What we do
When a guest asks an AI where to stay, what does it have to go on?
Presence Pro answers that for your hotel, shows you who gets named instead, and turns it into a short plan. Then it checks again every month.
Does your own site answer what guests ask?
An AI assistant leans on your own pages for facts. We read up to 14 of them and check whether they answer the questions guests actually ask, quoting the exact words and the page, so you can see what we relied on.
- Parking, breakfast, check-in and out, pets, accessibility, family rooms, getting there, Wi-Fi, EV charging, cancellation
- Where an answer is missing, the existing page it belongs on and wording to add
- Pages that disagree: two check-in times, two phone numbers, an expired offer
- Whether the site ties your hotel to the stations, airports and landmarks that matter
Machine access, checked plainly
What your robots.txt asks of AI crawlers, whether you publish an llms.txt (an informal proposal, so its absence is never a fault), whether the public web archive Common Crawl could read your site, and Google's mobile speed and accessibility scores.
Structured data, with the right caveat
We check what your markup says (a wrong phone number, bad coordinates), not just that it exists. We never present markup as a ranking lever, because nobody has shown that it is.
We ask an assistant the questions your guests ask.
Not "best hotels in Paris", which only big names win, but questions written from facts about your hotel: its kind (airport, spa, city-centre, resort), its price level, the named places it is close to, and what its own pages show it offers.
- Questions never contain your hotel's name, and are never so exact that only you could fit
- A four-star hotel is never asked a luxury question
- Each question is asked more than once, so one odd answer doesn't decide the result
- You can add questions of your own, put exactly as written
What you see
How often you were named, in what position, who was named instead and how comparable they are, what the assistant says when asked about you by name (and whether it's right), and the exact pages it cited.
What it means
A labelled sample of one assistant with live web search, on one day, shown with a plausible range. It is never a market ranking, and we never claim to know why an assistant chose a hotel.
Where is your hotel listed, and does it agree?
Assistants also draw on booking and review sites, maps, open data and press. We check 30 points across them, and say plainly what we could and couldn't verify.
- Booking and review sites, identified from public search results
- OpenStreetMap and Wikidata, matched only on name plus coordinates, postcode, website or phone
- Press and publications, weighed by quality, with sponsored pieces counting for nothing
- Awards, claimed against issuer-confirmed
- For UK hotels, the official food hygiene rating
Where it doesn't agree
Check-in times, star ratings and telephone numbers that travel sites show are compared with your own pages. A difference is marked "worth checking", because the listing may be the right one.
What we won't read
Booking.com and Tripadvisor prohibit automated collection, so we don't read their pages. Where we can't see something it's marked not verified, never missing.
Your own review export, analysed in full.
Every platform lets a hotel export its own reviews. Give us that file (a CSV or Excel sheet from Booking.com, Tripadvisor, Google or Expedia) and we analyse the whole thing.
- Reply rate overall, by platform and by rating, and how long replies take
- Critical reviews still unanswered, newest first, with a skeleton reply to finish
- How replies read: specific, thankful, apologetic, signed, or copy-pasted templates
- What guests praise and criticise by topic, and whether your site says it
Private by design
Reviewer names are never read. The review text is analysed for that run only and is not saved; only counts and rates are kept with your report.
Never an incentive
Draft replies never offer anything in return, and we never suggest fake or incentivised reviews.
Threads where travellers are asking right now.
We find recent threads on Reddit, Quora, Travel Stack Exchange and similar where someone is asking where to stay, and tell you whether each still takes replies.
- One card per thread, with a link to open it
- Honest wording to use, built from facts we've verified about your hotel
- Threads that are archived are listed apart
The rules we give you
Answer once, as the hotel, never as a guest. Never seed reviews or offer anything in return. Presence Pro never posts anything for you.
A few things, in the order worth doing them.
Every finding is scored on the risk to your reputation, how much travellers care, the effect on discoverability and how sure we are, then moderated by effort. Risk counts most: a poor food hygiene rating, a wrong phone number or a site that blocks crawlers outranks tidying markup.
- At most five actions in the management report, plus a low-effort list and a 30, 60 and 90 day plan
- Required fixes (wrong or conflicting information) kept apart from optional commercial opportunities
- Each with what we found, why it matters, the evidence, what to do, where, and an example
- Labelled Confirmed, Likely or Worth checking, so you know how much weight to give it
The score, honestly
A readiness score out of 100, always shown with how much of the full model it rests on, for example "72, provisional, 60% measured". Anything not measured is left out, not scored as zero.
Withheld when weak
A recommendation with no evidence, no specific action, or generic wording is withheld, and the withheld list is kept so nothing disappears silently.
See what changed, and trust the change.
Your core questions are frozen at your first report and asked again word for word every month, so this month is compared fairly with last.
- Results shown as bands (not appearing, rarely, sometimes, usually) with a plausible range
- A change is called only when the ranges don't overlap, otherwise it's "too early to say"
- Past reports comparable side by side
What you receive
A management PDF of about 5 to 10 pages that answers the key questions on page one, a technical and evidence appendix with everything behind it, and a data file.
What we don't do
Clear limits, so you can trust the rest.
We can't promise rankings
No one can promise an AI assistant will recommend a hotel, and we won't pretend to.
We don't test every assistant
The sample comes from one assistant. For ChatGPT, Gemini, Copilot and others you get a checklist to test by hand.
We don't post for you
No posting, no commenting, no review requests sent on your behalf.
We don't guess
"Not assessed" and "not found on the pages checked" are real answers. We'd rather say we couldn't tell.
See it for your own hotel.
Read a full sample report, or see the plans.