Key Takeaways
- Read the question list before the score. A number tells you how often you appeared. The list of which questions worked and which did not is the only part that tells you what to do next.
- Treat the rival column as the real finding. Who got named in your place is a description of what a model could find and repeat. It is a reading list, not a verdict on anybody being better.
- Never act on a single run. Assistants return different answers to the same question on different days. One month of movement is noise until a second month agrees with it.
- Click through to the answer the number came from. If a report will not show you the text behind a figure, you are being asked to take it on trust. Read the answer and check that the number describes it.
- Turn the report into exactly one page. The useful output of any check is a single question you now intend to win, and the page you will write to win it. Everything else is context.
An AI visibility report arrives as a document. There is a number near the top, a list of names, and several pages of detail underneath. Most people read the number, feel something about it, and never open the rest.
That is backwards. The number is the least useful thing in the file, and the part that tells you what to do on Monday is three pages down.
This post covers what each part of a report actually supports, the five misreadings that cost the most, and how to turn any of it into one page worth writing.
How do you read an AI visibility report without a dashboard?
In three passes, in this order: which questions named you, who was named instead, and the raw text behind one figure you find hard to believe.
The first pass tells you what already works. The second tells you what a model could find and repeat about your market. The third tells you whether the report is honest. None of the three needs a login, a chart, or a tool that updates while you watch it, and the whole thing takes about twenty minutes.
Everything after that is prioritization, which is a judgment call rather than a measurement.
The headline number, and what it does not mean
It is a count of appearances across a fixed question set on one day. It is not a percentage of your market, a ranking, or a grade.
Say a report opens with six of twenty-five, an illustrative figure. That means six questions produced an answer containing your business and nineteen did not. It does not mean you are 24 percent visible, because the twenty-five questions are a sample somebody chose, not the whole of what buyers ask. Change the sample and the number changes without anything about you changing.
The count is still worth having, for one reason: it is comparable to itself. The same questions, asked again next month, tell you the direction of travel. That is the only claim a number like this can carry. If you have never run the questions at all, the mechanics of doing it by hand are in how to check if AI recommends you.
The rival list is the real finding
Who got named instead of you is a description of what an assistant was able to find and repeat. Read it as a reading list rather than as a scoreboard.
Go through the names one at a time and ask a single question about each: what does a model know about this business that it does not know about mine. The answer is nearly always mundane. A stated price. A page about one specific situation. A profile on a site that gets read. A written client outcome with a number in it.
Google's guidance on AI features in Search is blunt about the underlying mechanism: to be shown as a supporting link, a page has to be indexed and eligible to be shown with a snippet, and there are no additional requirements. There is no separate contest being run. A business that turns up in these answers is a business whose facts are written down somewhere legible.
The same is true one layer down. Google's documentation on featured snippets answers the question people always ask next, which is whether there is a marking that gets a page picked, with a flat no: their systems decide whether a page would make a good extract. So a rival appearing nineteen times is not evidence of a technique you have not bought. It is evidence of material a machine could use.
The five misreadings that cost the most
Each one leads to a different and more expensive next action than the correct reading does. That is why they are worth naming.
The middle column is what people conclude. The right column is what the line actually supports. Illustrative report lines, not measured data.
Taking them in order. "Named in 6 of 25" does not mean you are 24 percent visible; it means six specific questions worked and you should read which six. "A competitor appears in 19 of 25" does not mean they beat you on quality; it means they are described in more places a model reads. "A directory outranks every business" does not mean the test is broken; it means the question was too broad to name anyone. "Your score fell since last month" does not mean something you did stopped working; answers move on their own, so read the trend rather than the step. And "named, but described wrongly" is not good enough, because a wrong description costs more than absence does.
| Part of the report | What it supports | What it cannot support | The action it should produce |
|---|---|---|---|
| The headline count | Comparison with the same set next month | Any claim about market share | Write the date down and move on |
| The list of questions asked | Judging whether these are your buyers questions | Proof that the sample is complete | Add the questions you know are missing |
| The questions you were named in | Knowing what already works | Knowing why it worked | Protect those pages, do not rewrite them |
| Who was named instead | Seeing what a model could find | Any judgment about who is better | Read their pages and note what is written down |
| How you were described | Spotting an out-of-date or wrong picture | Telling you where the wrong fact came from | Fix the page carrying the stale fact |
| The recommended fix | A starting point someone has reasoned about | A guarantee of a result | Do the first one only, then re-check |
The third column is the one worth rereading. A report that claims more than those middle entries allow is overselling, and the honest test of any vendor is whether their document admits its own limits. If it will not show you the answer text a figure came from, treat the figure as an assertion.
What to do in the first week
One page, one re-check, nothing else. The most common failure after a report is doing six things at once and learning nothing from any of them.
Pick the single question you most want to win, from the list you were not named in. Write the page that answers it: the question as the heading, the answer in the first two sentences, a price or a range, and what happens in the first month. Then wait a few weeks and ask the same question again. If you changed six things you will never know which one moved it. Choosing which single thing to do first is the harder half, and it is covered in what to do with your AI test results.
Keeping the record is a spreadsheet job: date, question, who was named. That is enough to see a trend across three months, and it costs nothing, which is covered in tracking AI visibility without buying software.
If you do not have a report to read yet, the free check produces one. It asks ChatGPT 25 real buyer questions from your niche at $0 and emails a plain-English readout: your score, who got named in your place, and the one fix to start with. It runs as soon as you ask for it, and the report lands within the hour between 7am and 5pm Central, or by 10am the next morning outside those hours.
Questions we hear the most
How do you read an AI visibility report without a dashboard?
Read it in three passes: the list of questions you were named in, the list of who was named instead, and the answer text behind one figure you doubt. The headline score is a summary of those three things, not a substitute for them.
What does a visibility score actually measure?
How often your business was named across a fixed set of buyer questions on a given day. It is a count, not a rating, and it says nothing about the quality of your work or your ranking on any search engine.
Is a low score bad news?
It is ordinary news for most coaches and consultants, and it is more useful than a high one. A low count with a clear list of which questions failed gives you a specific page to write, which a high count does not.
Why did my score change when I changed nothing?
Because assistants do not answer the same question identically every time. Some movement is normal, so a single step up or down means little until a second measurement in the same direction confirms it.
Should I worry more about being absent or being described wrongly?
Being described wrongly, usually. Absence costs you a mention, while a wrong description actively tells a buyer something untrue about your prices, your location or what you do, and it gets repeated.
Do I need software to keep track of this?
No. A spreadsheet with the date, the question and who got named holds everything a monthly re-check needs. Tools save time at scale, and they do not tell you anything a written record does not.
What does the AnswerHalo free check report contain?
It asks ChatGPT 25 real buyer questions from your niche and emails a plain-English report: your score, who got named in your place, and the one fix to start with. It costs $0 and it runs as soon as you ask for it.
Can I see the answers behind the numbers in a paid audit?
Yes. Every number in the $500 audit links to the answer it came from, so you can read the text yourself rather than take the count on trust. The audit runs 125 checks across five AI tools.
Your buyers are already asking. Find out what AI tells them.
The free check asks ChatGPT 25 real buyer questions about your niche and sends you the report within the hour, at $0. If it shows you are already getting named everywhere, we will say so in plain words.
Get my free check →