AEO Agency RFP Template (2026): Questions to Ask and How to Score Them

The best AEO agency RFP questions can only be answered with evidence: a good writer can make a weak team sound strong, but nobody can invent a prompt list or a client report. OnlyAEO is an Answer Engine Optimization (AEO) and Generative Engine Optimization (GEO) agency, so we answer this list too. The crawler and AI-feature guidance below reflects OpenAI's and Google's documentation as checked in October 2026.

What should you include in an AEO agency RFP?

Give bidders enough context to fit their plan to you: your objective, your audiences and their questions, your current stack, the scope you expect and how you will choose.

Gigawatt Group's GEO agency RFP guide suggests covering:

• Business objective. Pipeline, market entry or reputation.
• Decision prompts. Buyer questions for defining a problem, comparing options and shortlisting, without your brand name.
• Current ecosystem. Website, analytics stack, CMS, CRM, subject-matter experts, content library and approval constraints.
• Expected scope. Advise, implement, write, measure or run the whole program.
• Selection method. Weighted criteria, required evidence, interview steps, commercial terms and the decision timeline.

Our guide to choosing an AEO agency for B2B software helps with the long list.

How do you evaluate an AEO agency's AI visibility measurement?

An agency that can measure AI visibility shows you its prompt set, sampling method, headline formula and a real client report.

Why a traffic report is not enough

Most of what an AI answer does for or against your brand never becomes a click.

In Pew Research Center's March 2025 browsing study of 900 U.S. adults, Google users clicked a traditional result on 8% of visits with an AI summary against 15% without one, and clicked a link inside the summary on 1% of visits.

Our guide to measuring AI visibility shows how to set your own baseline.

Questions to ask on measurement

• Q1. Prompt set. Can we see the full prompt set you would track? Expect a shared, editable set organized by audience, funnel stage, topic, geography and platform, tracking inclusion, citations, competitors, sentiment and accuracy over time.
• Q2. Engines covered. Which AI engines do you track, by official API or scraping? The Business Rover's agency guide treats ChatGPT, Google AI Overviews and Perplexity as a minimum, adding Gemini, AI Mode, Copilot or Claude by audience, and notes that scraped coverage breaks more often.
• Q3. Sampling. How often do you run each prompt, and how do you report variation? The same guide calls weekly sampling with several runs per prompt a sensible floor.
• Q4. Headline formula. What formula drives your headline number, and will it stay fixed? A changed method can make a flat quarter look like growth.
• Q5. Sample report. Can you send a redacted client report with per-engine results? 9Sail's guide for law firm CMOs treats an agency that cannot produce a sample report with platform-level data as an automatic disqualifier.
• Q6. Report contents. Beyond traffic, what will our monthly report measure? Expect prompt-level presence, citation frequency, cited domains, accuracy, sentiment, share of voice, referrals and pipeline, with methodology and sampling limits visible.
• Q7. Declines. How do you report declines, and can we rerun your prompts? A report that never shows a decline is a sales document.

How do you test an agency's technical work?

Ask which AI crawlers to let in, how the agency checks access and what structured data it adds. Good answers cite the platforms' own documentation.

OpenAI's crawler documentation describes GPTBot for training, OAI-SearchBot for ChatGPT search and ChatGPT-User for user-initiated fetches that may not follow robots.txt, and blocking OAI-SearchBot keeps a site out of ChatGPT search answers.

Google-Extended governs Gemini training and grounding use and does not affect Google Search inclusion or ranking. Anthropic separates ClaudeBot, Claude-SearchBot and Claude-User in its guidance for site owners.

Google's AI features documentation says: "There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary." It adds that no AI text files or special schema.org markup are needed. In August 2023, Google also said it would show FAQ rich results only for well-known government and health sites.

• Q8. Crawler rules. Which AI crawlers should our robots.txt allow or block? A strong answer separates crawlers by job. See the robots.txt mistake that blocks AI crawlers.
• Q9. Access checks. How do you verify AI crawlers can reach our pages? Google's page lists allowing crawling in robots.txt and in any CDN or hosting infrastructure among the basics.
• Q10. Structured data. What structured data do you recommend, and why? Expect accurate labeling, plus product-level markup for e-commerce catalogs. Our piece on what makes a SaaS site citable covers the rest.
• Q11. Implementation. Which technical changes will you implement yourselves? Ask about crawlability, indexation, schema, canonicalization, rendering, architecture, sitemaps and analytics, and confirm who owns implementation.

