EN
EnglishFrançais
Case study · Real reading, anonymized

Named on half its buyers' questions. Absent where they compare costs.

This is a real citation test, run on the client of an SEO agency we had just written to, to show the agency what it would receive. The client and the agency stay confidential. It is the starting point of a Sprint, not its result: no Sprint was ordered on this client.

Client: employee-benefits platform Market: France Measured: 31 August 2026 Engines: Perplexity, ChatGPT, Claude
The result
55%
Named in a little over half of the answers.

The rate comes from Perplexity, with its 95% confidence interval: 30 to 75%. Two figures inside that range are the same result, which is why the rate is never published without it. ChatGPT and Claude, asked a smaller set of the questions, are read as signals, not as rates.

Where it is named, and where it is not

✓
Named on choice questions
Which platform for a company under 50 employees, a turnkey solution for 20 employees, the price of an outsourced platform. The questions a buyer asks when close to deciding.
!
Absent on cost per employee and on tax
Asked about the price of a platform, the client is named. Asked about the cost per employee per month, it is not. Two questions a buyer sees as the same give opposite results.
!
Absent on comparison and on proof
When the buyer asks how to compare the solutions, or which companies have adopted one, the answers rely mostly on direct competitors' sites.
!
Read, not kept
On one question, a page of the client's own site is among the pages the engine opened, and the client's name is still missing from the answer.
✕
Out of reach, and said so
Part of the absences cannot be moved by any outreach work. The reading names them, and nothing is sold on them.

What the agency can sell its client

The client is not invisible: it is named about one time in two. So "you are missing from AI" would be false, and an agency that pitched it would lose the client's trust the day the client checked. What can be sold is a defined scope: the cost, comparison, and proof questions, where a competitor answers in the client's place.

That is what a Citation Sprint works on: three off-page placements pursued against those gaps, then a re-test at 30 days on the same questions, with the same margin of error. The re-test is a comparison, not a promised improvement.

What this case does not show

The same reading, on one of your clients.

Send us one client domain. We run the citation test and email you what came back, competitors included. No charge, no account, no call. Your client is never contacted.

Send us one client domain →