AI Visibility Without a Direct Booking Path Still Sends Revenue Elsewhere.

Three different outcomes, one word

"Visible" is doing far too much work in this conversation. It covers three states that have entirely different commercial consequences, and conflating them is how hotels end up celebrating a result that costs them money.

Mentioned. The answer names the property. Something in the retrieved material knew it existed and thought it relevant.

Cited. The answer attributes something to a source, and that source is reachable. The traveller can click it. Whose domain it is decides who owns the next step.

Bookable. There is a visible route to reserve, and it belongs to you.

A property can be mentioned in every answer and bookable in none. That is not a partial success. It is a fully-functioning demand generation engine pointed at somebody else's checkout.

The failure that looks like a win

Picture the meeting. The commercial director has run the questions. The property is named, described accurately, positioned well against two competitors. Someone screenshots it.

Then look at what the answer actually cited: an OTA listing page and a review platform. And what it offered: a link to the OTA.

The property has just been recommended, on the strength of a product it spent years building and a reputation it earned, and the traveller has been handed to a channel that will take fifteen to twenty per cent of the resulting booking.

This is worse than absence in one specific sense: absence is a slow content problem that everyone can see. This one looks like success, gets celebrated, and continues.

Official source versus OTA source

The metric almost nobody publishes, and the one that translates immediately into a conversation hotels have been having since the mid-2000s.

Of the answers that cited anything at all:

• how many cited your own domain?
• how many cited an OTA?

Both over the same denominator — answers that cited any source — because answers showing no sources at all belong in neither number.

In our fictional sample audit, five answers cited something. Two cited the official site. Three cited an OTA. The property was named in four answers and a direct route was visible in one.

Read those numbers as a commercial director rather than a marketer: the property is not invisible. It is being described. And it is being described through the channel that charges it.

Why does an OTA get cited over the hotel? Usually one of three reasons, and all three are addressable:

The OTA page is genuinely better. More structured, more complete, more specific about room types, policies and amenities than the hotel's own page. This is embarrassingly common. Hotel room pages describe atmosphere; OTA listings describe facts. Answers need facts.

The hotel's page cannot be reached or read. Crawler blocked at the CDN, key content rendered client-side, the fact in a PDF. The OTA has no such problem.

The OTA carries more corroboration. Reviews, structured attributes, consistent data across a large domain. Authority is partly borrowed from the surrounding site.

Only the first is a content problem. The second is access. The third is source authority, which is slow. Most hotels have the first two, and they are fixable.

Landing and booking continuity

Suppose the official link wins. The work is not done, because the click lands somewhere.

The traveller arrived with a specific question — the anniversary weekend, the accessible room, the restaurant a local would book — and the answer promised the property could handle it. If the landing page is the homepage carousel, the thread is broken at the first step. They came with a question and were handed a brochure.

Then the booking path. This is territory hotels already know is weak, and the AI-sourced visitor makes it worse: they arrive later in the decision, with more confidence and less patience, because something they treated as advice has already vouched for the property.

The recurring breakages:

• Homepage landing rather than the room, offer or outlet the answer described
• A second search — dates re-entered, filters re-applied, work already done being demanded again
• Room names that mean nothing. "Deluxe Superior Executive" is not an answer to "which room has the harbour view?"
• Rate opacity. Taxes, resort fees and city levies appearing at step four, after the comparison was made at step one
• Policy uncertainty. The pet, child or accessibility question that motivated the whole search, unanswered on the booking path
• Mobile behaviour. Date pickers that fight the thumb, forms that zoom, engines that lose state on back-navigation

Each of these is an ordinary conversion defect. The reason to re-examine them now is that AI-assisted discovery raises the value of the traffic reaching them, which raises the cost of every one.

Room, rate and policy clarity

A specific point that sits between content and conversion, and gets missed because it belongs to neither team.

If an answer describes your property as suitable for a family with a toddler, and your booking path cannot confirm a cot, the answer has created a lead you cannot convert. The recommendation was accurate. The path could not honour it.

The reconciliation is worth doing explicitly: take the claims answers are making about your property, and check whether the booking journey supports each one. Where it does not, either the claim needs correcting at source or the path needs fixing. Both are actionable. Neither happens by itself.

