We tracked how ChatGPT and Google AI Overviews recommended a design-led London hotel. The audit showed it owned its postcode but not its category, that each AI was reading a different set of sources and that fewer than 2% of the evidence came from the hotel’s own website. A list of media titles would not have told us any of that. The audit told us where the hotel was really competing and what to do about it.
Most AI visibility reports end with a list of sources. Time Out, Tripadvisor, Condé Nast Traveller. It looks useful, but every agency already knows those titles matter.
What a hotel needs is the context around that list: which questions it is winning and losing, who it is really competing against, which AI reads what and how stable any of it is. That is market intelligence, and it is what this audit was designed to produce.
The audit at a glance
The client was a design-led lifestyle hotel in central London, known for its restaurant, bars and cultural programming. Its positioning is grown-up, stylish and fun, deliberately not a traditional luxury hotel.
- Built from the client’s own brief: its target guests, commercial priorities, competitor set, priority media and the language it wanted associated with the hotel
- 100 questions written the way real guests ask, from “a London hotel that feels cool but still comfortable” to “where should I stay for a weekend of galleries, restaurants and good bars”
- Each question asked repeatedly over five weeks on ChatGPT and Google AI Overviews: around 1,000 AI answers
- Up to 500 websites and 2,000 pages cited in each round of questions
- Up to 300 different businesses recommended in each round: hotels, bars, galleries, booking sites and review platforms
- None of the 100 questions named the hotel. Every answer was a genuine discovery answer.
Why unbranded questions matter
Most AI visibility reports lean on branded questions, where a guest already knows the hotel and asks about it by name. Those flatter almost every brand, because AI is very good at describing a hotel it has been told about. The harder and more commercial test is whether AI thinks of you when a guest describes what they want and names no one. That is the moment new bookings are won or lost, and it is the only moment this audit measured.
Part one: what the audit revealed
1. The hotel owns its postcode, not its category
Asked about its neighbourhood, AI reached for the hotel often. Asked about the things its brand is built on, it rarely did.
How often AI recommended the hotel
A question about the neighbourhood narrows the choice to a handful of nearby hotels, so this one was an easy answer. A question about style or atmosphere opens it up to every hotel in London. There, the hotel had to compete on reputation alone, and on fourteen questions it was never recommended, including “somewhere grown-up but still fun”, “a more local-feeling London trip” and “dinner, cocktails and staying the night”. The hotel’s own keyword list split along the same line. Its location terms matched where AI recommended it. Its category terms, such as “design hotel London” and “lifestyle hotel London”, matched where AI did not.
What it means
The brand’s positioning has not travelled beyond its address. On location questions the hotel competes with its immediate neighbours. On lifestyle questions it competes with the design-led hotels of Soho, Shoreditch and Marylebone, and it mostly loses. Of the ten competitors the hotel named in its own brief, only one regularly made AI’s top ten. The brand plans for one competitive set. AI judges it against two.
2. ChatGPT and Google are two different media markets
Google AI Overviews read the crowd. Tripadvisor was consistently its most influential source, and Reddit, Facebook, YouTube and TikTok all ranked among its most-cited sources. ChatGPT leaned on the editors. Time Out was its most influential source on average, alongside Visit London, DesignMyNight and Condé Nast Traveller. Tripadvisor mattered to both.
Most influential sources, average share of answers
Google AI Overviews
ChatGPT
Across 4,800 cited pages, only 1 in 10 was used by both. The results showed it. Asked for “a hotel that feels stylish but not stuffy”, the hotel’s core positioning, Google named it five times out of five. ChatGPT named it none.
4,800 pages cited: who used them
“A hotel that feels stylish but not stuffy”
What it means
One media plan cannot win both. Google is won through reviews, community and creators. ChatGPT is won through editorial roundups. That split decides where the PR budget should go.
3. The hotel’s story is told almost entirely by others
Fewer than 2% of the sources AI cited came from the hotel’s own website. Everything AI said about it, it learned from someone else.
Where AI’s evidence came from
What it means
This is not an SEO problem. It is a reputation and coverage problem, which makes it PR’s territory. It is the clearest case yet for the agency, not the web team, to lead on AI visibility.
4. AI knows the hotel, but not what is inside it
Competitors’ bars and restaurants were recommended as destinations in their own right. This hotel’s restaurant and bars each surfaced by name in under 1% of answers. Its cultural programming, the one thing no competitor can claim, did not appear in a single cited source.
What it means
AI is selling the hotel as a well-located room, not as a place to go. Its venues are the bridge from the location market to the lifestyle market, and that bridge is missing.
