Everyone’s Panicking About AI and Hotel Bookings. The Data Says Something Else.

Every conference panel this year has the same slide: a funnel in flames, an AI chatbot standing over the wreckage. The pitch is always the same too – the old search-discover-research-book model is dead, and if you don’t buy whatever’s being sold, you’ll be invisible by Christmas.

Here’s the problem. When you go looking for the data behind that story, it mostly isn’t there. And the data that does exist tells a more interesting, more useful story than the one being sold to you.

Google’s original four-stage model came from a specific piece of research: The 2014 Traveler’s Road to Decision, built with Ipsos MediaCT from a 5,000-person US survey. It described four moments, not a rigid funnel: Dreaming, Planning, Booking, Experiencing. That’s it. The “38 sites visited” figure so often bolted onto it is actually Expedia Media Solutions data from 2013, not Google’s. And the famous “7 hours, 11 touchpoints, 4 locations” rule everyone quotes at conferences? No verifiable Google source exists for it. It’s industry folklore, repeated so often it’s been mistaken for research.

So before we even get to AI, the baseline everyone’s measuring against was already softer than it’s treated.

Let’s deal with the two questions that matter: has the path to booking shortened, and has AI actually moved the money.

The path got longer, not shorter. Expedia Group’s Path to Purchase research – passively metered behaviour from over 70,000 travellers across seven markets, not a survey – found travellers viewed 141 pages of travel content globally, rising to 277 in the US, over roughly five hours spread across the 45 days before booking. If AI were collapsing the funnel into a single decisive interaction, this number should be shrinking. It isn’t.

Direct bookings haven’t been eroded by AI. SiteMinder’s 2025 data, drawn from over 130 million bookings across 20 markets, found direct revenue share stayed within 1.5 percentage points of the prior year in 95% of the markets they analysed, despite a full year of predictions that AI would swing it hard one way or the other. Hotel websites also generated a meaningfully higher average booking value than OTAs, £516 against £312, which tells you something about where the quality of demand still sits.

AI traffic is real, but it’s small. Lighthouse, working across an 80,000-hotel network, puts AI referral traffic at just under 1% of total hotel website visits. Adobe Analytics recorded eye-catching year-on-year growth in AI-referred travel traffic, but growth off a tiny base is not the same as scale. Adobe’s own data also shows AI-referred visitors converting well below other channels, although that gap has narrowed by roughly 70% since late 2024. This is a channel maturing, not a channel that’s already arrived.

Search’s share of research is genuinely shifting. Phocuswright found search engines’ share of the traveller’s research mix fell from 51% to 36% across 2025, while generative AI platforms rose from 6% to 15%. That’s a real, measurable shift in where people start looking. It is not evidence that hotels have lost the booking.

If you want one number to stop the scaremongering in a room, use this: SiteMinder now shows OTAs, not search engines, are the most common starting point for hotel research, at 26% against 21%. Word of mouth has doubled to 14%. The funnel didn’t collapse. It reordered itself.

You’ll have seen the stat that AI referrals convert at 11.4% against 5.3% for organic search, or that generative AI drove 357% growth and 1.1 billion visits to hospitality sites in 2025. Worth knowing: those figures are Similarweb’s all-web, cross-industry numbers, dominated by sites like Google, Reddit and Amazon, repackaged by a vendor as hospitality-specific and wrongly credited to Phocuswright. If a number sounds designed to make you panic-buy a subscription, check where it actually came from before you repeat it.

Neither extreme is right. Google’s own “messy middle” framework, published in 2020, describes buyers looping between expansive exploration and narrower evaluation, across multiple channels, until they reach confidence, rather than marching through four neat boxes. That was already true for hotels before generative AI existed. What’s changed is the number and type of loops, not the existence of a journey. AI has added a new exploration surface. It hasn’t replaced the evaluation and conversion layer where hotels still win or lose the booking.

Based on the evidence, not the hype, four things carry real, measurable weight right now:

1. Your website and booking engine. Still the highest-value conversion point you own, and the destination for AI-referred traffic once it does arrive. If your direct conversion rate sits near the industry average of around 2.2%, fix that before anything else.
2. OTAs and metasearch. Still 55-60%+ of independent bookings by most measures. Ignoring them because AI is fashionable would be a mistake, not a strategy.
3. Google Business Profile, listings accuracy and reviews. This is what AI assistants and search overviews actually draw from. Getting this right is the single highest-leverage “AI-readiness” move available to any hotel, chain or independent.
4. AI visibility, as a watching brief. Worth monitoring, not worth panicking over. Independent research from HotelWorld AI found only around one in six of the world’s hotel properties currently appear in AI search results at all, and chain-affiliated properties dominate what visibility exists. That’s a gap worth closing over time. It isn’t a fire to put out this quarter.

The industry doesn’t have a funnel-collapse problem. It has a discipline problem, dressed up as an AI problem. The fundamentals that always separated hotels who owned their distribution from hotels who didn’t – accurate data, a booking engine that doesn’t fight the guest, a review profile that reflects reality, a direct channel worth choosing – are exactly the fundamentals that also determine whether AI recommends you.

So the honest question isn’t “is AI killing the funnel.” It’s whether you’ve actually fixed the things that were already broken before you go chasing the next platform everyone says you can’t afford to ignore.

Are you ready to look at your own data before you believe someone else’s headline?

Sources:

• Google & Ipsos MediaCT – The 2014 Traveler’s Road to Decision
• Expedia Media Solutions/Compete (2013) – “38 sites” study, presented at Phocuswright Conference
• Expedia Group – Path to Purchase research (with Luth Research)
• Google – “Decoding Decisions” / The Messy Middle (with The Behavioural Architects, 2020)
• SiteMinder – Hotel Booking Trends 2025 report
• SiteMinder – Changing Traveller Report 2026
• Cloudbeds – State of Independent Hotels 2026
• Lighthouse – AI referral traffic data (80,000-hotel network)
• Adobe Analytics – Generative AI travel traffic reports
• Phocuswright – “Search Slips, AI Surges: Travel’s New Front Door?”
• Skift Research × McKinsey – “Remapping Travel with Agentic AI”
• HotelWorld AI – “World’s Best at AI 2025 Index”
• Pew Research Center – AI Overviews click-through study (March 2025)