TL;DR: An AI visibility report shows which buyer questions AI engines answer, who they name, and which sources they cite. Read the role you were given, not just the mention count. Almost every gap is one of four types, and each has a different fix.
An AI visibility report answers one question: when your buyers ask an AI assistant who to hire or what to buy, does your name come up? This guide explains every part of a report in plain English, then gives you the fix for each kind of gap it finds. If you have not run one yet, the free AI visibility check is the quickest way to get one.
Key takeaways
- A report tests a fixed set of buyer questions across several AI engines and records who gets named, how prominently, and which sources the answer used.
- Being mentioned is not the same as being recommended. Read the role you were given.
- Almost every gap is one of four types: AI cannot read your site, no page answers the question, your page is weaker than the winner, or AI relies on third-party sources where you are absent.
- One check is a snapshot. AI answers vary between runs, so a trend beats a single result.
- The report is the diagnosis. The value comes from doing the work it points to, then checking again.
What does an AI visibility report measure?
It measures how AI assistants answer the questions your buyers actually ask, and whether your business gets named. The report runs each buying question through several AI engines and records which businesses were named, in what order and in what role, and which websites the answer drew its information from.
A good report also connects that to your own search data. Once Google Search Console and Google Analytics are connected, it can show which pages earn clicks and impressions, which queries bring people in, how your rankings have moved, and how many sessions and enquiries followed. AI answers are usually built from pages search engines can already find and trust, so the AI side shows where you are missing and the search side shows what you have to work with.
The questions are the part most people skim, and the most important part. A report built on questions no buyer asks will look tidy and mean nothing.
What do the terms in the report mean?
Reports use the same handful of words and they are easy to confuse. Here is what each one means.
Mention and recommendation
A mention means your name appeared in the answer. A recommendation means the answer put you forward as a choice for the buyer. Reports usually split this into roles: recommended, mentioned in passing, compared against someone else, or cautioned against. A mention in a list of ten is not a win. A recommendation in an answer that names three businesses usually is.
Citation and cited source
A citation is when the answer links to or names a web page as the source of what it just said. The cited sources are those pages. Some will be yours. Many will not: directories, review sites, trade publications and competitor pages all show up. In my view this list is the most useful part of a report, because it shows where AI gets its picture of your market.
Share of voice
Share of voice is the proportion of tested answers in which you were named, compared with everyone else named. If you appear in 12 of 50 answers and your nearest competitor appears in 30, the gap is clear without further maths. Good reports split this by engine and by topic, because you can be strong on one topic and invisible on another.
Position in the answer
Position is where you appear inside the answer: first, in the middle, or last. Buyers read the top. Appearing fourth in a list of five is closer to invisible than a percentage makes it look.
What is an AI visibility score?
An AI visibility score is a single summary number showing how often and how prominently AI engines name your business across a fixed set of buyer questions. It usually blends how many answers you appear in, the role you are given, and where you sit in the answer, then expresses the result as a percentage or an index.
Treat it as a thermometer, not a diagnosis. There is no industry standard formula, so a score from one tool cannot be compared with a score from another. It is good for tracking your own direction over time on the same question set and engines. A score that moves from 14 to 22 over a quarter means something. A score of 22 on its own means very little.
How do you read the report, section by section?
Work through it in this order. It takes about twenty minutes and it stops you reacting to the first alarming number you see.
- Check the questions first. Cross off any a real buyer would never type. If most are wrong, nothing else in the report is reliable.
- Look at share of voice per engine. Note where you are weakest. Strong in one engine and absent in another is common, and points at different fixes.
- Read the roles, not the counts. Separate the answers where you were recommended from the ones where you were mentioned in passing or compared unfavourably.
- Read the cited sources. List the non-competitor sites that keep appearing: directories, review platforms, trade publications, association lists. These are the places AI trusts in your market.
- Line up the losses. For every question where a competitor was recommended instead of you, note which competitor and which of their pages was cited. Here is how to track which competitors are recommended.
- Cross-check with Search Console. For each losing question, see whether you have a page targeting it, and whether it gets impressions or clicks.
- Sort by buyer value. Put the questions closest to a purchase decision at the top, whatever the size of the gap.
That final list is your to-do list. Everything above it is preparation.
