AEO Agencies: 9 Questions to Ask Before You Sign

Before you sign with an answer engine optimization agency, ask nine questions that test how it measures AI visibility, what its method is, what it will actually ship, how it governs content and what happens when results stall. This matters because buyers now get vendor shortlists from ChatGPT, Perplexity and Google AI Overviews, then check Reddit threads and review sites that no brand controls. The agency market has grown faster than the methods behind it. Plenty of offers are an SEO retainer with a new label, a monthly screenshot of one prompt and a promise to "get you cited."

This guide reflects the procurement conversation as it stands in October 2026. It is written for teams that already run search and paid programs, and it covers strong and weak answers, the proof to request, scope and the first 90 days.

What you are actually buying from an AEO agency

An AEO agency sells work that gets AI systems to quote, cite and recommend your brand when buyers ask about your category. TechCrunch describes AEO as the practice of structuring a brand's content so AI engines can use it. In practice, the work covers your own pages and the third-party sources those engines read.

Five adjacent services now use the same vocabulary. Know which one you're being sold before you compare prices.

If a proposal treats GEO, AEO and "AI SEO" as one thing, that's fine. Check instead whether the deliverables differ from the SEO retainer you already pay for.

The nine questions to ask before you sign

The nine questions below test measurement, method, implementation, content, expertise, off-site presence, governance, channel fit and exit terms. Ask every shortlisted agency all nine in the same order, and score the answers side by side.

1. How do you measure our visibility in AI answers, and on which platforms?

Why it matters: AI answers change with each run, platform and phrasing. One screenshot proves nothing, and an agency that can't measure can't show you progress.

Strong answer: "We build a fixed prompt set mapped to your query clusters, such as 'best X for Y', 'X vs Y' and 'how to solve Z'. We run it on ChatGPT, Perplexity, Google AI Overviews, Gemini and Copilot, sample each prompt several times a week, and report the share of answers that cite or mention you against named competitors. The baseline comes before any content ships."

Weak answer: "We'll check whether you show up in ChatGPT and send screenshots."

Proof to request:

• An anonymized dashboard with citation share by platform and cluster, a competitor comparison and week-over-week change
• How they build and version the prompt list
• How they handle run-to-run variance, for example by averaging repeated samples
• Their tracking stack, checked against the AEO tools for share of answer you could license yourself

2. What is your methodology, and how does it differ from your SEO service?

Why it matters: AI systems retrieve passages, combine sources and often cite domains that aren't yours. A method built only for blue links misses most of that.

Strong answer: A clear sequence that starts with a baseline citation audit, then a source analysis of which domains the engines cite for your prompts, then a prioritized gap list. On-site page work and off-site presence follow, with a retest at a fixed interval. They name what changes versus SEO: passage-level answers, consistent entity descriptions across sources and third-party coverage.

Weak answer: Keyword research, backlinks and "content optimized for AI," with last year's SEO deliverables.

Proof to request:

• An anonymized audit showing baseline citation share, cited domains per cluster, page-level gaps and a ranked fix list
• A written workflow document, not a slide
• One rebuilt page, with the prompts it was cited for before and after

3. What will you implement on our site, and who does it?

Why it matters: Recommendations parked in an engineering backlog do nothing. Some of the work is technical. It includes structured data, robots.txt and CDN bot rules that can silently block OAI-SearchBot or PerplexityBot, and JavaScript-rendered content that some AI crawlers don't execute.

Strong answer: A structured data plan that lists JSON-LD types per template (Organization, Product or SoftwareApplication, Article with an author Person, BreadcrumbList) and maps the properties. It places the markup in page templates rather than injecting it client-side, and validates it with the Schema Markup Validator and Google's Rich Results Test. They name who deploys it, for example the agency publishing to your CMS with your engineers approving template changes.

Weak answer: "We'll send recommendations for your developers."

Proof to request:

• A sample structured data plan with validation results and a named deployment owner
• Experience with your CMS, whether WordPress, Contentful, Webflow or Shopify
• A bot-access audit of your robots.txt and CDN rules

4. How do you write content that answer engines quote?

Why it matters: Answer engines pull self-contained passages. A 2,000-word post that buries its answer in paragraph nine is rarely the one they pick.

Strong answer: Page types matched to buying questions, such as comparison pages, "X vs Y," alternatives, pricing explainers and integration pages, each built from an answer-first brief that includes:

• The target prompt cluster and the one-sentence answer the page must lead with
• A comparison table spec with the criteria buyers actually use
• Required facts, each with a primary source and date
• The SME to interview, the schema type and the next refresh date

Weak answer: "We'll publish 40 AI-optimized blog posts a month."

Proof to request: A real brief, plus two published pages and the prompts they are currently cited for.

5. How do you turn our expertise into evidence?

Why it matters: E-E-A-T is easy to say and hard to run. Engines and buyers both weigh who wrote a page and what it's based on.

Strong answer: A documented workflow covering each step:

1. A 30-minute SME interview before drafting
2. A draft, then SME review, then a fact check against primary sources
3. Real bylines, with author pages and Person schema linking to credentials
4. Original research where possible, such as a customer survey or anonymized product data
5. An update SLA, for example every top-20 page reviewed every 90 days

Weak answer: "Our writers cover every industry," or ghostwritten posts under your CEO's name without review.

Proof to request: The review workflow document, a sample author page and the update SLA written into the contract.

6. What do you do off our site?

Why it matters: AI answers frequently cite Reddit threads, review sites, YouTube and trade publications. If competitors own those sources, on-site work hits a ceiling.

