The Billboard Effect After AI: Do OTAs Still Send You Direct Guests?

In 2009, Chris Anderson at Cornell's School of Hotel Administration published "The Billboard Effect: Online Travel Agent Impact on Non-OTA Reservation Volume". He studied hotels operated by JHM Hotels that were listed on Expedia.com and found that when the hotels were listed on Expedia they saw more reservations through their own distribution channels too. For the four properties in the study, the increase was between 7.5% and 26%.

In 2011 he followed up with "Search, OTAs, and Online Booking: An Expanded Analysis of the Billboard Effect", using data on 1,720 reservations for InterContinental Hotels Group brands over the summers of 2008 to 2010. The headline finding was that for each reservation a hotel received through Expedia, the brand's own website received between three and nine additional reservations. Consumers in the data visited OTA sites a lot: the average consumer made twelve visits to an OTA's website, viewing 7.5 pages per visit.

Press coverage of that second study at the time reported that almost 75% of consumers who booked on the brand's site had visited the OTA first, and 83% had run a web search before booking. Those figures come from the coverage rather than the paper's abstract, so treat them as a summary.

The argument hotels took from this was simple. An OTA listing is a sales channel with a commission, and it is also advertising. Guests find you on Booking.com or Expedia, then go to your website to book, and some of those bookings would never have happened without the listing.

It is worth being exact, because the billboard effect gets quoted as a law of nature. The 2009 study covered four hotels. The 2011 study covered one large chain's brands, with loyalty programmes and strong brand websites, in a specific period.

None of that makes the findings wrong. It means the size of the effect depends on the hotel. A chain hotel with a loyalty programme and a good website gives an OTA browser a strong reason to book direct. An independent with a slow booking engine and no direct-booking benefit gives them much less of one.

The research also described a particular path to booking: a guest browses an OTA, notes a few hotels, searches for them by name, and compares. That path is the part that has changed most since 2011.

A guest who asks an AI assistant where to stay does not browse twelve OTA pages. They get a short answer with two or three hotels, often with links, and many go straight from the answer to a booking page. If the link goes to an OTA, the discovery and the booking happen in the same place, and the moment where a guest would have searched for your website may never come.

That cuts both ways. If an assistant names your hotel and links to your own site, you get a direct booking without the OTA having been involved at all, the billboard effect without the billboard. If it names you and links to Booking.com, the OTA collects the commission on a guest whose decision was made by the AI. And if it does not name you, the guest may never hear of you, on the OTA or anywhere else.

So the question a hotel should ask has shifted. In 2011 it was whether an OTA listing sends people to your website. Now it is also which page the AI answer sends people to, and whether your own site is a source the engines trust enough to cite.

We publish the results of real scans for a small number of hotels, and the citation pattern in them is consistent. When AI engines answered Prague hotel questions without naming Golden Well Hotel, they leaned on Booking.com and TripAdvisor most often, alongside a French-language travel blog and the hotel-editorial site thehotelguru.com. For Mosaic House, also in Prague, the sources engines pointed to were Booking.com, TripAdvisor, Expedia and the same editorial site. In both cases the hotel's own website was rarely the source the engine cited, even on questions the hotel was a good answer to.

Outside the big city the pattern changed. For Olympia Hotel in the spa town of Mariánské Lázně, engines cited Booking.com and local tourism directories, and in some answers a competing spa hotel brand's own website. For Bouda Mama, a family hotel in Pec pod Sněžkou, engines mostly cited other local hotels directly.

Those are four hotels, scanned at particular times on particular sets of questions. They are not a representative sample of the industry, and we are not going to turn them into a percentage. What they show is that when AI answers point somewhere other than the hotel, the place they point is very often an OTA, which is the reverse of the path the billboard effect described.

Across the 8 hotels in our published benchmarks, 4 were named in none of the AI answers we checked. A hotel that is not named gets no billboard at all.

You can test your own billboard effect without a research budget, using data you already have. The idea is the same as Anderson's: compare periods when something changed on the OTA with what happened on your direct channel.

1. Pull monthly branded search impressions and clicks from Google Search Console for your hotel's name, for at least the last two years.

2. Pull monthly direct bookings and direct room revenue from your PMS for the same period, using a fixed definition of direct.

3. Mark the months when your OTA visibility changed sharply: joining or leaving a programme like Genius or Preferred, a big change in OTA rates, a closed-out period, or a sharp change in review score.

4. Look at whether branded search and direct bookings moved in the months after. If they rose when OTA visibility rose, some of your direct business is billboard-driven. If they did not move, the OTA is mostly selling rooms, not advertising you.

5. Separately, check where AI answers send people. Ask a few assistants the questions your guests ask, and note whether your hotel is named and which site the answer links to.

This is not a controlled experiment, and seasonality will blur it, so compare against the same month last year. But it gives you a better answer than assuming the billboard effect applies to you because a study of other hotels found it.

Keep the OTA listings accurate and complete. They are still a discovery channel, and they are one of the third-party records AI engines read about you. A wrong check-in time or a missing amenity on Booking.com can end up repeated in an AI answer.

Make the billboard pay off. If guests find you on an OTA and then look you up, your own site has to load fast, show the price straight away and give a reason to book direct. Our direct bookings guide covers the offers and the stack.

Work on being cited directly. The things that make an AI engine cite your own site rather than an OTA are crawler access, facts stated clearly in structured data, consistent details across listings, and presence on the local and editorial sites engines already trust. Our guide to why AI recommends Booking.com instead of your hotel goes through each one.

And measure both sides on a schedule: your direct share and branded search from your own systems, and how often AI answers name your hotel and where they link. The billboard effect was a finding about one path to booking. The path is changing, so the number worth trusting is your own.