A credible AEO agency pitch shows how the agency finds real buyer questions, what it will fix and who implements it, how it measures mentions versus citations versus revenue, and what it cannot promise. Buyers are increasingly screening for guaranteed-inclusion claims and opaque scores. This post lays out a pitch structure, slide by slide, built from published buyer guides and Google's documentation, as of 14 October 2026.
Key Takeaways
- Buyers want method and evidence, not a tour of AI search trends they already know.
- Never promise inclusion in AI answers. Google says there are no additional requirements or special optimisations for AI Overviews and AI Mode, and no outcome can be guaranteed.
- Separate four things in every report: mentions, citations, accuracy and referral or revenue outcomes.
- Show the delivery team and the approval workflow. Buyers are told to meet the people who do the work.
- Include a slide on what you will not do. It builds more trust than a fourth case study.
Who is reading the deck
There are usually three readers: a marketing lead who wants a plan, a finance or procurement person who wants terms and a stop clause, and sometimes a technical owner who needs to know what lands on the dev backlog. Huble's B2B and enterprise buyer guide frames the selection criteria in exactly these terms: buyer-question research, technical and content delivery, measurement, governance, evidence quality and the delivery team (Huble). Design the deck so each reader finds their answer without hunting.
If you are on the buying side, my earlier posts on AEO retainer pricing and scope and red flags in GEO agencies are the mirror image of this one.
Slide 1: The buyer's problem, in their words
Open with a specific, checkable observation, not a market statistic. For example: here are the ten questions your buyers ask, here is what three assistants say today, and here is where you are absent, wrong or outdated. This is a mini audit, and it works because it is about the client.
Keep it honest. AI answers vary between runs and engines, so show the date, engine, country and number of runs behind each screenshot.
Slide 2: How you find the questions
Huble suggests buyers weight question research heavily (it proposes 30% in its scorecard) and want prompt libraries built from sales, support and procurement input, with each question tied to a role, market and intent. Show your process:
- Interviews or call transcripts with sales and support.
- Search Console queries and on-site search.
- Community threads and review sites where buyers phrase problems.
- A prompt library mapped to funnel stage and persona.
Name the sources. "Proprietary prompt database" with no provenance is the sort of opacity buyers are warned about.
Slide 3: Diagnosis before tactics
Show the discovery checks you will run before writing anything: indexation, snippet eligibility, canonical and duplicate issues, crawl access for the relevant bots, and the state of the site's key facts. Google's AI features documentation says a page must be indexed and eligible to show in Search with a snippet to be used, and that no extra technical requirements apply (Google Search Central). Agencies that skip this and jump straight to "AI-optimised content" are solving the wrong problem. I describe my own order of operations in my AEO audit process.
Slide 4: What you will change, and who does it
Use a table. This is the slide procurement reads closest.
| Workstream | Examples | Who implements | Who approves |
|---|---|---|---|
| Technical | Indexing fixes, redirects, canonicals | Client dev or agency | Client tech lead |
| On-page | Answer-first rewrites, comparison tables, updated facts | Agency | Client SME |
| Off-site | Reviews, listings, digital PR, community | Joint | Client comms |
| Measurement | Prompt panel, GA4 segments, CRM field | Agency | Client analytics |
Huble's guide says buyers should see clear ownership of who implements and tests each change, and what needs the buyer's developers. Put that in writing.
Slide 5: Measurement you can audit
This slide separates serious pitches from decorative ones. Present the KPI structure with definitions and denominators. The one I use is the three-layer model in my dashboard spec: presence, perception, outcome.
HubSpot's KPI list gives useful definitions you can adopt: AI visibility rate as prompts where your brand appears divided by prompts tested, and citation share as your citations as a proportion of all brands' citations on the same prompts (HubSpot). Adobe's list adds answer accuracy and AI share of voice (Adobe for Business). Seer Interactive argues that visibility alone does not show how a brand is described and that accuracy needs a reference brand canon (Seer Interactive).
Say what is correlation. Huble's question list asks what an agency can attribute to observed CRM touchpoints and what remains correlational. If your answer is "most of it is correlational, here is how we triangulate," say so on the slide.
Slide 6: Evidence, with the awkward bits left in
Case studies need baseline, method, period and what else changed. A traffic chart with no baseline is a decoration. Huble recommends buyers ask for exactly that. Discovered Labs, itself an agency, advises checking that case studies show pipeline metrics and not only traffic (Discovered Labs). Treat that as one vendor's view of what to test.
If you do not have case studies in a category, say so, and offer a paid pilot with a defined scope instead. I cover how to read this kind of evidence in AEO case studies: what counts as evidence.
