The Answer Box Now Has a Price List

For two years the industry has debated what AI search would do to hotel demand. That debate has moved on. Google has begun placing sponsored hotel promotions inside AI generated travel recommendations through a format reported as Direct Offers, with IHG named as the first hotel partner and access currently limited to invitation only.

The significance is not the ad unit. It is what the ad unit confirms. The AI answer is no longer an experiment sitting above the search results. It is now inventory, with a commercial model attached and large chains already inside it. Hotel visibility in AI search has become a paid and earned competition running in parallel, exactly as organic search and paid search have done for twenty years.

For an independent hotel with no invitation and no group media budget, the practical question is narrow and urgent. When a traveller asks an AI assistant where to stay, what determines whether your property is in the answer at all?

Where AI Actually Gets Its Hotel Recommendations

Research gives a clear answer, and it is uncomfortable. Lighthouse presented its AI Visibility Edge study at Luminate 2026, built on 4,545 ChatGPT prompts spanning nine global destinations, five traveller personas and 101 prompt variations per combination, tracking 49,707 hotel mentions across 2,721 unique properties.

The headline finding: 82 per cent of the sources AI draws on for hotel recommendations come from just two categories. OTA and metasearch platforms such as Booking.com and Expedia, and editorial and media sites.

Your own website is not in that majority. The places you pay commission to, and the places you have never pitched, are.

Most hotels are simply absent

The same study measured how many hotels in a market ever get surfaced at all. In Tokyo, 10 per cent. In Paris, 13 per cent. In Park City, roughly a third, meaning two thirds of the market were never mentioned across thousands of prompts.

Independent properties fare worse than the market average in most cities measured, from 6.5 per cent of mentions in Indianapolis to 10.5 per cent in Tokyo, with Brussels an outlier at 28 per cent. Chains dominate because their information is consistent, structured and repeated across exactly the sources the models trust.

Why Paid Placement Changes the Calculation

Paid placement inside the answer does two things to a market. It puts a floor under the visibility of whoever can buy in, and it compresses the space left for everything earned. If a group can pay to sit beside the recommendation while your property is not even in the candidate set, the gap is no longer about brand strength. It is about machine readable presence.

There is a second effect worth planning for. Once an answer carries commercial placements, the neutral recommendations around it become more scrutinised, better labelled and more valuable. Being the property the model chooses on merit is a genuine competitive asset, and it is one no invitation is required to build.

What Earns a Mention

None of the levers here are exotic. They are the unglamorous fundamentals that most properties have never completed.

• Make your facts machine readable. Structured data on your own site covering property type, location, room types, amenities, rates and reviews is what allows a model to state facts about you with confidence rather than guessing or skipping you.
• Fix inconsistency across the web. If your room count, facilities or address differ between your site, your OTA listings and your directory entries, the model has no reliable version of you to repeat. Consistency is a ranking factor in an answer economy.
• Treat your OTA listings as content, not just distribution. Since OTA and metasearch pages are among the most cited sources, a thin, poorly written listing weakens your visibility even in channels you would rather not depend on.
• Earn editorial mentions deliberately. Local press, destination guides, niche travel publications and roundup articles are the other half of that 82 per cent. One well placed feature can influence answers for months.
• Answer real traveller questions on your own pages. Prompts are conversational and specific. Pages that address family suites, late arrivals, accessibility, parking or walking distances give a model something concrete to match.

This is the same discovery shift we set out in how AI search is rewriting hotel discovery, now with a commercial layer bolted on top.

Visibility Is a Media Discipline Again

Being cited by an AI assistant looks like an SEO problem and behaves like a media problem. It rewards consistent presence across the platforms and publications a model already trusts, sustained over time, measured properly. That is the work Bookassist Digital Media exists to do, aligning search, metasearch and paid activity so that the same accurate version of your hotel shows up wherever the decision is actually forming.

The chains are not waiting for the format to mature. They are already inside it.

Find Out Whether AI Can See You

You cannot fix an absence you have not measured. Run the free AI Visibility Audit on your hotel website. It takes less than a minute, and it shows you what an AI system can and cannot establish about your property today, which is the difference between being recommended and being invisible.

---

Photo via Unsplash