Questions to Ask an AEO Agency Before You Sign

Questions to ask an AEO agency should reveal how it tests AI search, what it will actually deliver, and whether its claims stand up to verification. Use this checklist to compare proposals, request evidence and spot guarantees that no agency can control.

Answer Engine Optimisation (AEO) aims to help a brand appear in answers from tools such as ChatGPT, Perplexity, Claude, Gemini and Google AI Overviews. It overlaps with search engine optimisation (SEO), but a credible programme should do more than rename familiar SEO tasks. If you are comparing a specialist with an existing SEO partner, start with SearchFIT’s agency comparison, then use the questions below to assess the work itself.

You do not need every answer to be perfect in a sales meeting. You do need clear explanations, verifiable examples and a plan that fits your business. Ask for the evidence behind claims, not just a polished dashboard or a list of services.

1. Check that the agency understands your goals

• Which business outcome are you proposing to improve? Ask the agency to connect its work to qualified leads, sales, sign-ups or another outcome your team values, because more AI mentions alone may not help the business.
• How will you define a useful AI search appearance for us? A useful result might be a recommendation in a buyer prompt, a citation to a product page or a referral visit, and the definition should match your buying journey.
• Which audiences and markets are included? Ask it to specify the customer types, locations, languages and product categories covered, because an overall visibility claim can hide gaps in the markets that matter.
• Which AI search surfaces will you work on, and why those? The agency should explain its choice of ChatGPT, Perplexity, Claude, Gemini or Google AI Overviews based on your customers, rather than promising equal coverage without a reason.
• What is outside the proposed scope? Get exclusions such as paid media, digital PR, development, content production or analytics implementation in writing so you can see which responsibilities stay with your team.
• How does the plan fit with our existing SEO and marketing work? A credible proposal should identify shared foundations and distinct AEO tasks, avoiding duplicate fees for technical or content work already underway.
• What would make you recommend that we do not hire you? A specialist should be able to name cases where AEO is premature, where another channel deserves priority, or where your organisation cannot yet support the work.

2. Test the agency’s AEO methodology

• How do you distinguish live-web retrieval from information models may already know? Ask for a plain-English explanation of how the plan addresses sources retrieved during a search as well as the broader information associated with your brand.
• What does your process look like from diagnosis to action? Request the steps, owners and decision points, because a repeatable process is more useful than a collection of loosely connected tactics.
• How do you decide which questions to test? The agency should combine buyer research, sales language, customer questions and category prompts, rather than selecting prompts only because your brand already appears in them.
• How do you separate AEO work from standard SEO? Ask what is genuinely different in its research, technical checks, content structure, off-site work and measurement; new labels on familiar deliverables are not proof of a new method.
• How does the approach change between AI platforms? Ask for examples of how it adapts tests and recommendations to different products, since one set of results should not be treated as a universal view of AI search.
• What does your agency know about answer instability? It should explain that answers can vary by prompt, time and product experience, then describe how it handles that variation instead of treating one response as a lasting ranking.
• How do you decide what not to optimise? A sound method prioritises opportunities connected to the business and evidence, rather than producing a long task list simply to make a retainer look busy.
• Can you show us a sample strategy with sensitive details removed? A redacted plan can reveal the quality of the thinking, while protecting client information and making it easier to assess whether the agency can turn observations into work.

For a second perspective on the questions buyers can use to assess a provider, read eSEOspace’s guide to questions for a GEO agency. Treat any checklist as a starting point; the agency still needs to answer in terms of your market and goals.

3. Ask for a real baseline before signing

• Can you show our current visibility before the contract begins? Request a dated baseline rather than a promise to measure later, because you need a starting point to judge whether the work changes anything.
• Which prompts will you include in that baseline? Ask for buyer-intent prompts such as category recommendations, alternatives, use-case searches and questions about your brand, not just a few queries that mention your name.
• Will you test prompts where we are not named? Unbranded questions reveal whether the brand enters a category answer naturally, while branded prompts alone can overstate how discoverable you are to new buyers.
• How many prompts are included, and how were they selected? One industry guide suggests asking for a baseline covering at least 15 prompts; the agency should explain why that set represents your audience and how it will be maintained.
• Which platforms and product modes will the baseline cover? Ask it to state whether tests use live search or another mode, because results from different experiences are not directly interchangeable.
• Will you give us the exact prompts and test dates? Keep the full prompt list and timestamps so your team can rerun the tests and distinguish a real trend from a change in wording or timing.
• Will the report include the answers and source URLs? A result without the underlying response and cited pages is difficult to verify, interpret or turn into a useful content or technical task.
• Can you show what competitors appear alongside us? A baseline should help reveal which brands are recommended, which sources support those recommendations and where your own evidence or coverage is missing.

