Goppa Research

How to show a client their AI visibility

Producing the number is the easy half. You ask an assistant the buying questions in the client's category, count how often they are named, and you have a percentage. The hard half is walking into a room with it, because on current data that percentage is usually zero — and a zero presented badly ends the conversation instead of starting it.

First: you are probably in the same position

We measured 81 agencies that sell AI visibility, answer engine optimisation or generative engine optimisation, using the question their own prospective clients would type. 61 appear in none of the answers.

That is worth knowing before the meeting for two reasons. It means the client's zero is not evidence of anything unusual. And it means that if you are about to sell this service, someone may eventually run the same check on you — so the honest framing is also the durable one.

A zero alone is not a finding

"You appear in 0% of AI answers" is true and useless. It has no unit of comparison, no cause, and no next step, so the client is left with a bad feeling and a decision they cannot make.

Three things turn it into a finding.

1. The benchmark around them

The median appearance rate is zero in every category we measured, and 48 of 64 shops appear in none of the answers about their own category. Saying so first is not softening the news — it is the difference between "you are behind" and "this position is open", and only one of those is true.

It also pre-empts the question that kills the room, which is whether the number means anything at all.

2. The pages the assistant read instead

This is the part that makes it a plan. The assistant opened specific pages to build that answer; they are a short, finite list, and the client is on none of them.

Sort that list by how you get onto each page rather than by how often it appears. Across the categories we measured, 23 of 101 source citations are directories with no gatekeeper — you submit, you are in. Those are the ones to start with, and almost nobody does, because they are unglamorous.

The encyclopedia entries are the second most cited kind of page and the worst first step: they need independent coverage to exist, so they come last.

3. A date to measure again

A single measurement is a reading, not a verdict. Assistants vary by day and by phrasing, and a client who is told "you are at 0%" without being told that will treat the next number as proof you were wrong.

Say it upfront, set the repeat date in the same meeting, and the second measurement becomes the deliverable rather than a risk.

What not to do

• Do not promise a position. Nobody controls which brands an assistant names, and a guarantee you cannot keep is a refund with a delay.
• Do not sell the technical checklist as the fix. 32 of the 64 shops we measured score a perfect 100 on technical readiness, and 25 of those 32 appear in zero answers. Do the work — it is cheap and it is a prerequisite — and do not bill it as the thing that moves the number.
• Do not present one run as a trend. Two measurements with the same questions are a comparison. One is a starting point.
• Do not hide the method. The client will eventually ask which assistant, which questions, how many times. Having that on the page you hand over is what makes the number survive the second meeting.

The shape that works

1. The benchmark first: this is normal, here is the count.
2. Their number, with the sample size next to it.
3. Who is named instead, when anyone is.
4. The pages the assistant read, sorted by how open they are.
5. What you will do in the next 30 days, split into what you control and what you do not.
6. The date of the next measurement.

None of that requires our product. It requires asking the same questions twice, writing down what you asked, and not overstating what came back.

The limits of what we just told you

The 81 agencies and the 64 shops are the sets our own collection reached, not censuses, and a different collection would produce different sets. Three questions per agency is a coarse instrument: it supports the proportion appearing at all, not a ranking. Assistants vary by day and by phrasing, and a business we record as never named may well be named on a question we did not ask.

How to cite: Goppa Research (2026). How to show a client their AI visibility. https://trygoppa.com/answers/how-to-show-a-client-their-ai-visibility — CC BY 4.0.

No business is named for a poor result. We publish the counts and the method; the identity of individual businesses stays out of it.