Is your hotel showing up in AI answers? What Search Console and Bing report, and what they miss

Partly measurable. Google Search Console's generative AI performance reports and Bing Webmaster Tools' AI Performance report show how often your pages appeared or were cited in those companies' AI answers. AI visibility is whether AI answers mention or link your hotel. Neither report shows the question asked, clicks, or other assistants, so check answers by hand too.

A vendor offers to track your hotel's AI visibility for a monthly fee. Two reports you already have access to, from Google and from Bing, show part of the same picture at no cost. Knowing exactly which part they show tells you what the tool is really selling. After reading this guide you will know how to open both reports, read them page by page for a hotel website, and decide whether a paid tool adds anything.

What does AI visibility mean for a hotel, and how do you track it?#

AI visibility is whether AI answers mention, cite or link your hotel when a traveler asks about a stay. It is measured in three layers, and each layer answers a different question.

Layer one is hand checks. Type the questions a traveler would actually ask into the assistants people already use, then record the answer, whether you appear, what is said about you, and which sources are linked. This is only repeatable if you keep the same question list and note the date. If you have not built that list yet, start with keeping your hotel's facts consistent across AI assistants.

Layer two is search engine reports. Google Search Console and Bing Webmaster Tools now publish how often your pages appeared inside their own AI features. These are free, first-party and limited to those two companies' surfaces.

Layer three is analytics. Visits and bookings that arrive with an AI assistant as the referrer, plus the many that arrive looking like direct traffic. That setup is described separately in the guide on measuring visits and bookings from AI assistants in analytics.

The reason for three layers is simple. Layer one tells you what is said, layer two tells you how often your pages are used as a source, layer three tells you what it produced. Skip a layer and you mistake activity for outcome, or outcome for quality.

Two mistakes come up constantly. The first is treating one check on one day as a measurement: answers vary by user, location, language and date. The second is counting mentions without recording what is said. A mention carrying a wrong room count or a restaurant that closed last winter is worse than no mention, because it is confidently wrong.

What do Google's generative AI performance reports show?#

Two reports in Search Console count how often links to your site were shown inside Google's generative AI features: one for Search and one for Discover.

They were announced on the Search Central Blog on 3 June 2026 and first went to a subset of websites. A note added to that post says they reached all websites worldwide as of 31 August 2026, and Search Console Help (opens in a new tab) says the same. Check Help before you rely on the labels, because both tools are new and definitions are still moving.

The Search report counts impressions, meaning how many times links to your site were shown in a generative AI feature, covering AI Overviews and AI Mode. Impressions only: no clicks. Its dimensions are pages (the final URL after redirects, mostly the canonical URL), countries where the search originated, dates in Pacific Time, and devices (desktop, tablet, mobile).

Two search type filters split text-based web results from multimodal ones, for example Google Lens and Circle to Search. The report excludes Search Labs experiments, has the usual 1,000-row table limit, and can be exported. It may not appear at all if your site has too few such impressions or has been excluded through the Search generative AI control.

The Discover report counts impressions in generative AI features in Discover, by page (the canonical source page), country and date, with no device breakdown. The June post says Google may add metrics over time.

What does Bing's AI performance report show?#

Bing Webmaster Tools' AI Performance report counts how often your content was cited as a source in AI-generated answers across Microsoft's assistant, AI summaries in Bing and unnamed partner integrations.

It arrived as a public preview on 10 February 2026 on the Bing Webmaster Blog (opens in a new tab), with further features on 16 June 2026 and 13 October 2026. Confirm the current labels inside the tool before you quote them to anyone.

Total Citations counts how many times your content was shown as a source in the selected period. It does not show placement or presentation. Average Cited Pages gives the average number of unique pages from your site cited per day, and page-level activity gives citation counts per URL.

Grounding queries are the key phrases the AI used when retrieving content that was cited. Treat them as a sample of citation activity, not a full log of what travelers typed. The 16 June 2026 additions classify grounding queries into intents (informational, commercial, navigational, local) and topics, add Citation Share as your percentage of all citations for a grounding query, and add a Compare view that overlays an earlier period.