Strong answer: A source map of the domains cited for your priority prompts, with a plan for each. That covers Reddit participation within each subreddit's rules and with disclosed affiliation, review programs on G2 or similar sites, and PR aimed at publications the engines already cite.

Weak answer: "We'll build backlinks," or any hint of unattributed accounts posting on Reddit.

Proof to request:

• A source map for 20 of your prompts, produced during the sales process
• Their written Reddit and disclosure policy

7. Who approves what, and how is it recorded?

Why it matters: The agency publishes under your name, on your site and in public communities. One off-brand Reddit reply or unapproved product claim creates legal and brand risk.

Strong answer: An approval gate before anything goes live, named reviewers per content type, a claims and style guide, an audit trail of who approved what and when, and a way to roll back.

Weak answer: "We'll share a Google Doc," or "we post and you review afterwards."

Proof to request: The governance document and a redacted approval log from a current client.

8. Do you also run SEO and paid search, and how do they connect?

Why it matters: Many SEO agencies treat paid search as a separate service, so don't assume a "full-service" label covers Google Ads. Even with separate teams, paid query data shows which questions convert, and ad claims should match what AI repeats about you.

Strong answer: A plain list of what's in and out of scope. If paid sits elsewhere, they describe the handoff: shared query clusters, a shared claims library and a monthly sync with your paid team.

Weak answer: "We're full-service," with no named owner per channel.

Proof to request: A statement of work listing every channel and the person responsible for each.

9. What happens if it isn't working by day 90?

Why it matters: Without exit criteria, a weak program can run for a year on activity reports.

Strong answer: Leading indicators agreed for day 30, 60 and 90, a formal day-90 review, the right to re-scope and a reasonable notice period after the initial term. No citation guarantees, because no agency controls how the models choose sources.

Weak answer: "Guaranteed #1 in ChatGPT," or a 12-month lock-in with no review point.

Proof to request: Draft contract terms and a reference client whose engagement was re-scoped partway through.

Red flags in AEO agency pitches

The clearest red flags are guarantees, rebranded SEO, missing technical detail, vague reporting, no governance, single-platform testing and suppression tactics sold as AEO. The table below sets each one against what a credible agency offers instead.

That last row matters for brands with a reputation problem. Reverse SEO is a variant of search engine optimization focused on suppression, which is a different job with different risks. If a pitch blends the two, ask which budget pays for which.

What an AEO engagement should include

A sound AEO engagement names its deliverables, cadence and owners in the statement of work, and anything not written down should be treated as not included. At minimum, the scope covers:

• Content audit: every priority page scored for answer-first structure, sourcing and freshness
• Schema deployment: JSON-LD shipped to templates and validated, not only recommended
• Content refreshes: a set number of existing pages rebuilt per month
• Authorship: author pages, bylines, credentials and Person markup
• Source citations: a standard for every claim, with primary sources and dates
• Testing and reporting: the prompt set rerun weekly, with weekly data and a monthly narrative review

Pricing usually follows one of three models: a fixed-fee audit, a monthly retainer tied to deliverable volume, or a hybrid with an audit fee followed by a retainer. Compare them on output per month, not the headline fee. Our breakdown of AEO scope and pricing shows how to normalize proposals.

What you provide versus what the agency owns

How to measure success in the first 30, 60 and 90 days

Success in the first 90 days means a clean baseline, shipped work and early citation movement in priority clusters. Traffic is a lagging, partial signal, because many AI answers send no click at all. Track these metrics instead:

1. Day 30: prompt set agreed, baseline recorded, audit delivered, bot access fixed, first briefs approved.
2. Day 60: schema live on priority templates, first batch of rebuilt pages published, off-site source plan running.
3. Day 90: retest against baseline, cluster-level movement reviewed, scope adjusted for the next quarter.

As an illustration, take a cloud security brand that tracks 150 prompts across four platforms and starts with a 6% citation share in its "alternatives" cluster. If that cluster reaches 12% by day 90 while the "how to" cluster stays flat, next quarter's work shifts toward how-to content. That's a useful result even before pipeline data matures.

How Tellr runs answer visibility as one program

Tellr runs answer visibility as one governed program instead of a tracker that leaves the work to you. A senior team works on Tellr's own platform across the Reddit threads, Google results, AI answers and ad feeds your buyers read. Every reply, page and piece of creative goes through approvals with an audit trail, which is what questions 6, 7 and 8 test for. It is a managed program built for marketing teams spending $10k+ a month at companies worth $500M+ or with 200+ employees, so a smaller team will usually get better value from a self-serve tracker and in-house writing.

• Weekly tracking of who Google and its AI Overviews cite for your category's queries
• Comparison pages, reviews and answer-shaped articles built to be quoted by ChatGPT, Perplexity and AI Overviews, published to your CMS
• Subreddit mapping, a daily thread radar and guideline-checked replies behind an approval gate
• Category ad intelligence and ready-to-run creative that matches what AI says about you

Common mistakes buyers make when hiring an AEO agency

The most common mistake is buying measurement and calling it a program. A tracker tells you where you're missing, but someone still has to do the work that changes it. Other frequent mistakes:

• Choosing on a demo screenshot instead of a baseline built from your own prompts
• Signing without committing SME time, then getting generic content
• Scoping only on-site work when your category's answers cite Reddit and review sites
• Judging the program on organic traffic alone
• Skipping the build-or-buy question; see how enterprises weigh an agency or in-house team
• Hiring a generalist for a vertical with strict trust requirements

That last one deserves its own check, because the evidence engines need differs by vertical:

Run all nine questions, ask for the proof behind every answer, and write the scope, SLAs and day-90 review into the contract. An answer engine optimization agency that welcomes that level of scrutiny is usually the one worth signing.

FAQ