Every study on restaurants and AI search until now has one thing in common: it measured US restaurants. There has been no UK read on which British restaurants AI engines actually name, or which sources they lean on to decide. So we built one. Over two weeks in July 2026 we tracked 28 London restaurant and wine-bar groups across around 85 real diner questions, run daily across the major AI engines, and recorded who gets named and which sources the engines cite. The short version: even the most AI-visible group in London appears in only about one in eight relevant answers, and the sources that decide the shortlist are a small, nameable set. This is a first read from a young dataset, not a census, and we will refresh it each quarter.
TL;DR
- Across roughly 2,100 AI answers to real London dining questions, the most-named group appeared in just 12.5% of them. No tracked group cleared one in seven.
- The median tracked group appeared in about 4% of relevant answers, and several well-known names in under 1%. Being famous in the room does not mean being named by the model.
- Hawksmoor led the index, followed by Blacklock, Dishoom, Noble Rot and Humble Grape.
- The sources that decide the shortlist are concentrated: Time Out was retrieved in 42% of answers, ahead of SquareMeal, thatsup, Reddit and The Infatuation. Editorial, community and listings sites, not restaurants' own websites, build the answer.
What we measured, and how
We tracked 28 London restaurant and wine-bar groups, from steak and Indian fine dining to natural wine bars, across around 85 real diner questions of the kind people actually type, best restaurants in a neighbourhood, a private room for a number of guests, a great wine bar for a date. Those questions ran daily across the major AI engines for two weeks in July 2026, producing roughly 2,100 AI answers per brand to measure against. For each brand we recorded its visibility, the share of answers that name it, and for each answer we recorded which source domains the engine retrieved. No self-reported data, no surveys: just what the engines actually said, day after day. It is a young, two-week dataset and a first read, not a definitive UK census, which is exactly why we are publishing it openly and refreshing it quarterly.
Who does AI actually name? The London leaderboard
Visibility here means the share of relevant AI answers that mention a group. Read the top number twice: the most-named group in London appears in only 12.5% of the answers to questions it should be a candidate for.
| Rank | Group | AI visibility |
|---|---|---|
| 1 | Hawksmoor | 12.5% |
| 2 | Blacklock | 9.1% |
| 3 | Dishoom | 8.0% |
| 4 | Noble Rot | 7.7% |
| 5 | Humble Grape | 7.0% |
| 6 | Kiln | 6.8% |
| 7 | St. JOHN | 6.5% |
| 8 | Gymkhana | 5.9% |
| 9 | Big Mamma | 5.8% |
| 10 | Fallow | 5.6% |
The leaderboard rewards groups with deep, consistent footprints across the editorial and community sources the engines read. But the striking finding is not the order, it is the ceiling. Hawksmoor tops the index with the largest share of voice of any tracked group, and still misses roughly seven of every eight relevant answers.
The invisibility finding
Below the top ten, the numbers fall away fast. The median tracked group appeared in about 4% of relevant answers, roughly two-thirds of the 28 groups appeared in fewer than one in twenty, and several well-known names, venues that are busy and well-reviewed in real life, showed up in under 1% of answers. This is the UK version of the pattern US audits have reported: being genuinely good, and even genuinely popular, does not translate into being named by an AI engine, because the model is reasoning from sources, not from footfall. The mechanics behind it are in why your restaurant is invisible on ChatGPT, and the wider sourced numbers in the restaurant AI visibility statistics.
Which sources does AI cite for London dining?
This is the part no one has published for the UK, and it is the most useful. When an engine answers a London dining question, it retrieves a small, consistent set of sources. Here is how often each was pulled into an answer.
| Source | Type | Share of answers retrieved |
|---|---|---|
| Time Out | Editorial | 42% |
| SquareMeal | Editorial | 25% |
| thatsup | Community listings | 24% |
| Community | 22% | |
| The Infatuation | Editorial | 20% |
| DesignMyNight | Listings | 18% |
| Michelin | Guide | 17% |
| OpenTable | Booking | 15% |
| OliveMagazine | Editorial | 14% |
| Condé Nast Traveller | Editorial | 13% |
| Visit London | Institutional | 13% |
| The Nudge | Editorial | 11% |
| TripAdvisor | Community | 9% |
Three things stand out. First, editorial dominates: Time Out alone was retrieved in more than four in ten answers, and a cluster of UK titles follows it. Second, community is not a footnote, it is core: Reddit and community listings sites together rival the big editorial names, which matches what we found in Reddit and local forums behind restaurant AI answers. Third, your own website is almost never the source; the answer is assembled from third parties who write about you. If you want to be named, you have to be present where these engines actually look.
What the index means if you run a London restaurant
Two practical reads. If you are outside the top of the leaderboard, the gap is not a quality problem, it is a presence problem: you are missing from the editorial, community and listings sources the engines retrieve. And if you run a group, the ceiling matters, because even the leader's 12.5% shows how much headroom is unclaimed, and multi-venue groups have the added task of not competing with themselves, which we cover in why your restaurant group's venues cannibalise each other. The single highest-leverage move the data points to is earning a place in the sources at the top of that second table.
Disclosure: Schmitdy is our own AI search service at AI Heroes, and this index comes from the same daily multi-engine tracking we run for the restaurants we work with. We publish it openly because a UK benchmark did not exist, and we will refresh it quarterly. If you want your own group's number, where you sit on this leaderboard and which of these sources name you, the free AI search audit shows it.
