The open-access white paper argues that hospitality’s AI-era problem is not simply being found. It is keeping the same consequential stay intact across the systems that describe, sell, compare, summarize and deliver it.
A hotel can rank.
It can appear on Google.
It can have structured data, an OTA listing, a direct booking engine, destination coverage and an AI answer.
And the traveler can still receive the wrong conclusion.
Tamaga has published One Hotel, Seven Versions, an open-access public-source forensic study built around that problem.
The paper compares 16 traveler-critical facts across seven public and transactional representations of one accommodation property and asks a harder question than visibility:
Do the versions still describe the same stay where the guest needs certainty?
The publication is available as a readable online edition and a fixed downloadable PDF, with no email gate.
Hospitality has a version-control problem
The modern hotel no longer exists in one digital place.
It exists simultaneously as:
- an owned website;
- a direct booking engine;
- a structured-data layer;
- Google and local-search surfaces;
- OTAs and review platforms;
- destination and editorial sources;
- AI synthesis.
Those versions should not be identical. They have different jobs.
The failure begins when they change the consequence of a fact.
Connecting rooms on request
becomes:
Connecting rooms available.
An adapted room becomes an “accessible hotel.”
Sunday lunch becomes an assumed Sunday dinner.
A policy attached to one rate becomes a generic hotel promise.
The noun survives.
The condition disappears.
One Hotel, Seven Versions names the distance between operational reality and what these distributed versions allow a traveler or machine to conclude the Property Truth Gap.
The paper’s broader argument is that hospitality has built a distributed representation system without a reliable way to reconcile its consequential facts.
The dangerous answer is the plausible one
The paper opens with a deliberately difficult traveler request: a family arriving by train after 20:00 in Chamonix needs connecting rooms, step-free access, dinner that evening and flexible cancellation.
There is enough public material to produce a polished answer.
There is not enough evidence to establish every part of it.
The reviewed sources support pieces of the stay at different strengths and under different conditions. Communicating rooms are described as available on request. Public material supports selected access facts but not a complete step-free journey for a specific traveler. Sunday service evidence does not by itself establish the requested dinner. Late-arrival wording varies by representation.
The answer does not have to hallucinate to fail.
It can simply remove the boundaries between:
- capability and guarantee;
- description and live availability;
- general policy and selected offer;
- public evidence and human confirmation.
AI did not create the contradiction. It exposed the representation system behind it.
AI visibility is the wrong place to stop
The release comes at an unusual moment.
Google now gives site owners dedicated reporting for visibility in generative AI features. At the same time, its current guidance says there is no special AI file, special Schema.org markup or separate optimization requirement for appearing in AI Overviews or AI Mode; the same foundations—useful content, crawlability, visible information and accurate structured data—still matter.
So One Hotel, Seven Versions asks what happens after visibility.
Was the correct property resolved?
Was the fact current?
Was the condition preserved?
Did the live offer agree?
Did the guest reach the right action?
Could the booking be corrected or recovered?
A citation is not a recommendation.
A recommendation is not a booking.
And a completed booking is not proof that every consequential promise survived.
One public case, not a hotel score
The paper uses Hôtel Mont-Blanc Chamonix as an illustrative public-source forensic case because it is not digitally absent. It has the kind of distributed footprint hotels are routinely encouraged to build: a substantial owned site, room pages, a direct booking application, Google representation, major OTA listings, destination-office material, editorial coverage and an existing AI report.
Tamaga reviewed 16 traveler-critical facts across the seven representations.
The audit does not score the hotel.
It does not claim access to its PMS, CRS, private operational rules or staff procedures.
It does not claim that one case establishes prevalence across the sector.
And one representation—the structured-data example—is explicitly a third-party proposal from a supplied publication, not markup verified as deployed by the hotel.
Those boundaries are part of the research, not disclaimers added after it.
The surprising finding: agreement matters more than completeness
A natural response to AI-mediated discovery is to publish more:
more pages, more schema, more attributes, more feeds, more content.
The paper argues for a different first move:
Reconcile what already exists.
A structured-data graph can be explicit and still contain the wrong checkout time or bed configuration.
A destination office can accurately corroborate a spa while having no authority over room allocation.
An OTA can contain valuable commercial detail while flattening room nuance.
The official website can be the richest source and still distribute its truth across separate pages and seasons.
The booking engine can hold the freshest commercial state while remaining opaque to a static audit.
AI can successfully retrieve several of these versions and still produce an underqualified answer.
The objective is therefore not one universal database.
Different facts belong in different systems because they move at different speeds.
The objective is:
one accountable authority for each consequential fact, known propagation between systems, and visible boundaries when certainty ends.
The paper’s conclusion puts it more simply:
The agreement is the asset.
Three concepts for the next hotel web
The paper introduces three Tamaga concepts:
Property Truth Gap
The distance between what the property can deliver and what its representations allow a traveler or machine to conclude.
Omnichannel Decision Mesh
The pages, profiles, partners, sources and booking paths that support a traveler decision without losing meaning between channels.
Source-to-Service Chain
The path from description through resolution, corroboration, comparison, live offer, confirmation and action to recovery when something changes.
Together they move the discussion from:
How do I rank in AI?
to:
Can the knowledge survive from source to service?
The paper is also a refusal
One Hotel, Seven Versions does not claim that:
- SEO has been replaced by GEO;
- JSON-LD automatically improves AI recommendation;
- structured data is ground truth;
- an AI citation proves preference;
- a direct link proves a direct booking;
- an API guarantees safe agent action;
- one hotel establishes a sector-wide failure rate.
The harder proposition is less theatrical.
Structured data should make defensible facts explicit.
Search should make useful sources retrievable.
Live systems should own live commercial truth.
External sources should corroborate where they are actually authoritative.
Humans should retain judgment where the promise remains conditional.
The systems do not need the same words.
They need the same consequential meaning.
Open access is part of the method
Tamaga is publishing the research without a lead-generation gate.
The online edition is designed to expose the argument together with its methodology, evidence labels, figures, references, limitations, version history and corrections.
The PDF remains a fixed downloadable edition.
The editorial rule is simple:
Show what is sourced. Show what is limited. Make corrections possible.
The paper’s practical instruction is equally small.
Choose one consequential fact—connecting rooms, late arrival, step-free access, cancellation, one room configuration—and trace it through all seven versions.
Who owns it?
What conditions apply?
Which version changes it?
Which system can keep it true?
What still requires a person?
That is the smallest form of the research method.
It is also the starting point for Tamaga’s Seven-Version Snapshot.
Become the clearest source before trying to become the chosen answer.
And the final proposition is simpler still:
The hotel is not a page. It is an agreement.
Read the publication
- Read One Hotel, Seven Versions online
- Download the approved PDF from the Research edition
- Trace one consequential promise with a Seven-Version Snapshot
- See Channel Lens
Publication note
This is a first-party Tamaga publication announcement.
The white paper is a public-source forensic study and an illustrative case, not an audit commissioned by Hôtel Mont-Blanc Chamonix, a hotel rating, a legal-compliance review, an assessment of commercial performance, or a statistical estimate of the accommodation sector.
Time-sensitive observations in the paper are dated and bounded. The online edition should carry the current methodology, sources, figures, corrections and version history.
Context references
- Google Search Central. Optimizing your website for generative AI features on Google Search. 2026.
- Google Search Central. Introducing Search Generative AI performance reports in Search Console. 3 June 2026.
- Pew Research Center. Google users are less likely to click on links when an AI summary appears in the results. 22 July 2025.
- Tamaga. One Hotel, Seven Versions. Public-source forensic edition, August 2026.
Corrections
No corrections have been issued for this draft.