How to present an AI visibility service in your sales proposals (structure and examples)
The meeting went well, the client understood the problem, they asked you to "send me something in writing"… and that's where half of all AI visibility sales die. Because explaining live that the AI doesn't recommend their business is one thing, and putting it in a proposal in a way that a business owner —who'll read it diagonally, probably on their phone— understands the problem, the plan and the price without you being there is another.
SEO proposals have twenty years of well-worn templates. GEO ones (Generative Engine Optimization: working on a brand's presence in the answers of ChatGPT, Gemini, Perplexity and company) every agency is improvising. This article is the standard section, block by block, with sample text you can adapt in ten minutes.
The structure has five blocks, and the order matters: problem → what we'll do → what we'll measure → what we don't promise → price.
Block 1: the problem, with its data and the client's
No starting with your methodology. Start with the change in search behaviour, two or three sourced facts, and —this is what separates your proposal from a generic PDF— the client's specific picture.
Key data
Where your potential customer searches today
The way people search is changing fast. Frequent use of ChatGPT in Spain has gone from 4% to 28% in just two years (Funcas, III Survey on AI, 2026), and 37.9% of the population used generative AI in the last quarter of 2025 (INE). More and more purchase decisions start with a question to an AI, not a Google search.
We've checked where [company] appears when someone asks those questions. We tested [N] real questions —for example, "[tested prompt]"— in ChatGPT, Gemini and Perplexity. Result: [doesn't appear in any / appears in only one / appears, but with incorrect data]. We attach the dated screenshots.
It's not an isolated case: in a study of 9,865 Spanish SMEs, 91% appeared in only 1 of the 4 main AIs. The good part: your competition, in all likelihood, is in the same position — and that turns the problem into a window of opportunity.
Notice the mechanism: market fact (it's a real phenomenon), client fact (it's YOUR problem), context fact (you're not alone, but you can get ahead). If you don't yet have the client's check, don't send the proposal: that picture is what makes it inevitable.
Block 2: what we'll do
Concrete, in the language of deliverables, without jargon. Four or five lines of work, not fifteen.
Key data
What we'll do
- Full baseline (month 1): we'll define with you the [30-40] questions your potential customer asks and measure where you appear today across the main AIs. That documented starting point is what will let us see the evolution.
- Source correction: AIs feed on your website, directories, listings and reviews. We'll correct wrong or incomplete data and reinforce the sources they draw from.
- Citable content: each month we'll publish content that answers the detected questions, structured so AIs can cite it.
- Weekly tracking: we'll repeat the measurements every week, because AI answers change constantly and a one-off data point is worth nothing.
- Monthly report: a PDF report with the evolution, what was done that month and what's planned for the next.
Block 3: what we'll measure
This is where it's decided whether, in six months, the renewal is easy or impossible. Always separate what's under your control from what you can only observe — the long version of this distinction is in what KPIs to put in a GEO proposal.
Key data
What we'll measure
Work commitments (they depend on us and we guarantee them): questions monitored every week, corrections applied to the sources, content pieces published.
Outcome indicators (we measure and report them every month): in how many of the monitored questions you appear, across how many AIs, what the AI says about you and how it evolves against [competitor].
Block 4: what we do NOT promise
The block almost no one includes and the one that generates the most trust. In writing, before signing:
Key data
What we won't promise you
No one controls what an AI answers: neither us, nor any agency, nor OpenAI itself can guarantee that a specific brand appears in a specific answer. That's why we don't promise "appearing in ChatGPT in X weeks" or a guaranteed number of mentions. If anyone promises you that, be suspicious.
What we do guarantee is the work that's under our control: continuous measurement, the corrections, the content and an honest report where you can see the evolution, improving at whatever pace it improves.
This doesn't weaken the proposal: it sets it apart. The client is comparing your PDF with someone's who promises miracles, and this block turns the other promise into a red flag. What can genuinely be committed to in this service is developed in what to promise (and what not) about AI visibility.
Block 5: the price, anchored and without a menu
A monthly figure with everything inside, anchored to a known reference, without breaking it down by hours (hours invite haggling; the result doesn't).
Key data
Investment
AI visibility service: [€450]/month, no lock-in, with everything described included (weekly measurement, corrections, [N] content pieces/month and a monthly report). For reference, an SEO service is priced in Spain between €600 and €4,000/month (pacoruben.com); this service covers the channel that's growing on top of that base.
The initial audit that accompanies this proposal isn't billed: it's yours, whatever you decide.
The ranges by client type —local, SME, B2B— and the margin arithmetic are in how much to charge for AI visibility services.
The detail that multiplies the close: screenshots, not paragraphs
An AI visibility proposal with real screenshots of the client defends itself; without them, it's theory. The problem is producing them: testing 30 questions across 3 or 4 AIs, by hand, for every proposal you send, is hours you don't bill. This is what we built Surfeo for agencies for, which includes 3 pitch spaces designed exactly for pre-sales: you audit the potential client before they sign anything and the proposal goes out with their real picture inside — where they appear, where they don't and what the AI says about them.
Frequently asked questions
How long should this section be in the proposal?
Two pages, three at most with the screenshots. If the service goes inside a larger proposal (SEO + AI, web + AI), a page and a half: problem with the client's picture, plan in five lines, what you don't promise and price.
Do I put the screenshots inside the proposal or as an annex?
Inside, in block 1, the two or three most telling ones (ideally one where a competitor appears and the client doesn't). The rest, as an annex. The screenshot is the argument; don't hide it at the end.
What if the client compares my proposal with a cheaper one that promises results?
It's the likely scenario, and block 4 exists for that. Add a line if needed: "ask any provider how they guarantee appearances in a system they don't control, and ask to see their measurement". Honesty competes badly in headlines and very well in renewals.
Do I send the proposal without having had a meeting?
Better not. The proposal closes what the meeting opens. If there's no meeting yet, send only the mini-audit with two screenshots and ask for 30 minutes — the full flow is in the audit as a sales hook.
You've got the structure; what's missing is the client's picture that makes it irrefutable. Run the free visibility test on your next prospect's website and build the first proposal this week.
