What should an AI visibility report include?
See what an AI visibility report should show, from customer questions and cited pages to local rivals, useful changes, and gaps that a thin report can hide.
A report you can read
What this guide shows
A useful report gives a local owner enough context to understand each finding and make a business choice.
Each row names the question, the market and the date it was checked.
Your business and local rivals appear as the answer showed them.
The pages shown as sources sit next to the question that produced them.
New names, lost names and settings changes carry dates.

Short answer
The Answer
A useful AI visibility report shows the questions your customers ask, whether AI names your business, which of your pages AI cites, how you compare with local rivals in plain terms, and what changed since the last report. Our founder, an AEO specialist, reads reports this way.
Each finding names its platform, market and date. The founder closed over $100,000 in commission in July 2026, which he credits to AEO work on his own sites.We did it for our own business. Let's find out how to do it for yours with a free audit.
What does a useful AI visibility report show first?
A useful report is a record of what AI answers said about your business on a set of dates. It is not a score out of a hundred and it is not a chart of colors. It is a list you can read. Each row holds a question, a platform, a date, the names AI gave, and the pages AI showed. If one of those is missing, you cannot tell what you paid for.
Two kinds of data feed a report. The first comes from the platforms. Google added a Generative AI performance report to Search Console, and as of August 31, 2026 it is open to all websites worldwide. Bing Webmaster Tools has an AI Performance report, in public preview since February 10, 2026. The second kind is a saved list of questions that someone asks the AI services and writes down. A strong report uses both and labels every row with where it came from.
A report earns your trust by showing your real starting point, and our founder describes its job in exactly those terms.
“I can't promise you the exact result that I'm getting from this AEO service, but I can show you where you stand.”
Our founder, November 2026 interview
That starting point has five parts, and the next five sections take them in order. First, the questions your customers ask and whether AI names you. Second, the pages AI cites. Third, how you compare with local rivals. Fourth, what changed since the last report. Fifth, what a thin report leaves out. Plan on one sitting to read all five, with the question list first.
Start with Google's own figures, because they set the pattern for the rest. Its report holds four dimensions: Pages, Countries, Dates and Devices. Google counts an impression each time a user sees a link to your site within a generative AI feature.
Which questions did AI get asked, and did it name you?
The question list is the base of the report, so read it before anything else. Good questions sound like your customers, not like a keyword list. They carry the exact words a buyer would type, such as “who is the best probate realtor in los angeles.” If the list holds only broad phrases, the report tells you little about the people who call you.
Each question also needs a market. Google says your location shapes results even when you type no place, using device location with permission, saved Home or Work addresses, past activity and your IP address. ChatGPT search uses an approximate location from your IP address by default. A report built from the wrong city describes the wrong customers.
Example, with the row filled in. Question: “who is the best probate realtor in los angeles.” Platform: Google AI Overview. Date: October 29, 2026. Result: the answer listed our founder first and cited the city probate pages on his own real estate site. That is his note from one public search, not a Search Console figure. On November 10, 2026 a re-check listed him first again, but an earlier load that evening listed him second. The answers shifted between loads, so a good row records the load.
Read the “named” column in three states: named first, named lower, or absent. Absent is the most useful row, because it points to a question where a customer asked and AI sent them elsewhere. Gemini adds a wrinkle. With Memory on, it learns from a signed-in user's chats, so a check run from a personal account differs from a stranger's. Ask which account and setting the checks used. One answer is not proof, so a good report asks each question more than once.
A market label looks like OpenAI's own example. A “restaurants near me” prompt from the San Francisco area is rewritten to “top restaurants San Francisco.” A Los Angeles row needs the same kind of label, such as Los Angeles area, signed out, Memory off.
Curious what an answer says about your business? Does ChatGPT name you?
Which of your pages did AI cite, and where?
The pages section lists the addresses AI showed as sources. Google's report lists them by canonical page address with an impressions count. Bing's report lists page-level citation activity. Your own saved checks add the question that brought each page up. A page list with no question beside it tells you what was cited and nothing about why.
Platforms show sources in different ways, and the report must say which one produced each row. Perplexity puts numbered citations on each answer. Gemini shows a Sources button only sometimes, so no button means no links for that answer. ChatGPT responses that use web search may include citations, and OpenAI says those citations can be incomplete, outdated or incorrect. A good report lists “no sources shown” as its own result.
Illustration of a row, not a real result. Platform: Perplexity. Question: “water heater repair in Pasadena.” Date: a Tuesday in the check month. Page shown: your Pasadena repair page, numbered citation 2. Next to it, a column for the page you hoped to see. When the two differ, you have found a page job.
Three steps turn the page list into work. Sort the rows by page. Circle the page that appears most often. Open it and check that it names your service, your city and a way to reach you. Then look for pages you did not expect. They show which of your pages AI treats as an answer, and which ones sit unused.
Eligibility is the other half. Google says a page must be indexed and eligible to show in Search with a snippet to appear as a supporting link, and that no new machine-readable files, AI text files or markup are needed. A page missing from your list needs an indexing check before it needs a rewrite.
How do you compare with local rivals, in plain terms?
A rival comparison is the same question list run against the other names in the answers. It uses plain words: who was named, in what order, and whether you appear. It does not invent a score. The report should hold the same questions, the same market, the same dates and the same platforms for every business in the table.