The 13 October 2026 Topics Portfolio (beta) adds Topic Size, Sites Cited and Queries per topic, plus Opportunity Queries where other sites appear and yours does not, with volumes shown as ranges. Bing says the topic classification is AI-generated and the numbers approximate and directional, and hides topics without enough data.

Bing states that the counts do not indicate ranking, authority or a page's role in an answer, and clicks and traffic are not among the listed metrics.

How do you read these reports for a hotel website?#

Read them page by page, not site-wide. The useful signal is which of your pages are being used, for which kind of question, in which country.

Before you open either report, sort your pages into four groups: room and rate pages, location and getting-here pages, practical and policy pages (check-in times, parking, pets, cancellation), and food, spa and meeting pages. The grouping turns raw numbers into decisions, because each group needs different work. If the underlying pages need attention first, work through the hotel website SEO checklist for pages and technical basics.

In Google's report, filter by page and country together. A German-language page appearing in Germany is a different result from an English page appearing in the United States, and the two call for different work.

In Bing's report, read grounding queries next to their intents. Commercial and local intents sit closest to a booking decision, so a page cited there matters more than a page cited for a general question about the city.

Look for the pages that never appear. A room page with no impressions in either report is usually a content or structure problem you can fix, not a mystery about AI. Compare months, not days: on a single hotel site, a date range shorter than a month is mostly noise. And write the export date next to every number, because both tools are new and their definitions and labels are still moving.

What do these reports not show?#

They do not show the question as the traveler typed it, they do not show clicks, and they cover only two companies' AI surfaces.

Google's reports have no query dimension at all. You learn that a page was shown, never what was asked. Bing gives grounding queries, but as a sample, and they are the phrases the AI used to retrieve, which are not always the traveler's words.

Clicks are absent from both. Impressions and citations are not visits. To see visits you need analytics, and much of that traffic will arrive looking like direct traffic rather than a referral.

Coverage is partial by design. Google's reports cover Google's features; Bing's cover Microsoft's assistant, Bing's AI summaries and unnamed partner integrations. Answers given in other assistants appear in neither. Placement is missing too: a citation does not say where in the answer you appeared or whether you were recommended at all.

Competitors are mostly hidden. Citation Share does not reveal competitor domains, and Opportunity Queries shows that other sites appear, with volumes as ranges, not who they are. And small sites can see nothing: Google may withhold the report below an impression threshold, and Bing hides topics without enough data.

Should you pay for an AI visibility tool?#

Only after you have run the free reports and a hand check for a month, and only if the vendor will explain its method and show its question list before you sign.

Understand what such tools generally do: they send a fixed list of prompts to several assistants on a schedule, record whether and how a brand appears, and turn that into a score. The score is only as good as the prompt list and the sample behind it.

Ask for the method in writing: which assistants, which languages, which locations, how many prompts, how often, and whether the raw answers are stored so you can read them. Ask for the question list before paying. If it does not contain the questions your front desk actually hears, the score measures the vendor's imagination, not your visibility.

Ask what a mention means in their counting: a link, a name, a description, or a recommendation. These are different outcomes and should be reported separately. Then run a comparison. Take the trial or free sample and check ten of its answers yourself on the same day. If your results disagree, ask why before you discuss price. Finally, ask how prompts and answers are stored and who can read them, since they may contain commercial detail about your property.

If you are weighing an agency rather than a tool, the same logic applies: see the checklist for hiring an agency for AI search.

What is one monthly line to add to your marketing report?#

Add one line that states what the reports showed, what you checked by hand, and what you changed, with the date, so the number always arrives with its context.

Use a fixed template so month-to-month comparison is possible without re-reading old decks. Keep the same question list month to month. Add questions as you learn them, but never silently replace the old ones, or the trend disappears.

Record the export date and the exact report labels you used, because both tools are new and their names and definitions still change. Report the pages that stopped appearing as well as the ones that started: a page losing citations is the earliest warning you will get.

Do not convert any of these numbers into a revenue claim. There is no verified way to attribute a booking to an AI answer, so say what you measured and stop there.

What to do next#

Export both reports today, write down ten questions your front desk actually hears, and run them by hand on the same day. That gives you a baseline in all three layers within a week, and it turns any later conversation with a paid tool into a comparison rather than a leap of faith: the reports tell you how often your pages were used, the hand check tells you what was said, and analytics tells you what arrived.