How to Build an AEO/GEO Proposal Clients Can Understand and Approve

An AEO/GEO proposal is commercially credible when a buyer can see three things without a call: the evidence that a gap exists, exactly what you will deliver to close it, and how both sides will know whether it worked. It names fees, owners and exclusions, separates new work from the SEO they already pay for, and promises measurement rather than placements.

This guide shows how to build that proposal after you've diagnosed a client. It explains each section by the question it answers for the buyer, then gives you a complete proposal for a fictional home services company that you can copy. If you haven't decided yet whether to sell this service, or to which clients, start with How to Sell AEO and GEO Services to Your Existing SEO Clients.

The full proposal is in this article, and you can also download it as a Markdown file to edit in any text editor or paste into a document. No sign-up is needed.

What do you need before you write the proposal?

A proposal written before the diagnosis is a guess with a price on it. Collect six inputs first.

For the baseline, sample each question several times per engine over at least two runs. AI answers change between asks, so a single screenshot isn't evidence. Serpify's AI Visibility collects this for you: you add the questions and competitors, and it reports mention and citation rates with 95% ranges, the competitors named and the sources cited, for ChatGPT, Gemini and Google AI Mode.

What question does each section answer for the buyer?

Buyers read a proposal looking for reasons to say no. Each section should remove one.

Write the executive summary last, and keep it under 150 words. A busy decision maker may read nothing else.

How do you scope the work so it doesn't overlap with SEO?

Split the work into five kinds and price each one separately. The split also makes it easy to show the client what their existing SEO contract already pays for.

The rule: if a deliverable is already in the SEO scope, deliver it there and say so in the proposal. In the worked example below, six pages need rewriting, two fall under the retainer's monthly on-page allowance, and the proposal charges for four. That one line does more for trust than any amount of positioning.

How do you define measurement without overpromising?

Agree definitions in the proposal so nobody argues about them at day 90.

• Mention rate: the share of sampled answers that name the client.
• Citation rate: the share of sampled answers that link to one of the client's pages.
• Competitor mention rate: the same, for each named competitor, in the same answers.
• Answers collected: questions × engines × answers per engine × runs. 40 questions × 3 engines × 3 answers × 2 runs = 720 answers.
• 95% range: the interval the true rate is likely to sit in, given the sample size.

Sample size matters more than most proposals admit. In the worked example, Kestrel is named in 58 of 720 answers (8.1%, range 6.3% to 10.3%). Narrow the same rate to one engine and 240 answers, and the range widens to roughly 4% to 10%. Narrow it to one question asked 9 times, and almost any result is possible. So set expectations in writing: a change smaller than the range is noise, and per-question results are directional only.

Variability isn't a quirk you can tune away. SparkToro and Gumshoe's January 2026 research found less than a 1 in 100 chance that ChatGPT or Google's AI returns the same list of brands twice for the same prompt. That is why the proposal promises measurement, not mentions. Completing the work makes accurate information about the client easier to find; Google, for example, says AI Overviews and AI Mode use the same SEO foundations as the rest of Search. It doesn't mean any assistant will cite the client.

Report traffic separately from answer measures. Google counts AI Overviews and AI Mode within Search Console's overall Web performance data, and visits from assistants such as ChatGPT show up as referral traffic only when someone clicks a link.

Weak versus better proposal statements

Vague lines are what get proposals rejected or, worse, approved and then disputed. Rewrite each one into something observable.

The complete proposal template

Fictional example. Brightline Search, Kestrel Home Services, Copperline Plumbing, Summit Heating Co and every figure below are invented to show the method. They are not real businesses or results. The fees are illustrative and are not market rates or Serpify prices. The suggested terms are not legal advice.

Proposal: AI answer visibility for Kestrel Home Services

Prepared by Brightline Search for Dana Okafor, Operations Director, Kestrel Home Services. Valid for 30 days from the date sent.

1. Executive summary

Kestrel wants more booked boiler and heating jobs from its three branch areas. When we asked ChatGPT, Gemini and Google AI Mode 40 questions local homeowners ask, Kestrel was named in 8.1% of 720 sampled answers. Copperline Plumbing was named in 31.0% and Summit Heating Co in 22.5%. The answers mostly cite two local directories, a review site and competitors' service pages; Kestrel's pages are cited in 1.9% of answers, and three listings show an old phone number.

We propose a fixed project to correct and complete the evidence these answers draw on, then monthly monitoring of the same 40 questions so you can see whether Kestrel is named more often. Fees: $4,700 one-off, then $1,200 a month. We will report what changes; we can't promise how any AI assistant will answer.

2. Your goal and how we'll judge this work

• Business goal: more booked boiler installation, repair and servicing jobs in the Harbour, Millbrook and Eastvale branch areas.
• What this project can show: whether Kestrel is named and cited more often in sampled AI answers to the agreed questions, whether the facts in those answers are correct, and how many visits and enquiries arrive from AI assistants where your analytics can identify them.
• Decision maker: Dana Okafor approves scope and fees. Sam Reyes (Marketing Coordinator) approves copy and supplies facts.

3. Baseline: what AI answers say today

Collected over two weekly runs: 40 questions, three engines (ChatGPT, Gemini, Google AI Mode), three answers per question per engine per run, United Kingdom, English. 720 answers in total.

What we found in the cited sources:

• Answers to "emergency boiler repair near me" style questions cite two local directories where Kestrel's listing has an old phone number and no Eastvale branch.
• Answers to "how much does a new boiler cost" cite competitors' pricing guides. Kestrel publishes no pricing information.
• Kestrel's six service-area pages share near-identical copy and don't state call-out hours, areas covered or boiler brands serviced.

