Independent hotels like to say AI is rigged in favour of the big chains. The best data I have seen this year says otherwise. AI is not biased against independents. It simply does not know that many of them exist. That is a worse problem today, and a much better one long term, because it can be fixed.
What the Barcelona study found
At Skift’s Data + AI Summit in London last week, one finding stood out in Skift’s write-up. It came from Carlo Del Mistro, founder of Stiplo and former Chief Digital Officer at Ennismore.
For the Stiplo Barcelona Study, his team took every registered hotel, hostal and guesthouse in Barcelona, 659 properties in total. They asked ChatGPT, Perplexity, Google AI and Gemini 114 real-world travel questions, in English, Spanish and Catalan, 20 times each. That is 27,360 AI answers.
143 of those 659 properties never appeared once. Not in any assistant, not in any language. That is more than one in five.
The split is what should get every independent owner’s attention. 32.4% of independent properties never surfaced, against just 4% of chain hotels. Independents were eight times more likely to be invisible. Chains make up 37.6% of the register but took 62.2% of all the appearances.
Now the twist. Once an independent did appear, the AI described it just as warmly as a chain. Del Mistro puts favourable framing at 88% for independents and 85% for chains. The penalty is not in the judgement. It is in getting into the conversation at all.
This is not bias. It is admin.
Del Mistro made the point bluntly on LinkedIn: “the disadvantage is administrative. Administrative problems have owners, and they can be resolved.”
I agree, and I would go further. If you run an independent hotel and you cannot name the person responsible for whether ChatGPT knows your hotel exists, you have found your problem.
Chains do not get recommended because AI loves brands. They get recommended because they have more of the signals AI reads: more reviews, more pages, more mentions across the web. Independents can build those signals too. Most just have not been told to.
Your GEO checklist is probably the wrong to-do list
This is where I expect some pushback. The study found that schema markup, llms.txt files, AI-crawler settings and content freshness showed no meaningful independent link to AI visibility. Stiplo is careful to say this is association, not causation. Even so, plenty of agencies are selling exactly those tasks as “AI optimisation” right now.
So what did line up with visibility? Google review volume had the strongest link of anything measured: 0.64 on a raw basis, and around 0.35 after adjusting for things like star rating, chain membership and domain authority. Stiplo found Google review volume mattered about five times more than review volume on OTAs and metasearch. Search visibility came next, then domain authority. And hotels with crawlable pages answering real guest questions were visible 87% of the time, against 72% for those without.
The unglamorous truth is that AI visibility looks a lot like good reputation management and good travel SEO. Get more Google reviews. Write pages that answer what guests actually ask. Get talked about on other people’s websites. None of that is new. All of it is now more urgent.
Being named is not the same as being booked
There is a second sting. Even when an AI assistant recommended a hotel, that hotel’s own website was among the cited sources only 7.3% of the time. Booking.com was the single most-present website, appearing in 14.3% of all answers, and Reddit was the most-cited. Among answers that named a hotel, 48.1% leaned on an OTA or aggregator source. Only 29.6% included any hotel’s own website.
In other words, the AI may do the shortlisting for you, then hand the click to someone else. Last week I argued that the tiny share of traffic AI sends hotels is the wrong number to watch. This is why. The influence is happening before the click, and right now the intermediaries are better placed to cash it in. It is the same pattern I flagged when Hilton’s AI agent problem became a gift to the OTAs, and when Google wrote the rules for AI hotel booking without hotels in the room.
Why this matters more in Europe
Skift notes that around three-quarters of European hotels are independently owned. If almost one in three independents is invisible in a city as famous as Barcelona, think about a smaller destination with fewer reviews and less press coverage.
The good news is that travellers have not fully handed over their trust yet. Skyscanner CEO Bryan Batista cited a global survey at the summit in which 51% of travellers said they distrust the reliability and accuracy of AI travel recommendations. That gives independents a window. It will not stay open for ever.
What I would do this month
- Give one named person ownership of AI visibility. Not “marketing”. A person.
- Ask happy guests for a Google review, not just an OTA review.
- Build plain, practical pages that answer real questions: parking, distance to the stadium or cruise port, family rooms, late check-in.
- Own a niche. Stiplo found a stadium-side hotel got a 98-times visibility boost on football weekend questions. You do not need to win “best hotel in Barcelona”. You need to win the question your ideal guest asks.
- Check whether you show up at all before you worry about where you rank. Ask the assistants the questions your guests ask, and keep a record.
If you want a hand with this, my travel AI and GEO services page explains how I work with hotels.
The debate
Independent hotels have one big advantage here: they are specific, and AI loves specific answers. The independents that lose will not be losing to the chains. They will be losing to silence.
So is AI invisibility a technology problem, or a reputation problem that hotels have been ignoring for years?
Follow the newsletter: Dan Taylor on LinkedIn