Hotel Center and free booking links

The plumbing, and the part that no amount of content work substitutes for.

Google's free booking links exist to place a hotel's official site alongside OTA options in its hotel booking modules, at no cost per click. Eligibility runs through Hotel Center participation with rates available and a working landing page, and the rate and availability feed generally arrives via a connectivity partner. Integration timelines are typically reported in weeks rather than months.

The commercially useful observation: many hotels are already connected through their booking engine or reservation provider and are unaware of it. Others are eligible and simply not appearing, for reasons that are usually mundane — a feed that stopped, a mapping mismatch, an account nobody has opened since a previous commercial director left.

If the official rate is not present in the module, the traveller is shown channels. That is not a visibility problem, a content problem or an AI problem. It is a distribution problem, and it is the reason this capability sits beside distribution rather than beside marketing.

On agentic booking, the honest position: Google's Universal Commerce Protocol now covers lodging, the hotel remains merchant of record, and Google's own FAQ states that participating does not influence ranking. It is early. Assess readiness, claim nothing about eligibility not confirmed on your own account, and treat anyone selling protocol readiness as a ranking lever as having not read the documentation.

The direct economics

Illustrative arithmetic with stated assumptions. Not a client result, and not a benchmark.

Take a property with 118 rooms at 72% occupancy and an ADR of 200 in local currency. Annual room revenue is roughly 6.2 million.

Assume 55% arrives through OTAs at an average 17% commission. That is roughly 580,000 in annual commission.

Now assume a shift of five percentage points of channel mix from OTA to direct — 55% to 50%. That is about 314,000 of revenue moving channel, and roughly 53,000 of commission not paid.

Against that, direct is not free: the booking engine, payment costs, brand-search defence and the marketing to support it. Call it 30% of the saving, generously. Around 37,000 of contribution retained, annually, recurring.

The assumptions are all arguable and should be replaced with the property's own. The structural point is not:

A five-point channel shift is not a marketing campaign. It is the cumulative result of the official rate being present where the decision happens, the official path being visible in the answer, and the booking journey holding together once taken. Every one of those is a distribution and conversion decision, and none of them is content.

That is the entire argument for treating AI discovery as a commercial system rather than a visibility exercise.

What to measure

Do not measure "AI visibility." Measure the chain, and give every number a denominator and a date:

| Metric | Denominator | Reads as | | --- | --- | --- | | Shortlist rate | Answers observed | Are we considered? | | Citation rate | Answers where citation behaviour was observable | Is anything sourced? | | Official-source citation rate | Answers citing any source | Are we the source? | | OTA-source dependence | Answers citing any source | Who owns the next click? | | Direct-path visibility | Answers where a route was observable | Can they book with us? | | Answer accuracy | Assertions with a decided verdict | Is what they were told true? |

Then, separately, from first-party data: AI-attributable referral traffic where identifiable, landing-page continuity, booking-path completion by source, and direct share over time.

Two honest caveats. First, AI referral traffic is imperfectly identifiable and will stay that way — some surfaces pass no referrer. Second, Google's Search Console generative-AI report currently gives impressions without clicks, click-through rate or query data, and has been rolled out selectively by region; assume it is unavailable on a given property until it is checked.

Which is exactly why the observation panel exists. It is the only method that currently answers which question produced this, and it does so at the cost of being slow and manual — a trade we would make again.

The one-sentence version

Being mentioned is a marketing outcome. Being cited from your own domain, with a visible official path and a booking journey that holds, is a commercial one — and only the second shows up in the P&L.

If the audit stops at whether AI mentions the hotel, it has stopped one link before the money.

---

Sources and limitations. Google Hotel Center Help on free booking links (eligibility, no cost per click); Google for Developers, Universal Commerce Protocol for Lodging FAQ (merchant of record; participation does not influence ranking); Google Search Console Help on generative-AI performance reporting (available dimensions and rollout limitations); Google Search Central guidance on AI features (crawling must be permitted by CDN and hosting infrastructure). The economic illustration above uses invented figures with every assumption stated, and is not drawn from any engagement. The sample audit figures come from a fictional property. Katalyst has no disclosed hotel AEO client results.

Ready to see where your own chain breaks? Request an AI Visibility Snapshot.