How to sell AI visibility work without overpromising

Lead with the client’s customer, not the technology

The pitch that lands is not about models. It is one question: “When someone asks ChatGPT for a dentist in your city, what does it say?” Then you show them. You already have the answer, because you ran the report before the call.

The reaction that sells is the competitor’s name in a list they are not in. No amount of explaining what a generative engine is gets you to that reaction, and nothing after it needs explaining.

A practical consequence: run the report first, sell second. A sales conversation with the evidence in hand is a different conversation from one about a service you might provide.

What to promise

• The questions. Buyer questions, asked across three engines, in writing. The client can read them and say “my customers do ask that” — or tell you the questions are wrong, which is also useful and happens more than you would like.
• The work. A specific list of changes, split into what you will do and what only they can do, each with steps and a checkbox. The second column matters: claiming their own Google Business Profile is their job, and a client who was never told that will assume you failed to do it.
• The measurement. The same questions, re-asked after the changes are live (the re-test is being built; until then, run the free report again by hand). Before and after, on the same instrument.

What to refuse to promise

This is the part that separates an agency that will still have this client in a year from one that will not.

• A position. There is no rank. There is one answer and it is generated fresh. “We will get you to number one in ChatGPT” is not an ambitious promise, it is a meaningless one.
• A timeline. Nobody knows how quickly an engine picks up a new page, and the answer differs per engine and per question. Anyone quoting you thirty days made it up.
• A score going up. If you introduce a visibility number, every review meeting becomes an argument about the number rather than about the questions. We do not produce one, deliberately, and you should not invent one.

The strongest thing you can say in the first meeting is what you will not be able to control. It is the only claim in the whole pitch the client can verify immediately, and it makes the rest credible.

Packaging it

The shape most agencies land on, and the one the product is built around:

1. Audit up front — the report, read together, with your read on which items matter for their business. Typically $500–1,500.
2. Monthly retainer — you do the left-hand column, they tick off the right-hand column, and a re-test will run when something goes live.
3. A report each month with your logo on it — which is what the client is really paying for between the work landing and the answers changing.

Our pricing is per client site, not per article or per task, which is what makes that arithmetic work: 15 sites is $199 a month, and what you charge on top is yours.

Handling the awkward question

Sooner or later a client asks whether any of this actually works — does adding a comparison page and some markup really change what an engine says?

The honest answer is that the industry does not have good public evidence either way, because almost every vendor reports output (“we published 40 articles”) rather than outcome. We are building the re-test precisely because that gap bothered us, and it will produce answers we do not control — including, possibly, that some changes do nothing.

Saying that out loud costs you nothing with a serious client and buys you the only thing worth having, which is being believed the next time you claim something did work. Before your next call, run the report on their site — and maybe on your own first.