invisible across ChatGPT, Gemini, Perplexity, Copilot and AI Overviews | 83% | Uberall GEO benchmark | May 2026 | US |
| Share of locations ChatGPT recommends vs 35.9% in Google's local 3-pack | 1.2% | SOCi Local Visibility Index (350,000+ locations) | Jan 2026 | US |
| Share of locations Gemini recommends | approx 11% | SOCi | Jan 2026 | US |
| Share of locations Perplexity recommends | approx 7.4% | SOCi | Jan 2026 | US |
Four different firms, four different methods, and the same conclusion: a large majority of restaurants are simply not present when an AI engine answers a where-to-eat question. We walk through what causes it and how to fix it in why your restaurant is invisible on ChatGPT.
Does a higher star rating mean more AI visibility?
No, and this is the most counterintuitive result in the data. In the Local Falcon audit of 10,000 restaurants, the best-rated venues were hidden more often, not less.
| Statistic | Figure | Source | Date | Market |
|---|---|---|---|---|
| Restaurants rated 4.8 stars and above, invisible in AI Overviews | 69.0% | Local Falcon | May 2026 data | US |
| Restaurants rated 4.5 to 4.7 stars, invisible in AI Overviews | 60.0% | Local Falcon | May 2026 data | US |
| Restaurants with more than 1,000 reviews, still invisible | 70.9% | Local Falcon | May 2026 data | US |
The likely reason is that the highest-rated venues tend to be newer, with thinner review histories an engine hesitates to trust. The full explanation is in the star rating trap.
How many diners actually use AI to choose where to eat?
Enough that this is no longer a fringe behaviour, and here the data is genuinely local rather than borrowed.
| Statistic | Figure | Source | Date | Market |
|---|---|---|---|---|
| UK consumers who use AI apps to help decide where to eat or drink | 26% | CGA by NIQ | Aug 2025 | UK |
| UK consumers using Google Maps for the same, for comparison | 27% | CGA by NIQ | Aug 2025 | UK |
| UK diners who trust AI-generated review summaries | 60% | CGA by NIQ | Aug 2025 | UK |
| UK diners who would leave a review if an AI assistant asked at the right moment | 51% | CGA by NIQ | Aug 2025 | UK |
| German guests who use digital platforms to find a restaurant | approx 70% | YouGov, commissioned by METRO and DISH | Jan 2025 | Germany |
| German guests, of those, using classic search engines | 51% | YouGov, commissioned by METRO and DISH | Jan 2025 | Germany |
AI-assisted restaurant choice in the UK is now essentially level with Google Maps, and the German digital-discovery base is large too. The German-market detail is in our Germany GEO guide for hospitality.
What is actually live: AI Mode and agentic booking
The channel context that makes the numbers matter, with dates.
| Development | Detail | Date | Market |
|---|---|---|---|
| Google AI Mode live in UK Search and the Google app | Live | July 2025 | UK |
| Google AI Mode can find and book a UK restaurant table directly | Via TheFork, OpenTable, SevenRooms, ResDiary, DesignMyNight | April 2026 | UK |
| Google AI Overviews in Germany | Signed-in users aged 18 and over | March 2025 | Germany |
| Google AI Mode in Germany | Live | Oct 2025 | Germany |
| TheFork integrated into ChatGPT for booking | Selected European countries; no confirmed Germany date | Dec 2025 | Europe |
The honest gap in the data
Almost every invisibility statistic above is US-sourced. There is, as far as we can establish, no recurring UK or German study measuring how many local restaurants appear in AI answers, and no one anywhere is tracking it as a daily time series rather than a one-off snapshot. So the responsible way to use these figures is as directional for a British or German market, not as verified local truth. That gap is precisely why ongoing, market-specific measurement matters: the only way to know your own number is to run the real queries your diners use, across the real engines, and record who gets named.
How to cite these figures
Use the source, date and market columns as given. When quoting an invisibility percentage outside the US, say "US data, directional for the UK" rather than presenting it as a local finding. The demand-side UK and German figures (CGA by NIQ, YouGov for METRO) can be cited as local. For the difference between this kind of AI-answer visibility and classic search ranking, see AEO vs GEO vs SEO.
Disclosure: Schmitdy is our own AI search service at AI Heroes, so treat the offer here as the maker's case rather than neutral advice. We track these engines daily for the restaurants and groups we work with, which is how we know the snapshot numbers above move over time and vary by city and cuisine. If you want your own figure rather than a borrowed one, the free AI search audit shows how often the major AI engines name your venue for the questions your diners actually ask.