Slide 7: Governance and risk
Buyers in regulated or multi-market businesses will ask: how are claims approved, who verifies customer references, how is translation reviewed, and who has access to what. Huble treats compliance requirements as pass or fail gates that a high score elsewhere cannot offset. Include:
- A claims-approval workflow (who signs off statistics and testimonials).
- Access control: which accounts and data you need and when you lose access.
- AI use disclosure: Google's guidance on helpful content suggests explaining how content was produced, including automation or AI, when readers might reasonably wonder (Google Search Central).
Slide 8: What you will not do
This is the slide most decks lack. List it plainly:
- No guaranteed citation counts, rankings, or revenue.
- No fabricated reviews, undisclosed endorsements or mass placements designed to influence AI answers. Huble names these as red flags.
- No large volumes of near-duplicate pages. Google's spam policy describes scaled content abuse as many pages generated mainly to manipulate rankings, including AI-generated pages that add no value (Google spam policies).
- No reliance on special files for AI visibility. Google says you do not need new machine-readable files, AI text files or markup.
Slide 9: Team and cadence
Name the people who will do the work, their roles and time allocation. Huble says buyers should meet the delivery team, not only the pitching strategist. Add the cadence: a weekly working check-in, a monthly measurement review with the methodology box, and a quarterly plan reset.
Include what is subcontracted. If you use freelancers or tools that touch client data, say so.
Slide 10: Commercial terms
Buyers are advised to scrutinise contract terms. Discovered Labs recommends looking for month-to-month or short terms, client ownership of content and data, and itemised pricing, and lists auto-renewals with long cancellation windows as a red flag. I would go further and offer:
- A defined 60 to 90 day first phase with deliverables and a decision point.
- Client ownership of prompt libraries, reports and content on exit.
- A written handover on termination.
Price structure options are in my retainer pricing post.
Slide 11: The first 90 days
| Weeks | Output |
|---|---|
| 1-2 | Audit, prompt library v1, brand canon, baseline panel |
| 3-6 | Technical fixes, top-ten page rewrites, listing and facts clean-up |
| 7-10 | Off-site work, comparison and pricing pages, second panel run |
| 11-13 | Review against baseline, report with caveats, decide next phase |
Slide 12: Questions we expect
Close with the buyer's own list. Huble's questions are a good template: which engines, languages and regions do you monitor; how do you separate mentions, citations, accuracy and referrals; what was the baseline in a comparable engagement; which changes need our developers; what happens to our data when we leave. Answering them before they are asked signals confidence.
How to deliver it
Keep the deck to about a dozen slides and put detail in an appendix. Send the mini audit beforehand. Bring one sample report with real structure but redacted data.
One more point from the buyer side: some buyers will test you. Discovered Labs suggests running buyer-intent queries to see whether the agency appears in AI vendor lists. Do not panic if you do not appear, but be prepared to explain what you measure about yourself and what you have learned.
FAQ
What is an AEO agency pitch deck?
It is the presentation an agency uses to win AI search optimisation work. A good one shows method, measurement, ownership, evidence and limits, not only a trend overview.
Should a pitch promise AI citations?
No. Huble's guide says guaranteed inclusion in AI answers should trigger scrutiny, and Google states there is no special optimisation that ensures appearance in its AI features.
How many case studies do I need?
Quality matters more than count. Each should show baseline, method, period and confounders. If you lack relevant ones, propose a scoped pilot.
What should I put on the pricing slide?
Scope, phases, itemised pricing, client ownership of outputs, and an exit path. Avoid long lock-ins without performance checkpoints.
How do I handle questions about attribution?
Separate observed touchpoints from correlation. Use branded search lift, GA4 AI referral segments and a self-reported CRM field, and label each by confidence.
Should the deck mention llms.txt or special schema?
Only to explain you will not rely on them for AI visibility. Google's documentation says no new files or special schema are needed for AI features. Structured data can still support normal rich results.
Who should present?
Someone who will also do the work, plus the account lead. Buyer guides specifically advise meeting the delivery team.
How long should the deck be?
Around a dozen core slides, with an appendix for audit detail and sample reports.
Is this useful for freelancers too?
Yes. The structure scales down. A freelancer's deck can be five slides: problem, method, measurement, scope and terms, and what is not promised.
How do I show AI visibility for my own agency?
Run your own prompt panel, publish the method, and be willing to show the gaps. It demonstrates you apply what you sell.
Let's talk about your pitch
I am Younus Fardeen, an organic growth, SEO and AEO strategist with 4+ years of marketing experience in edtech and startup work. If you want a second pair of eyes on an AEO proposal, or help scoping one, see my work and use the contact form at https://younusfardeen.in.