AI visibility report and GEO metrics
An honest AI visibility report gives you a rate, not a verdict. In our measurements a company in the pool still appeared in only 11% to 53% of answers.
One check tells you what one reading said. We measured how often a company that can be named actually is named, and the answer depends on the question: between roughly 1 answer in 9 and 1 in 2.
11–53%
how often a company in the pool appears in any single answer, depending on the question asked
What we measured
- In our measurements a company inside the pool appeared in between 11% and 53% of answers, depending on which question was asked.
- 9 domains are cited across the 2 questions on this page and not one of them appears twice - nobody owns this answer yet.
- The pool behind one question ran from about 8 to about 53 companies, and 4 of the 5 pools were still growing at the fourth reading.
- 4 of the 5 question pools we measured were still growing at the fourth reading, so a single check samples a set that has not settled.
What a single reading can and cannot say
A report that says “you are not visible in AI” after one check is making a claim its method cannot support. If the pool for that question holds 28 companies and the answer lists 5, a company squarely inside the pool is missing from roughly 4 answers in 5.
What a single reading can say is concrete and useful: here is the exact question, here is the raw answer, here are the companies named, here are the domains cited. All of that is checkable.
What it cannot say is whether you are in or out. That takes repetition, and the honest output is a rate.
What we put in a report
The questions your customers would actually put to an assistant, chosen for your category rather than for our convenience. The raw answers, verbatim, whether or not you appear in them. The companies named and the sources cited, so you can see who currently owns the answer.
The surface and the model tier for every reading, because those are the variables that quietly change results. And where a number is uncertain, the uncertainty, printed rather than smoothed.
One thing we do not include is a single score out of a hundred. We have not found a way to compute one that survives being re-run.
Which metrics survive being re-run
The question most people actually type here is not “give me a report”, it is “what should I be measuring”. Our answer comes from having re-run the same readings and watched which numbers held.
Appearance rate holds. How often you appear across many readings is stable enough to compare month to month, and it degrades gracefully - with few readings it simply has wide error bars.
Pool size holds, as a floor. It tells you how many companies could be named at all, which is the denominator every other number needs. Ours ran from about 8 to about 53 by question.
The cited set holds. Who is quoted in the answer changes slowly compared with who is named, and it is the most actionable thing in a report.
Presence in a single reading does not hold. Two readings of one identical question 10 minutes apart shared nothing in our sharpest pair. Any metric built on one check inherits that.
A composite score out of 100 we have not been able to build so that it survives being recomputed, so we do not publish one.
Proof that a small player can be in the answer
On one of the 2 questions this page answers - “ai visibility report” - a site with almost no off-page footprint is already cited. It is the reason we treat entry as realistic rather than aspirational, and the reason we publish the negative results that make the route narrower but honest.
We do not attach off-page estimates to any named company. Those figures are a third party's approximation, not our measurement, so where we publish them at all they are anonymised - the full table is on our page about off-page authority.
The questions this page answers
geo metrics
2,900/moai visibility report
70/mo
2 questions, 2,970 searches a month between them. All 2 answers we examined on them were produced by a web search, which means a page can reach every one.
Who an assistant cites on these questions today
- discoveredlabs.com 1
- linkedin.com 1
- similarweb.com 1
- walkersands.com 1
- ayzeo.com 1
- clappingdogmedia.com 1
- loamly.ai 1
- semrush.com 1
- seoreviewtools.com 1
9 distinct domains are cited across these questions; the 9 above are the most frequent, with the number of answers each appears in. Counted from the answers captured in our corpus, not from a market census.
What to ask before you buy
- Ask whether a report is one reading or many, and insist on seeing the raw answers behind it.
- Treat any single-number visibility score with suspicion unless the vendor will show you the same score recomputed a week later.
- Ask what the report says when the answer is “we cannot tell yet”. Ours says that.
Honest limits of what we measured
- Appearance rates come from capture-recapture on 3 to 4 readings per question in one trade, on one assistant. They are the right order of magnitude, not precise probabilities.
- The estimator assumes every company is equally likely to be caught, which is false, so the rates are approximate and the pools are floors.
- A report measures the assistant surface. It says nothing about classic Google rankings, which we do not measure.
Questions we get asked
What does the report cost?
It is priced per category, because the work differs by how competitive the answers are. The price for each category is shown on that category's own page.
Is a one-off report worth buying at all?
As a first reading, yes: it gives you the questions, the raw answers and the incumbents, which is more than most businesses have. As a standing claim about your visibility, no - that needs repeated readings, and we would rather say so than sell you the wrong thing.
Can we get it white-labelled?
Yes, for agencies running this across a book of clients. Terms are a conversation rather than a price on a page.
What happens after the report?
You can act on it yourself - the raw answers are yours - or we can do the work and re-read. Either way the next reading is the test, not our opinion of the first one.
Start with one free reading
One business, real questions, the raw answers back - before you decide on anything paid. If you are running this for a book of clients, say so and we will talk terms rather than quote a page price.
Free reading
Takes about a minute.
Related
Pricing is per category, because the work differs by how competitive the answers are. See every category we measure and what a reading costs →
Published 31 August 2026. Every figure on this page is recomputed from raw observation files by a single script, so any of them can be traced back to the readings behind it. We make your site citable. We do not promise placement in any assistant’s answer - we measure whether you are in it, and show you the raw answers either way.