Which gap is it, and what fixes it?
Nearly every line on that list is one of four gap types, each with a different fix. Getting the type right matters. Writing a new page will not help if the real problem is that AI cannot read the pages you already have.
Gap 1: AI cannot read your site
Symptom: you have good pages on the topic, they rank reasonably in Google, and AI still never names you. The cause is usually access. AI crawlers blocked in your robots rules, pages that only render after JavaScript runs, missing or broken structured data, or pages that are not indexed.
The fix is technical, and normally the fastest win available, because the content already exists. Allow the AI crawlers you want reading your site, put the important content in the served HTML, add correct structured data, and confirm the pages are indexed.
Gap 2: no page of yours answers the question
Symptom: the question is clearly a buying question, AI answers it confidently, and you have nothing that addresses it. Search Console shows no impressions for anything close.
The fix is to write the page. One question, one page, with a direct answer in the opening lines rather than paragraphs of preamble. AI engines lift short, self-contained answers, so put the answer first and the detail underneath.
Gap 3: your page exists but a competitor's is better
Symptom: you have a page, it gets some impressions, and the answer cites a competitor's page instead.
The fix is to study the page that won and close the difference. Note what it covers that yours does not: pricing ranges, comparison tables, specifics about a location or a sector, named process steps, real examples, an FAQ. Then improve your page until it answers the question more completely than theirs, and make the answer easy to extract.
Gap 4: AI relies on third-party sources where you are absent
Symptom: the cited sources for a question are not company websites at all. They are directories, review sites, roundups, trade publications or association pages. You are not on any of them.
The fix happens off your own site. Get listed where AI keeps looking, claim the profiles that already exist, collect reviews on the platforms that matter in your sector, and get covered in the trade publications your buyers read. This is the slowest of the four, and often the one competitors have quietly been doing for years. Our guide on how to get your business recommended by AI covers it in more detail.
An illustrative gap list. The table below uses a made-up regional accounting firm and invented results to show the shape of a finished gap list. It is illustrative only, not real data from any business.
| Buyer question | Engine | Who was recommended | Cited source | Gap type | Fix |
|---|---|---|---|---|---|
| Best accountants for small businesses near me | ChatGPT | Three local firms, not us | A local business directory | 4: third party | Claim and complete the directory listing, gather reviews |
| How much does a small business accountant cost per month | Google AI Overviews | A national firm | A competitor's pricing page | 2: no page | Write a clear pricing guide page |
| Do I need an accountant or can I do my own bookkeeping | Perplexity | Two firms, we appear fourth | A competitor's guide | 3: weaker page | Rewrite our guide, add a comparison table and FAQ |
| Accountants for construction companies | Claude | Two specialist firms | Two trade publications | 4: third party | Pitch a trade publication, add a sector page |
| Year end accounts checklist | Google AI Mode | Nobody named, generic answer | Two generic blogs | 1: not readable | Fix crawler access and structured data on our checklist page |
What are the best ways to track brand mentions in AI search?
Run a fixed set of buyer questions across every engine your buyers use, on a repeating schedule, and record the full answer each time: who was named, in what role, in what position, and which sources were cited. Consistency is what makes it tracking rather than spot-checking.
Three habits separate useful tracking from noise. Keep the question set stable, because if you change the questions every month you cannot tell whether you improved or the test changed. Never put your own brand name in the question, because that produces a mention every time and measures nothing. And record the sources, not only the names, because the cited source list is what turns a tracker into a plan.
You can do this by hand with a spreadsheet and a calendar reminder. It works, and it is dull, which is why most businesses do it twice and stop. A tracking tool or an AI visibility team keeps it running.
How to win brand visibility in AI search
Winning visibility means being the source AI reaches for when your buyer's question comes up. It is earned page by page, in this order:
- Make your site readable. Crawler access, served HTML, correct structured data, indexed pages. Nothing else works until this does.
- Answer real buyer questions on their own pages. One question, one page, answer in the first few lines.
- Be more specific than the page currently winning. Numbers, ranges, sectors, locations, steps and worked examples get quoted. Generic advice does not.
- Get present in the third-party sources AI cites. Directories, review platforms, trade publications and association listings for your sector.
- Keep existing pages current. Refreshing a page that is sliding is usually cheaper than writing a new one.