These are observations from a sample, not a full record of what every homeowner sees.

4. The opportunity

Kestrel is named least for the questions that come before a booking: cost, emergency availability and "who services my boiler brand". Those are also the questions where the cited sources contain missing or wrong facts about Kestrel. Fixing those facts is in your control. Whether the assistants then name Kestrel more often is not, which is why we measure it every week.

5. Scope and deliverables

A. One-off implementation project: $4,700

Six service-area pages need this work. Two of them are rewritten under the existing SEO retainer's on-page allowance in month one, so this proposal charges for four.

B. Monitoring and ongoing improvement: $1,200 a month

• The 40 questions sampled every week on ChatGPT, Gemini and Google AI Mode, three answers per question per engine.
• A monthly report (see section 9) and a 30-minute review call.
• Four hours of improvement work a month, chosen from the report's priority list. Unused hours don't roll over.

C. Optional, not included in the fees above: $1,800

• Outreach to three local guides and comparison articles that AI answers cite, asking for Kestrel to be considered. We pitch; publication is the publisher's decision.

6. How this relates to your SEO retainer

Your current SEO retainer covers technical SEO, on-page work on two pages a month, two blog articles a month and Google Business Profile updates. This proposal doesn't charge for any of those. The table shows where each piece of work sits.

7. 30/60/90-day plan

8. Responsibilities and dependencies

Kestrel will:

• name one person (Sam) to supply facts and approve copy within 5 working days;
• confirm prices, call-out hours, areas and brands for the new pages;
• give Brightline access to the website, Search Console, Google Analytics and the directory accounts listed in A4.

Brightline will:

• deliver each item to its acceptance criteria and log the date it went live;
• run the same questions with the same settings every week;
• flag any delay on either side in the weekly update.

Dependencies: dates in section 7 move by the length of any delay in approvals or access. Third-party sites decide when, and whether, they accept listing changes.

9. Measurement and reporting

Every monthly report includes:

• Tracked question set: the 40 questions, engines and market.
• Answers collected: how many answers each rate is based on.
• Mention rate: the share of sampled answers that name Kestrel, with a 95% range.
• Citation rate: the share of sampled answers that link to a Kestrel page, with a 95% range.
• Competitor comparison: how often Copperline Plumbing and Summit Heating Co are named in the same answers.
• Work log: what was completed and when.
• Observable changes: rates, cited sources and facts that changed after the work.
• Traffic and enquiries: visits and enquiries from AI assistants where Google Analytics can identify them.

How to read the numbers: AI answers vary from one ask to the next, so rates come with a range. A change smaller than the range is treated as noise. A change after a piece of work shows that the two happened in that order, not that one caused the other. The sample covers the tracked questions and settings, not every homeowner's answers.

10. Fees and payment

• First 90 days, without the option: $4,700 + (3 × $1,200) = $8,300.
• First 90 days, with the option: $8,300 + $1,800 = $10,100.
• Payment: 50% of the project fee ($2,350) on acceptance and 50% ($2,350) when items A2 to A5 are accepted. The monthly fee is invoiced at the start of each month.
• Suggested terms (confirm in your contract): monitoring runs month to month after the first three months, with 30 days' notice to stop.

11. What's not included

• Changes to the existing SEO retainer's scope or fees.
• Paid advertising, including ads shown inside AI assistants or search results.
• More than four improvement hours a month, or pages beyond A2 and A3 (quoted separately).
• Review generation, reputation management or responding to reviews.
• Any promise that an AI assistant will mention, recommend or cite Kestrel.

12. Acceptance and next step

To go ahead, reply to confirm:

1. Scope A and B (and C if wanted).
2. Sam Reyes as the approver for facts and copy.
3. A 45-minute kick-off call in the first week to sign off the question set.

We'll then send the agreement and the first invoice, and the first weekly sample runs within five working days of sign-off.

Checking the fees add up

Before sending, check the arithmetic the client will check:

• Project items: $600 + $1,400 + $1,200 + $900 + $600 = $4,700.
• 90 days: $4,700 + 3 × $1,200 = $8,300, or $10,100 with the $1,800 option.
• Payment split: $2,350 + $2,350 = $4,700.

Common proposal mistakes

• Pitching the acronym. "GEO services" means little to a homeowner-facing business. "Being named when people ask who fixes boilers in Eastvale" does.
• No baseline. Without before numbers, the day-90 report has nothing to compare against.
• One-answer evidence. A screenshot of one answer is easy to dispute and easy to reproduce differently.
• Charging twice. Billing for page work the SEO retainer already covers, under a new name.
• Mixing one-off and monthly fees. The client can't tell what stops and what recurs.
• Promising placements or uplift. It sets up a dispute you can't win.
• Hiding dependencies. If approvals take three weeks, the plan should say so.
• No exclusions. Unlisted work becomes expected work.

Pre-send checklist

• The goal is the client's, in their words.
• The baseline states the questions, engines, market, runs and answer counts.
• Every rate has a range.
• Every deliverable has acceptance criteria and an owner.
• Overlap with the SEO retainer is shown, and nothing is charged twice.
• One-off, monthly and optional fees are separate, and the totals add up.
• Exclusions include "no promise of AI mentions or citations".
• The next step is one clear action.
• Any fictional or sample data is gone, replaced with the client's own.

Your next step

Start with the evidence. Create a Serpify workspace, add the client as a project, enter your buyer questions and competitors, and let the weekly runs build the baseline for sections 3 and 4. New accounts get a 14-day trial of the advanced features with 10 AI questions and no card, which is enough to test a client's most important questions; a full 40-question baseline needs a paid plan. When the baseline is in, copy the template above (or the Markdown download), replace the fictional details and send it.