Multi property AI visibility for hotel groups: what aggregates and what does not
A group is judged on the portfolio and answered on the property. Any tool that hides that gap behind one average is selling you a number you cannot act on.
AI sessions measured across the sites we track, last 30 days.
Quick answer
Multi property AI visibility means measuring, property by property, whether assistants name each hotel, and being able to read the portfolio without opening 40 tabs. In reporte.ai the traffic side aggregates today: an All websites mode adds every property together and breaks AI sessions down property by property, and a single API call returns them all. The answer side does not aggregate: questions are measured 1 property at a time, on purpose. On 4 September 2026 one account in the platform held 87 of the 100 websites being measured, so the multi property account is running at that scale rather than being a diagram.
What a group buys that a single hotel does not
A single hotel wants to know whether it is named. A group wants to know which of its properties is the outlier, and that only works if every property is measured the same way. Comparability is the actual product here, and it is fragile: change the question set for one property and its series stops meaning anything next to its siblings.
So the group setup fixes three things and keeps them fixed. The questions per property, the surfaces, and the cadence. In the pass of 3 and 4 September 2026 that meant the same 57 questions on the same 5 surfaces across every domain, 285 measurements, 0 errors.
What aggregates today and what does not
Aggregated across all properties: AI traffic, sessions, page views, pages per session and the count of websites, plus the per property breakdown and the landing pages assistants send visitors to. Available in 1 API call with a Bearer key, so a BI tool or a monthly deck can pull the portfolio without anybody logging in.
Not aggregated: the measurement of what assistants answer. When the panel is in All websites mode it asks you to choose a property before it measures, and that is deliberate, not missing plumbing. A question about a city hotel in Seville and a question about a resort in Tenerife are different questions, and averaging their answers into one portfolio percentage would hide the property that is failing, which is the only property the report existed to find.
If a single portfolio wide mention rate is what your board wants on a slide, say so early. Today you would build it yourself from the per property figures, and we would rather tell you that now than after the contract.
Measuring one surface makes the portfolio look better than it is
In the pass of 3 and 4 September 2026 the same 57 questions surfaced the measured domain 10 times on ChatGPT, 14 on Google AI Mode, 15 on Gemini, 16 on Claude and 22 on Perplexity, across 7 domains of our own travel network rather than one property. Between the lowest and the highest surface there is a factor of more than two, and a single reading proves nothing: the gap counts once it repeats week after week.
Applied to a portfolio, that is not a rounding error. A group that measures ChatGPT alone and reports it as AI visibility is describing one fifth of the evidence, and the property it decides to invest in may be chosen on the surface where it happens to look worst. The 5 surfaces are measured together for that reason. The traffic side says the same thing from the other end: across the sites we measure ChatGPT sends around nine of every ten AI sessions of the last 30 days while it names the least, so where the visits come from and where the mentions come from are not the same ranking.
The brand you own and the brand the OTA rents
Scale does not protect a group inside an answer. In a sweep on 3 September 2026 we found Booking.com ads placed inside ChatGPT answers with hotel names in the ad title, and 0 of the 5 Spanish chains we checked, Meliá, RIU, Barceló, Iberostar and Palladium, had an ad of their own in those answers.
For a portfolio that is a leak repeated per property. The fix is not to outbid anyone, it is to make each property's own pages the thing the assistant reaches for, which is what the readiness score measures per site and the content plan turns into work with a number behind it.
For traffic, yes. An All websites mode adds every property together and breaks AI sessions down property by property. For the answer side the measurement runs 1 property at a time and does not aggregate today.
Because a question about a hotel in Seville and a question about a hotel in Tenerife are different questions. Merging them into 1 portfolio percentage would hide the property that is failing.
On 4 September 2026 a single account held 87 of the 100 websites the platform was measuring, and the All websites mode was serving that account.
You get a portfolio picture that is wrong by more than double. In the pass of 3 and 4 September 2026 the same 57 questions, spread across 7 domains of our own travel network, surfaced the measured domain 10 times on ChatGPT and 22 times on Perplexity, and one pass alone proves nothing.
Yes. A Bearer API key returns AI traffic for every website in the account in 1 call, with the per property breakdown and the host of each one, so a portfolio report does not need anybody logging in.
Not by size. In a sweep on 3 September 2026 we found Booking.com ads inside ChatGPT answers carrying hotel names in the ad title, and 0 of the 5 Spanish chains we checked had an ad of their own.
By keeping the question set and the surfaces fixed so the series is comparable. The pass behind these figures used the same 57 questions and the same 5 surfaces on every domain, with 0 errors.