Ninety-one percent of hotel groups already use AI, but only 28 percent have a company-wide AI strategy and just 13 percent can report measurable ROI. Those figures come from the h2c AI Opportunity Study 2026, published October 2, which surveyed 113 hotel groups representing more than 8,200 properties.

The gap widens further down the funnel. Seventy percent of the surveyed groups expect AI booking channels to increase direct bookings. Only 22 percent systematically monitor how their brand appears in AI search. In effect, most of the industry is betting on AI to drive direct traffic while never checking whether AI can accurately see their hotels at all.

A new diagnostic framework from consultancy AIDURIX argues the industry has skipped a step. The author, a veteran of more than two decades in and around hotels, says owners and GMs keep asking the wrong first question — "Which AI tool should we buy?" — when the better one is: what is AI already doing to the guest journey, and where is the hotel losing the guest?

Where does the guest stop being yours?

The framework breaks every booking into eight links:

• Recommendation: who suggests the hotel first
• Live availability: who shows whether a room exists
• Displayed price: who decides the number the guest sees
• Transaction: where the "book" button lives
• Payment: who acts as merchant of record
• Confirmation: whose name is on the email
• Guest communication: who the guest talks to pre-arrival
• Post-stay relationship: who invites them back

Ownership can quietly transfer at any link, and branding will not reveal which — a page carrying the hotel's photos, name and story can still belong to someone else. The commercial stakes differ sharply by link. Losing the guest at link four costs a commission. Losing them at link eight costs the relationship and every future stay attached to it.

Why intermediaries are changing shape, not disappearing

The distribution backdrop is shifting. HOTREC's European Hotel Distribution Study 2026 found 51 percent of hotels say OTAs frequently or occasionally undercut their rates — up 8 points since 2023. Meta's new Muse agent books flights through a direct airline connection but, for hotels, browses consumer sites such as Expedia and Hotels.com the way a person would, according to Skift.

At the Destination AI summit on September 29, Hilton's CIO said large language models put OTAs "under real threat." In the same breath, he acknowledged that AI personal agents will likely become paid channels themselves. The intermediary is not disappearing; it may simply change its name.

What does the diagnostic cover?

The method looks in two directions. Outside-in tests what AI assistants can actually see, show and complete for a hotel. Inside-out assesses whether the hotel can act on it — culture, collaboration, signals, governance and systems. Both matter: a hotel visible to AI but slow to answer guest messages has made its weakness easier to find, while a strong operation AI cannot see remains, in the author's phrase, a beautiful secret.

The author lists seven recurring problems the diagnostic addresses, from AI misstating amenities and pet policies to bookings drifting to intermediaries without a traceable moment of transfer, siloed teams, symptom-chasing tool purchases and unverified ROI promises. Anything unproven gets labeled a hypothesis to test, not a number to bank.

The framework also rejects single readiness scores — "a hotel cannot be '64 per cent ready'" — and stresses that findings must be retested and dated monthly, since AI assistants change month to month.

A 15-minute starting point

Operators can test the concept with two or three AI assistants, asking a real guest query, recording the date, exact wording and results, then repeating in a month. The output is evidence, not a score.

The winners, the author argues, will not be the hotels with the most AI, but the ones that know exactly where their guest journey stands — diagnosing first, deciding second, and letting staff handle what no algorithm can.