4. Demand proof that can be checked

• Can you show a case study with a clear starting point? Ask for the client’s initial situation, the measurement method and the period covered, because an improvement claim without a baseline does not show what changed.
• What specific work caused the reported change? The agency should connect outcomes to interventions such as content updates, technical fixes or earned mentions, rather than imply that a single tactic explains every result.
• Can the case study be independently checked? Ask for enough detail to validate the method, such as sample prompts, dated examples, citation URLs or a client reference where permitted.
• Will the result hold up in a live rerun? Ask the agency to repeat a claimed prompt while you observe, or provide a history of results, because a single screenshot may capture a temporary answer.
• How do you report variation between repeated tests? Look for a consistent method that shows whether a brand appears across a prompt group, not a selective screenshot presented as a fixed position.
• Can you show a case where your work did not produce the expected outcome? A thoughtful explanation of limits and adjustments is often more useful than a portfolio containing only success stories.
• Are the examples from the same type of business as ours? Differences in category, customer journey, regulation and website setup affect how transferable a case study is, so ask what is comparable and what is not.

A good agency can describe proof without asking you to accept a secret score. Optimist’s AEO agency guide is another reference for understanding the services agencies may offer; use it to frame follow-up questions about the actual work and evidence in your proposal.

5. Check technical access and crawlability

• How will you check whether AI crawlers can access important pages? Ask for a technical review of relevant bot access, page responses and content availability, because a blocked or incomplete page cannot support the visibility you want.
• Will you inspect the HTML delivered to a crawler? The agency should check whether key information is present in the page response, not assume that a human-visible page is equally accessible to every automated system.
• How do you assess JavaScript-heavy pages? Ask it to test whether important content appears without relying on client-side rendering, then state when a development change is necessary rather than treating a bot-access issue as a copywriting problem.
• Will you review robots.txt and relevant response errors? Request a record of any access restrictions or technical failures and the proposed fix, because recommendations should be tied to what the site actually returns.
• Which pages and templates will you inspect? Confirm coverage of product, category, pricing, comparison, documentation and other commercially important pages, not just the homepage or a sample blog post.
• What role do structured data and schema markup play in your plan? The agency should explain where structured data can clarify page meaning and where it cannot replace clear content, accessible pages or corroborating information.
• Do you recommend an llms.txt file, and what problem would it solve for us? Ask for a use case and implementation plan rather than accepting a file as a magic switch that guarantees crawling or citations.
• Who will implement and verify technical fixes? Establish whether the agency, your developers or another supplier owns each change, and require a post-release check so recommendations do not remain in a ticket queue.

6. Evaluate content recommendations

• How will you choose which pages to update or create? Ask for a prioritised list tied to buyer questions and business value, because publishing more pages is not a strategy unless each page serves a clear purpose.
• Will content answer the question directly and accurately? The agency should show how it will make key facts easy to find while preserving the context, caveats and detail needed for a trustworthy answer.
• How will you use our subject experts? Ask who checks claims, product details and regulated information, because an agency should not invent expertise or publish unverified statements in your brand’s voice.
• How do you prevent near-duplicate pages made for similar prompts? Require a plan to consolidate overlapping topics, since many thin pages can confuse readers and create maintenance work without adding useful evidence.
• What editorial standards will apply to AI-assisted writing? Ask how the agency checks accuracy, originality, tone and citations, and who takes responsibility for every published claim.
• How do you plan to keep product facts current? Set out how prices, features, availability and policies will be checked and updated, because outdated details can undermine both user confidence and future recommendations.
• Can you show a before-and-after content example? A real example should explain what changed and why, not just display a rewritten paragraph with no connection to a user question or measurement.
• How will content support both AI answers and conventional search? The agency should explain how useful information, clear structure and sound technical foundations can serve people and search systems without producing awkward copy for machines.

7. Probe the off-site and reputation plan

• What work will happen beyond our own website? Ask for a clear share of effort devoted to relevant external sources, because an on-site-only plan may miss the reviews, comparisons and independent references buyers consult.
• Which third-party sources matter in our category, and why? The agency should identify credible industry publications, directories, review sites or communities based on your audience, rather than promise a generic number of placements.
• How will you earn coverage or mentions? Request the outreach approach, target types and quality checks, because a list of placements is not meaningful if the sources are irrelevant or lack editorial independence.
• Are any placements paid, sponsored or otherwise incentivised? Require disclosure rules and a record of costs, because paid inclusion should never be presented as independent endorsement.
• What is your approach to reviews and customer feedback? Look for ethical ways to request honest reviews and respond to concerns, not scripts that pressure customers or attempt to manufacture sentiment.
• How do you handle community discussions and forums? The agency should explain how it will listen, contribute transparently where appropriate and avoid disguised promotion that could harm trust.
• What evidence will you provide for off-site work? Ask for outreach records, live URLs and the status of each opportunity so you can distinguish a pitch sent from coverage actually secured.
• How will you protect our reputation if a publisher declines or gets facts wrong? Agree how corrections, approvals and escalation will work before someone represents your business externally.