Sample layout, not data. Question: “best probate realtor in Pasadena.” You: named second. Rival A: named first. Rival B: not named. Pages shown: your city page and one guide from Rival A. Each cell is a fact someone saw on a date. The last line tells you what to build next, here a guide that answers the question Rival A's guide answers.
The market matters here more than anywhere else. Google says results differ by time, location, language, device and personalization. Checking a city from another city, or by typing the city name, is not the same as a customer searching from that city. A rival table made from the wrong place compares the wrong shops.
Scoreboards from outside checks are the shape to ask for. The HomeLight outside-in check on October 29, 2026, two weeks after launch, found 7 of 8 probate and inherited-home pages ranked first and cited in Google's AI Overview. It used one search per article, one day and public results only, so it states its own limits in the same breath as its result.
Your city and trade shape the table. Is your city or industry still open?
What changed since the last report?
The first report is your baseline, and every later report compares itself to it. A change line lists four things: questions where you are newly named, questions where you dropped out, pages added or lost from the cited list, and platforms that began or stopped showing sources. Put each next to its date.
The platform reports help here because they keep dates. Google's report shows dates by day, week or month in Pacific Time, and the newest data is preliminary. Bing's report shows trends over time. Use date ranges of the same length so a short month does not look like a drop.
Next to each change, write the cause: a page you published, a setting you changed, or nothing you did. A change with no cause is a note for next month, not a decision. Keep your own change log with the report, because a platform report cannot know what you edited.
One setting deserves a line in every report. Search Console has a Search generative AI control under Settings that excludes your site from AI Overviews, AI Mode and generative AI features in Discover. Excluded sites get no impressions from those features but stay in regular Search, and a subdomain follows its closest parent unless its owner changes it. Google says a change usually applies within a day or two, sometimes longer.
See the story behind our own result. Read our founder’s case study
What does a thin report leave out?
A thin report is a score, a screenshot or a colored chart. It shows a result and hides the work behind it. Hold it against the five parts. It leaves out the question list, the platform for each finding, the market and dates, the pages shown, the rival names and the change line. When all that is missing, you cannot repeat the check yourself.
It also leaves out the limits of its own numbers. Google's report counts impressions, which are links shown to a user, and its pages do not mention clicks or position. Bing says its citation counts do not indicate ranking, authority or placement in the answer. Any report that turns either figure into a traffic number or a rank is adding something the platform did not say.
Be careful with a score labeled as a Google AI rank. Google says no third-party tool has access to its internal ranking or AI systems, and it advises caution with any tool that claims to use internal Google metrics. A report built on a score like that is a guess with a number on it. The honest Google figures are the ones in your own Search Console report.
The test is simple. Hand the report to a colleague who has never heard of your business. If they can say which questions were checked, where, on what date, and which page was shown, the report is full. If they ask where a number came from, it is thin.
What should you see before you accept a report?
- The market and the dates for each question checked
- The exact wording of every question
- The platform behind every finding
- Whether your business was named, and in what position
- The pages shown as sources, with their addresses
- The same questions asked for local rivals
- A change line that compares with the last report
- A plain note of any gap, such as a platform that showed no sources
- The settings checked, including the Search Console generative AI control
Frequently Asked Questions
Is Google's Search Console AI report a full AI visibility report?
No. It is one part. Google's Generative AI performance report counts impressions for AI Overviews and AI Mode, with pages, countries, dates and devices. It does not cover ChatGPT, Gemini or Perplexity, and it does not list the questions asked. A full report adds those.
What does Bing's AI Performance report show?
It shows Total Citations, Average Cited Pages per day, a sample of grounding queries, page-level citation activity and trends over time. It covers Microsoft Copilot, AI summaries in Bing and select partner integrations. It was announced in public preview on February 10, 2026, and you open it inside Bing Webmaster Tools.
Why would a report list pages that never ranked for my question?
Google says AI Overviews and AI Mode may use query fan-out. They issue several related searches across subtopics and data sources, then show a wider set of supporting links. So a cited page can answer a related question, not the exact words typed. Read those pages as clues to what AI treats as relevant.
Should a report check my Google Business Profile?
Yes, as one line. Google says Business Profiles can help services be visible in AI responses and other Search results. Google also says indexing and serving are not certain. A report should note whether your profile is claimed, current and matches the name and city used on your site.
How much access should I give someone preparing my report?
Give read access to Search Console. Google's hiring advice says to grant only read access during an audit, not write access. Google also says to ask how the provider measures success. A report writer needs to see your figures. They do not need to change your settings.
Sources
- Google Search Central: AI features and your websiteGoogle
- Search Console Help: Generative AI performance report (Search)Google
- Search Console Help: Search generative AI controlGoogle
- Bing Webmaster Blog: Introducing AI Performance in Bing Webmaster Tools Public PreviewMicrosoft
- Gemini Apps Help: View related sources from Gemini AppsGoogle
- Gemini Apps Help: Get personalization with memory of your past Gemini chatsGoogle
- Perplexity Help Center: How does Perplexity work?Perplexity
- OpenAI Help Center: Searching the web with ChatGPTOpenAI
- Google Search Help: Understand and manage your location when you search on GoogleGoogle
- Google Search Help: Why your Google Search results differ from othersGoogle
- Google Search Central: Optimizing for generative AI features on Google SearchGoogle
- Google Search Central: Do you need an SEO?Google
- The Answer Engine founder case studyThe Answer Engine
Next step
Make your next report useful
We ran this on our own business, and now we want to learn what it means for yours. Bring your market and see where your business stands.
Get your free Answer Engine audit