- Measure again, and let the results choose the next round of work.
Worth being honest about the limits: none of this is fast, and nobody can promise you a place in an AI answer, because the engines rewrite their answers constantly. If you are weighing this against traditional search work, our comparison of AEO, SEO and GEO explains where the three overlap.
How often should you re-check, and why is one check only a snapshot?
Monthly is a sensible minimum, and weekly if your market moves quickly or you are actively publishing. AI answers are not fixed. Ask the same engine the same question twice and you can get two different sets of businesses named, in a different order, drawing on different sources. This is how much AI answers vary between runs.
That variability is normal, and it means a single check tells you roughly where you stand, not precisely. So do not panic about one bad result or celebrate one good one. Treat the trend across repeated checks on a stable question set as the number that matters, because random variation averages out and real change does not.
Re-checking also catches what goes wrong quietly: a page that slips out of an answer, a competitor publishing something better, a change that blocks a crawler, a new question your buyers have started asking.
What do people get wrong when reading these reports?
Four mistakes come up again and again, and each makes the report less useful than it looks.
Testing with your own brand name in the question
If the question contains your name, the answer will contain your name. That measures nothing about discovery, which is the point. Ask what a buyer who does not know you exist would ask.
Checking one engine only
Buyers are spread across ChatGPT, Google AI Overviews, Google AI Mode, Perplexity and Claude, and these engines do not agree with each other. Strong visibility in one is regularly paired with nothing at all in another.
Checking once
A one-off audit is a photograph of a moving thing. Useful for finding the obvious gaps, useless for telling you whether your work is paying off.
Chasing questions no buyer asks
It is satisfying to be recommended for a broad industry question with no buying intent behind it. It does not produce enquiries. Sort your gaps by how close the question is to a purchase, not by how big the gap is.
Questions people ask about AI visibility reports
How can I track if my competitors are getting mentioned in AI search results?
Run your buyer questions across the engines and record every business name in each answer, not only yours. Over a few rounds you will see which competitors are named most often, for which questions, on which engines, and which of their pages the answers cite. That cited page list shows what you are being measured against.
Is an AI visibility audit the same as an SEO audit?
No. An SEO audit looks at how your site performs in search results: rankings, technical health, content and links. An AI visibility audit looks at what AI assistants say when your buyers ask a question. They overlap, but a clean SEO audit paired with poor AI visibility is common.
Can I do this with a free AI visibility check?
A free check is a good starting point. It shows the questions, the engines, who is being recommended and where you are missing. What it cannot do is tell you whether you are improving, because that needs the same questions run repeatedly over time.
Why does my report show me in one engine and not another?
Each engine builds answers from a different mix of sources, refreshes them on a different schedule, and weighs reviews, directories and publisher content differently. A strong third-party listing can carry you in one engine and count for little in another.
Do I have to fix every gap in the report?
No, and you should not try. Sort by buyer value, take the technical fixes first because they are quick and help everything else, then work down the list. Most reports contain more work than you have capacity for.
How long before fixes show up in AI answers?
Technical fixes can appear within days once pages are re-crawled. New and improved pages need to be indexed first, and meaningful movement generally compounds over 2 to 3 months. Nobody can honestly promise a date, and you should be wary of anyone who does.
Where the Growth Engine picks up
A report gives you the diagnosis. The Visiby Growth Engine does the work it points to, then runs the check again and uses the result to choose what comes next.
It tracks the same buyer questions across ChatGPT, Google AI Overviews, Google AI Mode, Perplexity and Claude, works from your own Search Console and Analytics numbers, fixes the technical and AI-crawler issues that stop AI reading your site, and publishes 20 to 25 buyer-focused pages a month to your own site, plus refreshes of pages that start slipping. You approve the content. It starts at $1,000 per month, there is no contract, and everything created stays yours.
Find a gap. Close it. Check again.
Book 15 minutes and we will walk through what your report found and what the first month would cover. If you have not run a check yet, start with the free AI visibility check.
Last updated: 18 September 2026.
Raunaq Arora
Raunaq Arora is a senior software and AI engineer at Visiby, where he builds the AI-visibility measurement pipeline and dashboard. He writes about how AI-visibility tracking is measured and tooled. Run the free AI visibility check →