For agencies

Your clients ask where AI puts their brand.

Add a client's domain. We ask the questions their buyers ask, four assistants, three times each. You get a link you can send them with every answer in it.

Run a free audit

Want to see what clients get? See AI visibility

  • One domain per client
  • Client opens a link, no login
  • Free to start, no card

Where it breaks

Clients ask about AI. You answer with a screenshot.

The question comes up in every review now. Without a number per client you cannot prove the work.

  • “Does ChatGPT mention us?” comes up mid-meeting

    You take a screenshot. Next month you take another one. Neither proves anything moved.

  • Your rank tracker cannot see it

    Ranking third is no help when the answer names two brands and links to neither of them.

  • Asking once is not measuring

    Ask the same question twice and the brands change. Two runs match less than once in a hundred, so one check tells you almost nothing.

  • You cannot see whose competitors they are

    Every client has rivals you have not heard of. Guessing the list at setup means measuring the wrong fight.


What changes

Something you can show, and something to sell next

Visibility scored separately for each assistant, for one client's domain.

You answer it in the meeting

Named or not, on which assistant, against whom. One real run had a brand at 31% on Google AI Overviews and 0% on ChatGPT, the same day.

A tracked prompt set re-running on a schedule with its history.

The monthly line is real

The same questions run again every week and every answer is kept. So the deck shows a line, not a fresh screenshot each time.

Ranked moves, each naming the page to write or the source to earn.

Every gap names the next job

A page to edit, a thread to answer, a listicle to get onto. That is your next scope of work, with the answer that proves it.


How it runs

Set a client up in an afternoon

Four steps from a signed client to something you can send them.

  1. Add the client's domain

    One domain per client. Each keeps its own questions, its own rivals and its own history.

    A separate benchmark per client

    Step 01

  2. We find the questions

    From Reddit threads, People Also Ask, and the rivals already ranking. 49 get tested, 26 pass, 5 are kept. You can strike out any that are wrong for the client.

    Real buyer questions, each with a source you can check

    Step 02

  3. It runs every week

    Five questions, four assistants, three runs each. Sixty answers, all saved.

    An answer history you can point at

    Step 03

  4. Send the client the link

    One link, no login for them. It opens the numbers and the answers behind each one.

    A report the client can read without you in the room

    Step 04


What you get

The parts an agency uses


FAQ

Agency questions

What comes up before the first client goes in.

How many clients can I add? +

As many as you like. Each client is a domain with its own questions, rivals and history. There is no per-client limit and nothing is shared between them.

Can the client see it without a login? +

Yes. Each report is a link you can send. They open it and see the numbers with the full answers underneath. You can revoke the link at any time.

Can I put our branding on the report? +

Not yet. Today the report carries ours. White-labelling is in build and not shipped, and we would rather say that here than after you have pitched it.

Which assistants do you check? +

Google AI Overviews, Gemini, ChatGPT and Perplexity. Each question goes to all four, three times each, which is twelve answers per question.

Can I use this in a pitch before they sign? +

That is the best use of it. Run the free audit on a brand you want to win. If the assistants are not naming them, you are walking in with the problem and the first fix already found.

Do I have to know their competitors? +

No, and you should not guess. The rival names come out of the answers themselves, so you often find brands the client never mentioned and the answer treats as the alternative anyway.

How do I know a number really moved? +

Each question is asked three times, and one that gets named in a single run goes back out for nine more before it counts. Gaps under about 5 to 7 points are noise, and the report shows the sample size next to every figure so you can tell.


Take one client's number to the next meeting

Run the free audit on a brand you already look after. If there is a gap, you have your next scope of work.