The Shortlist Effect: The AI sent her. Your analytics say Google did.

Half of AI users check Google before they book. When they reach your site, your report credits the search, not the AI that sent them.

by Markus Busch

Chris Anderson could only prove the billboard effect because Expedia agreed to switch four hotels off and on again, week by week, until the difference showed up in the reservations. No hotel can ask ChatGPT for that favor.

That leaves a GM with a problem. A traveler gets three names from an AI and checks the hotel's own website before she books. The hotel can put breakfast and an upgrade on that page to keep her. Whether any of it works depends on seeing her arrive, and the standard reports only show the last click before she did.

Much of the Shortlist Effect is already in your numbers. It's filed under Google.

ChatGPT tags its own links

The part you can see is small and clearly labeled. According to OpenAI's own publisher FAQ, "ChatGPT automatically includes the UTM parameter utm_source=chatgpt.com in referral URLs." When she taps a link in the answer, Google Analytics shows the visit as coming from chatgpt.com. The same FAQ ties those links to sites that let OpenAI's search crawler, OAI-SearchBot, read them. If your web agency blocked it, ask why.

That traffic is real and growing, and it is still a sliver. In July, Lighthouse, which sells hotel software, put AI tools at just under 1% of visits to hotel websites.

Half check Google first, and arrive as a search for your name

Many travelers never tap that link. In Yext's survey of 3,848 adults this spring, 53% searched Google or Bing after an AI recommended a business, and only about 5% acted without checking anything. Yext sells listings and AI visibility, and the survey covers local businesses in general.

Follow her through it. ChatGPT names your hotel. She opens Google and types the name. She clicks your website in the results.

Your analytics record a Google search for your hotel's name. The report gives the credit to Google, and nothing in it mentions that an AI put the name in her head ten minutes earlier.

Google counts its own AI as ordinary search

It gets harder inside Google. When a traveler asks Google's AI Mode for a hotel and clicks through to you, Search Console doesn't report it separately. Google's changelog says plainly that "data from AI Mode is now counting towards the totals in the Search Console Performance report", mixed in with every other web search.

So the ChatGPT user who checks you on Google and the Google user who asked Google's own AI land in the same place on your report. Both look like search.

Your booking engine may lose her at the last step

There is one more place the trail goes cold, and it's inside your own site. Many independent hotels run their booking engine on the provider's domain, not their own. Without what Google calls cross-domain measurement, a visitor who moves between two domains is counted as two users and two sessions. The visit from chatgpt.com ends on your homepage. The booking starts fresh on the engine's domain, and its source becomes your own website.

So a hotel can have AI guests booking every week and a report that credits every one of those bookings to itself. Ask whoever runs your analytics one question: does a booking keep the source of the visit that led to it? If the answer is no, fix that first. The other numbers depend on it.

Three numbers the hotel already has

None of this needs new software. It needs three numbers someone at the hotel already has access to.

Bookings from chatgpt.com. In Google Analytics, filter sessions by source and find chatgpt.com. Look past the visits to the bookings. Compare how often those visits book with the rest of your traffic. Across US travel sites, Adobe found AI-referred visitors converted only 1% less than other traffic in July. If yours book far less often than your other visitors, the page is losing them.

Searches for your hotel's name. In Search Console, filter the queries that contain your hotel's name and compare the last three months with the same months last year. Search Console keeps sixteen months of data, so last year's figures are already there. Searches for your name grow when more people hear it somewhere. If they are rising while searches for your destination are flat, something upstream is telling more travelers about you.

One new answer at the booking form. If your booking engine asks how the guest heard about you, add "an AI assistant, like ChatGPT" to the list. Ask the same question at check-in and note it in the profile. It's the only one of the three that reaches the guest who typed your name into Google, because only she knows why she typed it.

Taken together, the three give you something Anderson had to borrow from Expedia: a before and after. Run them for a month before you put the direct package from the last piece on your homepage, then for a month after.

Pasted links arrive with no source

Some of the traffic can't be traced at all. A link copied out of the ChatGPT app and pasted into a browser usually arrives with no source, and analytics files it as direct. Searches for a hotel's name also rise for reasons that have nothing to do with AI: a press mention, a new review, a busy season. What guests tick on a booking form is what they remember. None of the three numbers proves the Shortlist Effect on its own. Read together and over time, they show whether it is growing at your hotel.

Read also: The Shortlist Effect: She found you through AI. She won't book through it. Yet. · The Shortlist Effect: The commission is already in your budget. Spend it on her first. · ChatGPT now sends 92% of AI referral traffic — but a quarter of it lands on the wrong page

Enjoying this analysis?* Hospitality.today delivers daily insights on hotel distribution, AI trends, and travel commerce — straight to your inbox. Subscribe for free at Hospitality.today →