8. Make measurement useful to the business

• Which visibility measures will you report, and how are they calculated? Ask for plain definitions of mentions, citations, citation share or other measures, because a proprietary label is not useful if nobody can reproduce it.
• Will the report show the actual prompts and source pages? Require enough underlying data to trace a result back to the question and cited URL, rather than a summary score with no audit trail.
• How will you track AI referral visits? Ask whether analytics can identify referrals from AI platforms and how the agency will check that channel groupings are configured correctly.
• How will you connect visibility to leads, sales or sign-ups? Discuss analytics events, CRM fields and customer surveys, since referral data alone may not capture someone who discovers you in an AI answer and converts later through another route.
• Will you include self-reported attribution? Consider asking new customers where they first heard about you, because not every influence will appear as a trackable website visit.
• How will you report sentiment and factual accuracy? A brand mention is not automatically positive or correct, so reports should flag misstatements and distinguish a recommendation from a neutral reference.
• How often will we review results and make decisions? Set a useful cadence for sharing findings and agreeing next actions, rather than receiving frequent reports that do not change the work.
• Can we export the raw history and use our own analytics? Data access keeps your team able to validate trends, compare providers and continue measurement if the engagement ends.

For a practical starting point on choosing measures, see SearchFIT’s guide to AEO KPIs. The important test is whether each measure has a clear definition, a business use and a way to check the underlying result.

9. Get the deliverables and responsibilities in writing

• What exactly will you deliver in each phase? Ask for named outputs such as a prompt baseline, technical findings, page recommendations, content briefs, outreach activity and reporting, not a vague promise to improve visibility.
• What is the 30-60-90 day roadmap? Request an itemised plan showing what happens in each period, who owns it and what your team must supply, so you can assess whether the proposed pace is realistic.
• Which tasks depend on our developers, writers or legal reviewers? Identify dependencies early, because an agency cannot deliver a technical fix or approved claim without access and decisions from your organisation.
• How many pages, prompts or markets are included? Write down the scope and any limits so both sides understand what is covered and what would trigger a change in fees.
• What is the approval process for content and outreach? Agree who reviews drafts, how feedback is handled and when approvals are needed, protecting your brand while avoiding unnecessary delays.
• How will the agency prioritise work if capacity is limited? Ask what gets done first and what is deferred, because a transparent queue is better than silently dropping lower-priority commitments.
• What does completion mean for each deliverable? Define whether a task is finished when a recommendation is written, implemented, published or verified, since those are different levels of work.
• What will you need from us before starting? Ask for an access, data and stakeholder checklist so the engagement does not lose time waiting for credentials, approvals or basic business information.

10. Understand pricing and value

• How did you build the proposed fee? Ask the agency to connect price to people, scope, tools and delivery, because a monthly number without a workload breakdown is difficult to compare.
• What does the fee exclude? Clarify additional charges for development, media costs, content, software, travel, digital PR or extra markets before comparing proposals.
• How should we interpret your price against the market? First Page Sage’s AEO and GEO pricing report, based on 1,146 quotes across 214 agencies, puts the median monthly agency retainer at $6,850 and the traditional SEO median at $4,740; use those figures as context, not a required rate or guarantee of value. See its AEO and GEO pricing report.
• What result or capability does each part of the fee buy? Ask for a link between cost and outputs such as analysis, technical implementation or outreach, rather than paying extra for a branded method with no practical difference.
• Can we phase the work or start with a defined audit? A contained first phase can help your team evaluate the agency’s diagnosis before committing to a longer programme, provided the audit has useful outputs and clear ownership.
• What would make the programme cost more later? Get the likely triggers, such as added markets, platforms, content volume or development needs, so a low starting quote does not hide predictable extras.
• How does your software or reporting fee work? Ask which tools are included, who owns the account and whether you can retain access to the data if you stop working together.