5. Answers vary. The odds do not.
Ask AI the same question twice and you may get a different answer: 82 of the 100 questions did. But the patterns behind those answers held throughout. The gap between location and lifestyle questions never closed. Tripadvisor led Google’s sources throughout and Time Out stayed among ChatGPT’s top two. The hotel’s own website never rose above 2%.
100 questions: did the answer change?
What held throughout
- The gap between location and lifestyle questions
- Tripadvisor as Google’s leading source
- Time Out among ChatGPT’s top two sources
- The hotel’s own website below 2% of sources
What it means
AI does not pick a winner once. It re-weighs the evidence every time someone asks. A brand cannot control a single answer, but it can change the odds on all of them by changing the evidence: the reviews, the roundups and the coverage AI keeps reading.
6. A handful of lists do much of the work
The Times’ “best boutique hotels in London” was one of Google’s most-cited pages. Its “coolest hotels in London” list fed both ChatGPT and Google. The hotel was on neither.
Lists AI kept reading
The Times
Best boutique hotels in London
Cited by Google · Hotel not listed
The Times
The coolest hotels in London
Cited by ChatGPT and Google · Hotel not listed
What it means
Some titles carry outsized weight because AI returns to them again and again. One placement on the right list is worth more than many one-off mentions.
7. The media wishlist is not AI’s reading list
Before the audit, the hotel listed the media that mattered most to it: Condé Nast Traveller, Travel + Leisure, Wallpaper, Monocle, FT How To Spend It, Vogue and Vanity Fair, plus guides such as Mr & Mrs Smith and Michelin. Only Condé Nast Traveller carried real weight with both AIs. Monocle was not cited once. Meanwhile Time Out, Reddit, Visit London, The Independent and The Times, which shaped far more of the answers, were not on the list at all.
The hotel’s priority media: average share of answers citing each title
What it means
Prestige and AI influence are different currencies. The titles a brand is proudest to appear in still reach guests directly, but many of them barely register with AI. Not every source is neutral either: two of the guides AI cited on both platforms are booking platforms owned by major hotel groups, one of them the hotel’s own, yet AI reads them as independent curation. An AI visibility plan has to be built on the sources AI actually reads, and the audit is the only way to know which those are.
Part two: from intelligence to strategy
The findings describe a simple journey and the point where it breaks.
Asked about its neighbourhood
Recommended
58% of answers
Asked about its style
Mostly missed
16% of answers
Asked about its venues and culture
Almost absent
Under 1% of answers
The strategy moves the hotel along that line: from a convenient address to a lifestyle choice to a destination.
It runs as seven workstreams, in priority order.
01
Take the positioning beyond the postcode
Rebuild the PR narrative around the lifestyle questions guests actually ask. Agree one language brief, built on style, design, music, food and grown-up fun, and use it in every release, landing page and caption so third parties begin to repeat it.
02
Make the venues destinations
Win standalone coverage for the restaurant and bars as places to go: reviews and features in Time Out, Hot Dinners, DesignMyNight and Square Meal. Pitch the cultural programming as the story no competitor can tell.
03
Run two media plans, one for each platform
For Google: steady, recent Tripadvisor reviews, genuine presence in Reddit travel threads and creator content on TikTok and Instagram. For ChatGPT: editorial roundups in Time Out, Condé Nast Traveller, The Independent and Rough Guides.
04
Get on the lists AI keeps reading
A targeted push to the travel desks of The Times and The Telegraph for their recurring London roundups, alongside the Condé Nast Traveller and Time Out lists AI already cites.
05
Defend the home turf
Keep transport guides, local area guides and niche travel blogs current and accurate, and make sure every “where to stay near” list includes the hotel.
06
Give AI something to read on the hotel’s own site
Publish clear, fact-rich, question-led pages for each venue and for the cultural programme, written in the same language brief.
07
Read the story every month
A monthly read-out, not a scoreboard: what AI is saying about the hotel, who it is recommending instead and which sources are shaping the answer. The question is whether the story is moving in the right direction, not whether a number went up a point.
Why this is market intelligence, not a media list
A list of influential titles would have told this hotel to pitch Time Out and look after its Tripadvisor reviews. Sound advice, but generic.
The audit told it six things no list could:
- Who it is really competing against, question by question, and that the competition changes with the question
- Which questions it is winning and which it is losing, and why
- Which AI reads which sources, so the plan can be split rather than duplicated
- Which of its prized titles AI actually reads, and which it barely notices
- Where its story is actually being told, and how little of it the hotel itself controls
- How settled the market is, and where positions are still there to be taken
That is the difference between knowing which media matter and knowing what to do next.
Run before a launch, a repositioning or a new PR year, an audit stops being a scorecard and becomes a strategy.
If your agency’s clients are asking about AI search, let’s talk about what an audit could show them.
Book a 30-minute call