When someone asks an AI chatbot such as ChatGPT or Perplexity whether a bar or restaurant is worth visiting, its answer is mostly assembled from websites that a venue operator doesn’t control.
A new analysis of 2.9 million AI citations finds that just three types of website supply over a third of everything AI tells customers about a UK hospitality venue, while a venue’s website accounts for relatively little.
Searchable, the AI visibility tool, logged 2,894,938 AI citations in response to customer questions about UK hospitality venues, such as what services they offer and where they are located. Three website types account for 34% of all citations, and at least one appears in 82% of answers.
Booking and venue-hire platforms lead at 14% of citations, ahead of review sites at 11% and editorial guides at 9%.
An AI chatbot’s citation is a source that a model references when answering. This research measured citations across 279,967 distinct AI responses to customer queries. The venue’s own website accounts for 8% of citations. More starkly, 62% of AI answers never cite it at all.
Social media contributes to 5% of citations, and reference sites such as Google Maps appear in 1.4%. The remaining sources consist of over 83,000 websites, with none recording repetitive influence on the same level as the category leaders.
Review website Tripadvisor accounts for 8% of all citations alone, and was used as a source for every venue that the AI chatbots queried. Overall, the ten most-cited individual websites account for 30% of every citation, and at least one of them appears in 81% of all AI answers.
“A venue’s website still acts as a window into the business,” said Chris Donnelly, co-founder and CEO of Searchable. “But customers are increasingly arriving there after already being told what’s inside the venue by an AI platform .”
According to OpenAI, nearly half of all ChatGPT prompts now involve people asking for information and recommendations. This means AI reputation has become a front-of-house issue for venue operators to monitor. The pool of options being returned to AI chatbots is relatively narrow too. AI recommendations often return a little as three venues in a reply, which is a smaller shortlist than traditional search ever offered.
What venue operators should do
Donnelly suggests a few practical steps to ensure venues are appearing accurately in the right places for AI search platforms to learn about them.
“First, work out which of the three types is describing you, because they aren’t equally fixable. A booking listing is yours to correct today, while a guide entry is someone else’s editorial call.
“Second, treat booking platform listings as reputational copy. They’re usually the single biggest source AI reads about a venue. So a listing set up years ago and never revisited is being read by the models every day and potentially surfacing inaccurate information
“Third, get the basic facts identical everywhere, including opening hours, address, service times. Where listings disagree, the model picks one or hallucinates from a combination of conflicting facts.”