11. Read the contract for limits and ownership

• What do you explicitly not guarantee? A responsible provider should not promise a fixed position or permanent inclusion in an AI answer, because the agency does not control how those systems generate results.
• What outcomes are you committing to deliver instead? Look for controllable commitments such as agreed analysis, implemented work, documented testing and reporting, rather than a guaranteed outcome the agency cannot directly command.
• Who owns the content and technical assets? Confirm that you retain rights to paid-for copy, briefs, schema code, files and other work product, subject to clearly stated third-party licences.
• Who owns the prompt history and reporting data? Specify that you can export the raw records and understand any limits on data access after termination.
• What is the contract term, notice period and renewal process? Ask for the practical exit steps and deadlines so you are not surprised by automatic renewal or an avoidable extension.
• What happens to unfinished work when the contract ends? Agree whether drafts, outreach records, recommendations and implementation files will be handed over, completed or removed.
• Can the agency use our name or results in its marketing? Set approval requirements for case studies, logos and performance claims, particularly where customer or commercial information is sensitive.
• How are confidentiality, data handling and account access covered? Review who can see your analytics and systems, what safeguards apply and how access is removed when staff or suppliers change.

For a useful counterpoint on guarantees and other warning signs, review AutoLander’s AEO agency red flags. Check its guidance against the actual promises, deliverables and legal terms in your own proposal.

12. Check the people, tools and communication

• Who will do the work after the sales call? Meet the people responsible for strategy, technical checks, writing and outreach, because a strong pitch does not tell you who will manage the account day to day.
• What experience does each person bring to our needs? Ask for relevant examples and the person’s role in them, rather than relying on a long agency client list with no link to the proposed team.
• Which tools will you use, and what do they actually measure? The agency should explain any platform or tracking product in plain terms, including its limits, data sources and whether you can export results.
• Do you have special access to an AI platform’s ranking system? Be wary of claims of private ranking access or a way to control a model’s answers; ask for public documentation or verifiable evidence of any exceptional claim.
• How will you communicate blockers and changing priorities? Agree who raises issues, where decisions are recorded and how the team will adjust when platform behaviour or business priorities change.
• What happens if the lead strategist leaves? Ask about documentation, handover and cover, because your programme should not depend on knowledge held by one person alone.

13. Decide whether your existing SEO agency can do the work

• Can our current agency show evidence of AEO-specific testing? Ask it to demonstrate prompt selection, cross-platform testing and citation analysis, rather than simply adding AI language to an SEO presentation.
• Does it work on external brand signals as well as the website? Check whether it can plan credible third-party coverage, reviews and entity information, or whether its offer ends at on-page updates.
• Can the current team work with your developers and analytics setup? An agency that understands your technical and measurement environment may be a better fit than a specialist unable to coordinate with the people who must implement changes.
• Are there capability gaps that need a partner? Ask whether digital PR, technical development, subject-matter review or data analysis should be supplied by another team, then define who coordinates that work.
• Would a software-plus-internal-team approach suit us? If you already have writers, developers or SEO support, compare the cost of adding measurement and expert guidance with outsourcing every task; SearchFIT’s pricing and resources can help you explore that option.
• What would we stop doing to make room for this programme? A new retainer needs time and resources, so ask the agency to identify which current activity should change instead of treating AEO as free extra work.

14. Score the proposal before you sign

• Does the agency show evidence rather than relying on a confident pitch? Give more weight to dated tests, traceable citations and work samples than to claims about proprietary insight or impressive-looking scores.
• Can you explain the plan to your team in plain English? If the agency cannot describe what it will do and why without jargon, it may not have a sufficiently clear strategy for your stakeholders.
• Do the deliverables cover technical, content, off-site and measurement needs? The right balance depends on your situation, but a proposal that ignores a known blocker should explain why it is safe to do so.
• Are the measurement rules agreed before work begins? Confirm the prompt set, reporting definitions and business measures in advance, so neither side can quietly change what success means later.
• Can you verify the work without the agency’s permission? Make sure your team can access source URLs, analytics, prompt logs and created assets, which reduces dependence on a closed reporting system.
• Would you still buy the work if AI visibility grew slowly or unevenly? Choose a programme that creates useful assets and insight along the way, not one whose value depends entirely on a quick change in model answers.
• Are the contract, scope and sales promises consistent? Resolve differences before signing, because a verbal promise is not a substitute for an agreed deliverable or written term.

What a credible proposal should include

Before you sign, ask the agency to provide three practical documents: a dated prompt baseline with responses and citation URLs; a case study that explains the starting point, actions and verification method; and a 30-60-90 day roadmap with owners, deliverables and dependencies. These requests do not guarantee that the work will succeed. They make it easier to judge whether the agency has a real method and whether you can hold both sides accountable.

If a proposal promises guaranteed placement, hides the prompt set, offers only generic blog volume or will not clarify data ownership, pause before committing. If it sets realistic limits, shows evidence you can inspect and spells out who will do what, you have a better basis for choosing a partner. Use SearchFIT’s agency comparison as a next step if you are weighing an agency against a software-supported approach, then take this checklist into your buying conversation.