Your restaurant is probably invisible on ChatGPT, and quite possibly missing from Google's AI Overviews too, for a simple reason: the pages that describe your food, your hours and your vibe aren't written in a form these engines can extract with confidence, and nobody is chasing the citations that get a name mentioned in an AI answer. A Google Business Profile and a page of five-star reviews used to be enough. It isn't any more, and UK diners are asking AI where to eat in growing numbers while the tools that used to warn you about a visibility problem, like Google Search Console, have no idea this one exists.
TL;DR
- More than a quarter of UK diners already use AI apps to help decide where to eat or drink, level with Google Maps.
- Google's AI Mode has been live in the UK since July 2025, and from April 2026 it can book your table directly through TheFork, OpenTable, SevenRooms, ResDiary and DesignMyNight, no click to your website required.
- Independent US audits found that roughly three in four restaurants never appear in Google AI Overviews, and a similar share are invisible on ChatGPT.
- Star rating doesn't save you: the very highest-rated venues were hidden from AI Overviews more often than solidly good ones, not less.
- Five fixes close most of the gap: crawlable menus and schema, one consistent identity everywhere, review platforms the engines actually read, editorial "best of" citations, and genuine community presence.
How many UK diners are actually asking AI where to eat?
About a quarter already, and that's rising fast. In CGA by NIQ's August 2025 UK hospitality survey, 26% of British consumers said they now use AI apps to help decide where to eat or drink out, a usage rate level with Google Maps (27%). That's not a niche behaviour any more. It's someone typing "wine bar near London Bridge" or "where to eat in Shoreditch tonight" into ChatGPT instead of scrolling Google, and getting a shortlist that either includes your restaurant or doesn't, with no obvious way for you to find out which.
The same CGA data shows diners are starting to trust what AI tells them about a venue almost as much as a review: 60% already trust AI-generated review summaries, and 51% said they'd leave a review if an AI assistant asked at the right moment. The behaviour is shifting on both sides of the search, discovery and reputation, at once.
Why does Google's AI Mode change things specifically for restaurants?
Because it's not a search experiment any more, it's a booking channel. Google rolled AI Mode out to the UK in July 2025, putting it as a tab on Search results and inside the Google app. Then, in April 2026, Google went further: AI Mode can now find and book a UK restaurant table directly, through reservation partners including TheFork, OpenTable, SevenRooms, ResDiary and DesignMyNight.
That's the part worth sitting with. If a diner asks for "best Sunday roast in Clapham" and AI Mode can complete the whole job, find a table, check availability, book it, without ever landing on your website, then your website stops being the moment that wins or loses the booking. The moment that matters happens earlier: does the engine even know your restaurant exists, and does it trust what it knows enough to suggest you?
How many restaurants actually disappear from AI answers?
Most of them, according to every independent audit run so far, though every one of these studies to date has sampled US restaurants rather than UK ones, so treat the exact percentages as directional for a British high street rather than a verified local number.
A Local Falcon audit of 10,000 US restaurants, built from data collected in May 2026, found that 74.9% never appeared in Google's AI Overviews for any of the searches tested. The counterintuitive part: restaurants rated 4.8 stars or higher were hidden more often (69.0% invisible) than solidly good ones rated 4.5 to 4.7 stars (60.0% invisible). Local Falcon's read is that the very top-rated spots tend to be newer, with thinner review histories, exactly the kind of thin signal an AI engine hesitates to trust. Being excellent doesn't protect you if the engine can't yet verify it.
A separate Local Falcon study, published in March 2026, compared close to 190,000 ChatGPT results against 16.4 million Google results for restaurant-type queries and found 83% of restaurants completely invisible on ChatGPT, against just 14% invisible on Google. Uberall's May 2026 GEO benchmark of multi-location restaurant brands landed on the same headline number from a different angle: 83% of restaurant locations were entirely invisible across ChatGPT, Gemini, Perplexity, Copilot and Google AI Overviews combined. And SOCi's own January 2026 Local Visibility Index, covering more than 350,000 locations, found ChatGPT recommended just 1.2% of the brand locations it evaluated, against a 35.9% average appearance rate in Google's local three-pack. Gemini recommended around 11% and Perplexity around 7.4%, both still a fraction of what a restaurant expects to see in ordinary local search.
| Study | Engine(s) | Headline finding | Market |
|---|---|---|---|
| Local Falcon (May 2026 data) | Google AI Overviews | 74.9% of restaurants never appear; 4.8+ star venues hidden more often (69.0%) than 4.5 to 4.7 star venues (60.0%) | US |
| Local Falcon (March 2026) | ChatGPT vs Google | 83% invisible on ChatGPT vs 14% invisible on Google | US |
| Uberall (May 2026) | ChatGPT, Gemini, Perplexity, Copilot, Google AI Overviews | 83% of restaurant locations entirely invisible in AI-generated recommendations | US, multi-location/QSR |
| SOCi (Jan 2026) | ChatGPT, Gemini, Perplexity | ChatGPT recommends 1.2% of locations vs 35.9% appearance in Google's local 3-pack | US |
Why doesn't Google Search Console or my analytics show this?
