What should a monthly AI visibility report include?

One page, six lines, in this order: the questions asked and in which engines, which engines named the hotel, who was named instead, the score and its direction, the fixes made this month, and the one fix for next month.

Owners do not read dashboards. They read one page, and they read it for one answer: is our hotel in the answer when a guest asks, and if not, who is. Everything on the page serves that answer.

  1. What we asked, and where. The questions a guest types, in the words a guest uses, and the engines checked: ChatGPT, Perplexity, Gemini, Claude and Google AI. “Quiet boutique hotel in Zurich near the lake, under 400 francs” is a question; “Zurich hotels” is not. List ten questions, the same ten every month, so the page compares like with like.
  2. Named, or not. For each question, which engines named the hotel. A grid of ten questions by five engines, a check or a blank in each cell, tells the owner everything in one look.
  3. Who was named instead. The three hotels each engine gave when it did not give ours. Owners know these names, and this line turns an abstract metric into a competitor with a face.
  4. The score, and which way it moved. One number, from zero to a hundred, with last month’s beside it. The AI presence score is one way to build it; any consistent method works, as long as the same method runs every month.
  5. What we fixed this month. Three lines at most: the category on the Business Profile, the description on the website, the reviews answered. Fixes are the only part of the page the owner is paying for.
  6. The one thing for next month. One fix, named, with the reason. An owner who reads six pages of plans trusts none of them; an owner who reads one line and sees it done next month trusts the next line.

That page is the deliverable of every program on this site. The appendix, with the full answers recorded per engine and the dates, exists for the day an owner asks “show me”, and for nobody else.

How do I explain AI search visibility to a board?

Say it in one sentence, then show the names: when a guest asks ChatGPT, Perplexity, Gemini, Claude or Google AI where to stay in our city, three hotels get named, and here is whether we are one of them.

A board does not need to know how an engine works. It needs to know that guests ask engines, that engines answer with three names, and that the hotel is or is not among them. The fastest way to make that real is to put the actual answer on the screen: the question, the three names, the date. In every board we have sat in, the first reaction is the same, someone reads a competitor’s name aloud.

Then answer the two questions that always come next. “How many guests ask this way” is the one nobody can answer with a number, and the honest reply is that the engines do not publish it, and the hotel’s own front desk can ask every guest how they found the hotel, which is the only count that exists. “What does it cost to be in the answer” is the program page for the hotel, and the hotels page says the three prices plainly.

Avoid the two things that lose a board. The first is jargon: no “entity”, no “schema”, no “retrieval”. Say “the facts Google holds about us” and “the words guests use in reviews”. The second is a promise with a date. Six to eight weeks until the first engine names the hotel is the range we see; say the range, and say that the score keeps moving after that.

Share of voice is the share of answers in which an engine names your hotel, across the questions you track and the engines you check, in a month.

Ten questions in five engines is fifty answers. Named in twenty of them is a share of voice of forty percent. Named in none is zero, which is where most independent hotels start. The number is simple on purpose: an owner can recompute it from the grid on the first page, and it compares the hotel to the ones named instead, because their share of voice comes from the same fifty answers.

Two things make the number honest. Ask the same questions every month, so the denominator does not move. And record the answers with the date, because engines change what they say, and a share that fell may have fallen because the engine changed, not because the hotel did. The monthly tracking page covers how to record them so the history holds up.

How do I show owners the bookings that started in ChatGPT?

You cannot see them directly, because a guest who asks an engine then books direct or on Booking, and the referrer is lost on the way.

That is the fact to state first, before an owner discovers it alone. The engines do not send a tracking parameter with a guest, and the guest often closes the chat, opens the hotel’s site by name, and books. In the booking engine that guest looks like a direct booking with no source. Whether AI is sending you guests sets out what can and cannot be measured.

Show the two things that can be seen:

  • Direct bookings against the timeline. Put the month the engines started naming the hotel on the chart of direct bookings. A rise that starts there is not proof, and it is the strongest signal an owner will get.
  • What guests say at the desk. “How did you find us” asked at check-in, with “ChatGPT” and “Perplexity” as answers the staff can tick, is the only count of AI-referred guests that exists. Ask it for three months and the owner has a number that no dashboard offers.

Say plainly on the report that the bookings line is an estimate built from those two signals, and keep the estimate in the appendix. The first page stays with what was measured: named, or not.

What goes in a quarterly review for a group?

One table: every hotel down the side, the five engines across the top, a check or a blank in each cell, and a score per hotel.

A group board has the same question as a single owner, multiplied by the number of hotels, and it has less time. The table answers “which of our hotels are in the answer” in one look, and the blanks answer “where do we work next”. Under the table, three things:

  1. The three hotels that moved most, up or down, with one line on why. A hotel that fell because a chain opened next door needs a different plan from one that fell because its Business Profile lost its category.
  2. The fixes per hotel for the next quarter, one line each. The board approves a list, not a strategy.
  3. The pilot, the standard, the rollout. For a group that is starting, the review says which three hotels were the pilot, what the standard is now, and which hotels come next. The group panel on the hotels page shows that shape.

One caution for groups: do not average the score across hotels. An average hides the two hotels that are invisible behind the eight that are fine. Show every hotel; the board can read a table.

The page, when you have to write it tomorrow

If the first report is due and nothing is set up, do this: pick ten questions a guest would ask about your city and your kind of hotel, ask them in the five engines, write down the three names each gives, and count how often yours appears. That is the grid, the competitors and the share of voice, and it takes an afternoon. The free check on this site does the first pass for you: it asks the five engines about your hotel and your city and sends you which hotels they named and what those hotels had, and that is the first page of the first report.

See also: The AI presence score explained | Track hotel AI search visibility | AI presence score