The first GEO conversation with an existing client is the easiest sales conversation you'll have this year, provided you don't set it up as a sales conversation. The client already trusts you, they already have the question ('what does ChatGPT say about us?') and you arrive with the answer instead of an offer.
Why existing clients are the best first clients
Three reasons. First, you already know their market, their services and their competitors; the prompt set you track is right from the start. Second, you already have access to their website and often to their content, so you can execute actions straight away. Third, you don't need to build trust; you only need to show that you see something they didn't see yet.
Pick clients you currently do SEO, content or PR for. There GEO is a natural extension: the same levers, a new yardstick. Don't start with the client who complains most about results, but with the client with the clearest market and the most competition. That's where the difference in AI answers is biggest and the story sharpest.
Preparation: the baseline
Add the client to the agency portal a week before the conversation and let tracking run. You didn't need to ask them anything for that: you only track public AI answers to public questions. After three tracking days (Monday, Wednesday, Friday) you have enough.
Before the conversation, read ten answers word for word. You're looking for three things you can quote: a question where a competitor is recommended and the client isn't; a fact that's wrong (a service the client no longer offers, a wrong location); and a source that comes back in many answers where the client is missing. With those three quotes you have the whole conversation.
The conversation in three screens
Open the live view or dashboard on a screen and show exactly three things, in this order. Don't talk about GEO, AI models or methods; talk about the client.
Screen 1: where you stand. 'These are fourteen questions your customers ask AI. On these questions you're mentioned in 18% of the answers. On ChatGPT 24%, on Perplexity 9%, on Gemini 21%.' Show one literal answer where a customer asks for a recommendation and the client isn't in it. Nothing works as well as the answer a customer gets today.
Screen 2: who does get mentioned. The competitor list with share of voice. 'Summit Climate Co. is in 46% of the answers. Redwood Home Services in 31%.' Often there's a name the client didn't know or didn't consider a competitor. That moment, the 'who is that?', is when the client feels the problem.
Screen 3: where the difference sits. The sources AI cites. 'These six websites come back in almost every answer. Summit is on four of them and you're not.' This is the most important screen, because it makes the problem solvable. The client doesn't see 'AI doesn't like me' but 'I'm missing in four places'. And that's work you can do.
Stop there. Almost every client now asks: 'So what do we do about it?' Only then do you talk about the service.
The four objections you'll hear
'Our customers don't use ChatGPT.' Answer with the numbers from the tracking itself: not how many people use AI, but what the answer looks like when someone does. And point to the referrers in Google Analytics: chatgpt.com, perplexity.ai and gemini.google.com are almost always already in there. The client may not use it; their customers do.
'Isn't that just SEO?' Show the difference instead of explaining it: 'You rank #2 on Google for this query, and you're in 0 of 9 AI answers to the same question.' That one example says more than an explanation of levers. If the client asks further: SEO is a precondition, not a guarantee; AI reads other sources and chooses differently.
'Can you guarantee we'll be in ChatGPT?' No, and say so. What you guarantee: tracking runs, the actions get executed, and every month the client sees exactly what changed and why. Clients value that honesty more than a promise you can't keep.
'What does it cost?' Name one amount for one package (see pricing GEO as a service), with what's in it: tracking, own login, weekly report, three to five actions a month, monthly report. Cancel monthly, with the agreement that the first three months are a fair test.
Closing without selling
Close with a concrete first step instead of a signature: 'This week I'll start the listing on comparison site X and open up the technical side; next month I'll show you what that did.' The client then doesn't have to decide on 'GEO', only on one action that makes sense. The service follows from the first report.
Give the client their own login in your branding straight away. A client who can look for themselves will look. And a client who looks sees the same thing you do: the competitor that gets mentioned, the source where they're missing. That does the selling for you.
After the conversation
Send a short email the same day with the three screens as numbers (visibility per AI model, the top-3 competitors, the four sources) and the agreed first action. Nothing more. Turn on the weekly report for the client in the portal, so they automatically get an update in your name on Friday. And start the first action that same week; the speed of that first action decides how seriously the client takes the service.
How to report afterwards without burying the client in metrics is in GEO reporting: which KPIs to show your client.
Open with the baseline, not a proposition. Show three things: where the client stands, who does get mentioned and on which sources the difference sits. Then the client asks what you're going to do about it.