Find out what AI says about your brand — and how much of that answer you can trust.
One link in. 45 answers out of 3 models. You get a range and a margin of error, not a single number — because one chatbot reply is noise, and we show you the arithmetic.
No sign-up, no card. The result opens on screen in about two minutes.
Every AI visibility tool gives you a number. Almost none of them tell you the number is mostly noise.
LLM output is non-deterministic: the same question, asked twice, gives different answers. Measuring it once is not a measurement, it is a sample of one.
It gets worse. The same brand, on the same data, reads 20% by share of mentions, 16.8% by weighted position and 31.4% by citations. The figure you are shown is decided by the method, not by reality — and the method is usually not published.
So we do not report a figure. We report a range, the sample it came from, and the smallest change that would mean anything.
Sample report — illustrative values
Mentioned in 31% of answers · 95% interval 18–44% · measured over 45 answers
At this sample size, only a change larger than 19 pp means anything. Anything smaller is noise.
The free check, in full, before you give us anything
3 models × 5 prompts × 3 repeats = 45 answers. Then, on screen:
- How often you are mentioned, with a 95% interval
- The top 3 brands named instead of you
- Up to 3 places where a model states something your own site contradicts
- One recommendation, tied to the specific answer that caused it
- A link to the method, in full
We generate the prompts from your site. You are not billed per prompt, and you do not have to guess which ones to track.
What 45 answers cannot do: track a change over time. The interval is ±13 pp, so only a swing above 19 pp would be readable. That limit is arithmetic, and it is the honest reason the paid tier exists.
«The AI is telling people something about us that isn't true»
For most owners this hurts more than placing fourth on some share-of-voice chart. It is also the one thing almost nobody sells you: exactly one competitor detects false claims about a brand, and only on an enterprise plan.
We read the facts off your own site, compare them with what the models assert, and list every disagreement in the plain form: your site says X, the model says Y, in this answer.
Up to three of them are in the free check. Not behind the payment.
A weekly delta that says «significant» or «indistinguishable from noise» — and means it
Monitoring starts at $29 across this market. A significance threshold is offered by nobody.
On the paid tier we run 4 models × 15 prompts × 5 repeats = 300 answers. That puts the interval at ±5 pp and the significance threshold at 7 pp, at 95% confidence.
Every week you get one line per metric: the change, and whether the change is real. When it is not real, we say so — including the weeks when the answer is «nothing measurable happened».
How we get to the number, published in full
How the sample is built, why that many runs, how the interval is computed, what counts as a significant change, and what we do not measure at all.
We query models over their APIs. That is not the same thing as the AI Overviews block or what you see in the ChatGPT web app, and we do not claim it is.
Read the method before you trust any of the numbers above. That is the point of publishing it.
Read the method →Things we deliberately don't sell you
- We don't promise to move you up in AI answers. Nobody can prove that link, and everybody promises it anyway.
- We don't measure AI Overviews or the assistants' web interfaces. We measure model answers over the API, and we say so in the method.
- We don't compete on volume. Two competitors run orders of magnitude more prompts than we ever will. We measure less, and we tell you the error bar.
- We don't write your content for you.
- We're not another suite. One tool, one job, one price.
One plan. $39 a month.
subscription, one plan
Free, no sign-up: a 45-answer check of one site, the interval, the top 3 competitors, up to 3 factual disagreements, one recommendation, the method.
$39/month: the 300-answer audit — interval ±5 pp, significance threshold 7 pp — every factual disagreement with a recommendation for each, the full competitor picture with intervals, and the weekly delta report.
What you're paying for is sample size and a threshold, not a longer feature list. 45 answers show you the picture. 300 answers let you tell a change from noise.
Monthly, cancel anytime. Cancelling keeps your existing reports readable at their permanent links.
Questions we actually get
No. The result opens on screen. You can have a copy emailed to you afterwards if you want a permanent link.
We generate them from your site — brand, products, region, competitors. You aren't billed per prompt and you can't accidentally track the wrong five.
No, and be careful with anyone who sells you that. The AI crawlers — GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot, Google-Extended — overwhelmingly skip it and read your HTML instead. What does work: extractable structure — definitions, comparisons, FAQs, lists — plus statistics and citations in the first 200 words, and mentions on third-party sources models already trust.
It isn't smaller, it's honest. Those tools give you one number with no error bar. We'd rather give you a range you can act on.
Once per domain per week. Ask again sooner and you get the report we already ran, not a new one.
Yes — in 7 days. Just know that two 45-answer checks are only comparable if the gap between them is bigger than 19 pp. We'll say that in the email too.
Find out what the models are saying about you. It takes one link and about two minutes.
Run the free checkNo sign-up, no card.