AI visibility
Find out whether the assistants recommend you
Four assistants and Google's AI Overview, asked about your market: the answer, the sources cited, and who was named instead of you.
- surfaces asked - four assistants and Google's AI Overview
- 5surfaces asked - four assistants and Google's AI Overview
- Perplexity, the cheapest and the widest on sources
- WeeklyPerplexity, the cheapest and the widest on sources
- ChatGPT, Gemini and Claude, at 4-8x the price each
- MonthlyChatGPT, Gemini and Claude, at 4-8x the price each
- every answer stored as written, not as a score
- Full textevery answer stored as written, not as a score
Five surfaces, one screen
Each assistant is asked about your market with web search enabled, and the answer is stored whole - not a score derived from it. You read what a customer would have read.
Google’s AI Overview is watched separately, on the keywords you already track, so the answer above the results is measured on the same terms as the results.
What comes back
Four things, and none of them is a number you have to take on trust.
The answer, kept
Every response stored as written, with its citations. A number tells you something changed; the text tells you what.
Cited, or not
Whether your domain appears in the sources on each surface, and where in the list it sat.
Who was named instead
The domains cited for your subject, ranked by how many of the five surfaces mention each one.
What it cost
Every answer records its own vendor cost, so you can see the price of the question you asked.
A week-by-week record
The same question asked again on a cadence, so a change in who gets cited is visible rather than remembered.
Your competitors, side by side
The same citation counts for the competitors you track, on the comparison page.
Why the count is surfaces, not citations
On one real run Perplexity returned twenty citations and Claude returned one. Ranked by citation count, that list measures how talkative each vendor is rather than who the market considers an alternative.
So a domain is scored by how many of the five surfaces name it at all - and the panel shows which, because “ChatGPT and Perplexity but not Gemini” is something you can act on and “two of five” is not.
And the visitors who actually arrive
Whether you were quoted inside an AI answer is not available to any tool. Search Console’s API has no AI dimension at all - we probed both of our own properties for every plausible value and each was rejected, while known-good values returned normally in the same run.
Arrivals are measurable. With Analytics connected, the report shows the people who came from an assistant and whether they stayed - and they stay far better than anybody else. On one connected site, 87% engaged against 15% for every other channel.
Measured on one connected site over 28 days. The channel was 0.3% of its traffic - which is why the rate is the number worth printing and the count is not.
The words for this
Three names for the same shift, used by different people. Here is what each one means and how this panel answers it.
LLM SEO
LLM SEO is optimisation for answers rather than for links: the work of getting a large language model to name your site when somebody asks it a question you would have wanted to rank for. Most of the discipline is unchanged - a page the model can read, a claim it can check, a source it is willing to repeat.
What is different is the feedback loop. A ranking is public; a citation is not. Search Console has no dimension for it, and we probed both of our own properties for every plausible value before writing that sentence - each was rejected while known-good values returned normally in the same run.
So it is measured here directly instead. Four assistants are asked about your market with web search on, each answer is stored as written, and its sources are matched against your domain: whether you were named, by which of them, and which domains were named in your place.
AI search optimization
AI search optimization is the wider term, and it covers the answer above the ten blue links as well as the assistants beside them. Google’s AI Overview is the surface most sites meet first: it appears on a share of ordinary queries, it answers them in place, and the sites it cites are the ones that get the visit.
It is watched here on the keywords you already track, at most once a week per keyword, on the same nightly run as your rankings - so the answer above the results is measured on the same terms as the results, and the two can be read against each other.
The optimisation half is not a lever any tool can pull for you. What this one does is close the loop: which of your tracked keywords carry an AI answer at all, which of those cite you, and what changed since last week.
Generative engine optimization
Generative engine optimization - GEO - is the same idea under the name the research literature gave it, and it is worth knowing because it is the phrase most industry writing now uses. A generative engine is any system that answers in prose and cites as it goes: ChatGPT, Claude, Gemini, Perplexity, Google’s AI Overview.
The practical difference from ranking is that there is no position to hold. You are named or you are not, the set of sources changes without notice, and the only honest measurement is repeated observation rather than a score derived from one.
Which is what the record here is for. The same question is asked on a cadence - weekly for Perplexity, monthly for the three that cost four to eight times as much - and every answer is kept whole with its citations. Nobody can promise you a citation, us included. What you get is the evidence.
What it does not do
It does not promise to get you cited. Nobody can - the assistants pick their own sources and change them without notice.
What it does is tell you, week by week, whether they picked you and who they picked instead, from answers you can read in full rather than a number you have to trust.
Questions
- Perplexity, ChatGPT, Gemini and Claude, each with web search on, plus Google's AI Overview on the keywords you track.
- The AI Overview is observed on the keywords you already track, at most once a week per keyword, on the same nightly run as your rankings. The assistants are asked when you press AI Analyze, because that call costs real money and a scheduled sweep would spend it answering a question nobody asked that day.
- The full answer each assistant gave, every source it cited, whether your site was among them, and a list of the domains that were cited instead - with how many of the five surfaces named each one.
- No. Rank tracking measures the ten blue links. This measures the answer above them and the four assistants that never showed links at all. A site can hold position three and be absent from every AI answer about its own category.
Start free
One site and 100 credits - up to 200 pages a scan, about 5 scans, and the whole report. No card. Paid plans add more sites, bigger crawls and the AI rewrites.