Because there's no "impression" to log. Search Console counts a click or an appearance in classic search results, and your analytics count a visit. An AI Overview or a ChatGPT answer that mentions three competitors and skips you doesn't touch either system. Nothing fires, nothing drops, no alert appears. The only way to know you're missing is to actually ask the engines the questions your diners ask and watch what comes back, which is what our daily multi-engine tracking exists to do: run the real queries, across the real engines, and record who gets named.
The 5 fixes that close the AI-visibility gap
None of this needs a rebuild. It needs the same five things done properly, in an order that reflects what an engine actually needs before it will risk citing you.
| Fix | What it means | Quick win |
|---|---|---|
| 1. Crawlable menus and schema | Menu, hours, cuisine and price band in text and structured data, not locked inside a PDF or an image | Add Restaurant and Menu schema, publish menu items as real HTML text |
| 2. Entity consistency | Identical name, address, phone number and cuisine description everywhere you're listed | Audit your Google Business Profile, website and every booking platform for mismatches |
| 3. Review surfaces engines read | Reviews on the platforms AI engines actually retrieve from, not just the ones you check | Prioritise Google reviews and TripAdvisor, ask happy tables at the table, not by email three days later |
| 4. Editorial citations | Getting named in "best of" and "where to eat" roundups that engines cite as sources | Pitch local food writers and trade press with a genuinely specific angle, not a press release |
| 5. Genuine community presence | Real mentions on Reddit, local Facebook groups and neighbourhood forums | Answer local subreddit and forum threads honestly, as the owner, when someone asks for recommendations |
Crawlable menus and schema. If your menu only exists as a photographed PDF or an image carousel, an AI engine can't read your dishes, your prices or your dietary options, and it will quietly recommend a competitor whose menu is plain text with Restaurant schema attached. This is the cheapest fix on the list and the one most restaurants skip.
Entity consistency. Engines cross-check a restaurant across sources before they'll mention it, and a mismatch reads as risk, not as a rounding error. A London wine bar group we work with had three different phone numbers and two spellings of one venue's name spread across its own website, its Google Business Profile and its TheFork listing. That's exactly the kind of inconsistency that makes an engine unsure enough to leave you out rather than risk citing something wrong.
Review surfaces the engines actually read. More stars doesn't automatically mean more visibility, as the Local Falcon 4.8-star finding shows. What seems to matter more is depth and recency of reviews on the platforms engines retrieve from, alongside your own site and listings. Chase volume and freshness on the surfaces that matter, not just a five-star average on the one you personally check.
Editorial citations. AI engines lean heavily on "best of" and roundup content when they answer discovery queries like "best Sunday roast in Clapham". Getting into those lists, through genuine local food press coverage rather than a paid placement, is one of the highest-leverage moves available, because one citation can feed dozens of different AI answers for months. For the fuller picture of how this differs from classic SEO, see our guide to AEO vs GEO vs SEO.
Genuine community presence. Reddit threads, local Facebook groups and neighbourhood forums get pulled into AI answers more than most restaurants expect, because engines treat community consensus as a trust signal. You can't buy your way onto these surfaces convincingly. You earn a mention by being useful and honest there as yourself, over weeks, not with a single drive-by post.
Who should actually do this work?
That depends on how much time your team has and how technical the fixes are, which is exactly what we cover in who actually fixes AI search visibility. Some of the five fixes above are a single afternoon with your website and your booking platform logins. Others, editorial citations especially, take ongoing relationship-building that's closer to PR than to SEO.
Disclosure: Schmitdy is our own AI search service at AI Heroes, so read the offer below as the maker's case for it rather than neutral advice. We map the questions your diners actually ask, fix the entity and schema issues, pursue the editorial and community citations, and read five AI engines daily so you know whether any of this is working, rather than guessing. If you'd rather see where you currently stand before spending anything, the free AI search audit shows your restaurant's current visibility across ChatGPT, Google AI Overviews and the other major engines.
