AI visibility for SEO agencies
A client asks why they rank first in search and appear nowhere in a ChatGPT answer. That question needs a measurement, not an opinion: who is named instead of them, for which questions and on which model. This page explains what the measurement gives an agency, what to show a client and where its limits are.
What an agency gets that search tools do not provide
A search tool tells you where a page sits in a list of links. A language model returns no such list; it returns an answer in which companies are named or absent. The metric is therefore different: is the client mentioned, at what place in the answer, in what share of the questions and on which of the three models.
The most useful part of the report is usually the list of companies named instead of the client. That is not guesswork about competitors but a record of the names the models actually produce for the buyer questions. It often differs from the list the client has in mind.
What the measurement adds to agency work
- A starting number when taking on a client: a score from 0 to 100 made of forty technical points and sixty points for presence in the answers.
- A technical part with concrete fixes: access for six AI crawlers, content without script execution, structured data, sitemap and llms.txt.
- The delta against the previous measurement, which is evidence of work done rather than a promise. Every run after the first shows the change per individual question.
- Questions in the language of the client market. The query language is derived from the text of their home page, with English always included; two languages at most.
Where the limits are, before you find them at a client
The free analysis is meant for a single domain, not for sweeping a whole client portfolio. One e-mail address may submit five distinct domains within thirty days, and one IP address three domains per day. An agency wanting to measure twenty clients for free cannot, and that is deliberate.
There is no white-label report carrying your logo and no shared dashboard covering several clients at once. Each domain is its own measurement and its own subscription. If you need either, that is information for before the decision rather than a surprise after it.
We do not offer daily tracking. Monitoring runs every fourteen days or weekly, because each run is a full paid analysis and because one day of difference says less than the variation between repetitions of the same question.
How to build the measurement into an offer
The most direct use is the first conversation. The free measurement asks eight questions on three models and runs a technical check; the result is concrete enough to show whether there is anything to discuss. Where the result is poor, an engagement is easier to justify than with a description of a trend.
For an ongoing client, recurring monitoring is more useful, because the work shows up in the delta. Technical fixes appear in the score quickly, presence in the answers more slowly, because it depends on sources outside the client site. Say that in advance, or the first report will read as a failure.
Frequently asked questions
Is there a white-label report with the agency logo?
No. The report carries our layout and a statement of when and how the measurement was made. You may of course use the data in your own document, but there is no automatic export in your own branding.
How many clients can I measure for free?
Five distinct domains per e-mail address within thirty days, and at most three domains per IP address per day. The free analysis is meant as a first look at one domain; a client portfolio is served by the paid report or a per-domain subscription.
Does it work for clients outside Slovenia?
Yes. The query language is derived from the text of the analysed domain, not from your own language, so a German, Croatian, Italian or Polish client gets questions in their language and in English.
What can I promise a client on the basis of a measurement?
You can promise the fixes the measurement identifies and a repeat measurement that verifies them. Nobody can promise a place in an answer: the answer is composed by the model, not by a tool vendor, and it can change without any change on the site.
Measure a client domain
Eight questions on three models, a technical check and a score from 0 to 100. No card required.
Check AI visibility