How do you judge an agency's content and accuracy?

Judge content by how the agency picks topics, researches, checks facts and gets approval.

• Q12. Content plan. How do you decide what to publish? Look for a plan built from buyer questions that gives each page a role: parent or supporting page, the intent it owns, internal links and update rules.
• Q13. Research. What research happens before a page is written? A strong answer checks what AI engines say now and reads the pages they cite.
• Q14. Fact checks. How do you check every claim, number and link? Expect named steps: sources traced, unsupported claims cut, links tested, your approval.
• Q15. Governance. How do you handle regulated claims and wrong AI answers about us? Expect approval owners, evidence standards, review cadence and escalation paths. Healthcare brands should also ask about YMYL content standards.
• Q16. Consistency. How do you keep our voice and facts consistent? Look for one record of facts and approved claims that every writer uses.
• Q17. Off-site work. What work happens off our site, and where? Ask which third-party sources the engines cite for your buyer questions and how the agency earns a place there.

What should an agency say about guarantees and results?

A credible AEO agency says plainly that it cannot guarantee AI citations. It should also explain what happens when results stall.

Guarantees and timelines

• Q18. Guarantees. Do you guarantee results? The strong answer is no. Gigawatt Group lists guaranteed citations as a warning sign and explains why: "No agency controls the output of independent AI platforms".
• Q19. Stalled results. What happens if nothing is moving after the first 90 days? Gigawatt Group says that window should establish a baseline, fix urgent technical and factual gaps, prioritize authority content and set a reporting cadence. Then expect a defined diagnosis, as 9Sail suggests: which signals are checked first, by whom and how fast.
• Q20. Lasting gains. Do gains persist if we stop working with you? An honest answer separates what you keep (pages, fixes, listings) from your place in AI answers, which nobody can promise since AI models change continuously.

Proof and references

• Q21. Client proof. Which prompts is a current client cited for, and how has that changed? The question exposes agencies that added AEO to their services without changing how they deliver.
• Q22. Industry reference. Can we speak with a client reference in our industry? 9Sail suggests a reference call in your own industry before scoring this criterion as met.
• Q23. Past failures. Tell us about an engagement that went wrong and what you changed. Candor here predicts how the agency will report a bad month.

What should you ask about the team, security and contract?

Ask who does the work, what access they need, who else they serve and what they will sign, then put the answers in the contract.

• Q24. Delivery team. Who exactly will do the work? Request the actual delivery team, experience, time allocation and division of responsibilities, with staffing that covers the scope.
• Q25. System access. What system access do you need, and how are credentials handled? Expect a written list at the lowest workable permission.
• Q26. Conflicts. Do you work with our competitors, and what is your conflict policy? 9Sail treats an agency that cannot describe its conflict policy as an automatic disqualifier for enterprise law firms.
• Q27. Pricing. How is pricing structured, and what is out of scope? Ask for transparent pricing, assumptions, optional scope and ownership of deliverables. In 9Sail's view, pricing tied purely to output counts signals production hours more than a managed visibility program.
• Q28. Contract terms. Will the contract fix report dates, a named lead, methodology and data export on exit? Strong agencies agree in writing.

How do you score AEO agency RFP responses?

Score the answers on a sheet you set up before any response arrives: weight each question group from 1 to 3 and score each answer from 1 to 5.

9Sail recommends a weighted scorecard over gut-feel scoring, since the field is too new for a committee to hold consistent standards without one. Published weights differ: Gigawatt Group gives measurement and business alignment 20%, while 9Sail gives AI citation methodology 25% in a scorecard for law firms.

Run the scoring in order

1. Apply disqualifiers. Remove any bidder that refuses to share its prompt set, cannot show a report with per-engine data, guarantees citations, rankings or placements in AI answers, or has no conflict policy in a competitive or regulated category.
2. Score each answer. Give 5 for the evidence asked for, 3 for a sound process without proof, and 1 for a generic or evasive answer.
3. Multiply and add. Multiply each score by its group weight and total each bidder.
4. Run a working session. Give each shortlisted agency the same page and a few buyer questions to run live, scored on the same sheet.

Our guide to AEO reporting for enterprise procurement covers what to expect after signing.

How does OnlyAEO answer these questions?

OnlyAEO answers these questions the way this page asks every bidder to, so score us on the same sheet. We do not offer a results guarantee and do not promise specific rankings, citations or placements in AI answers.

• Monthly cycle. Every month a fresh report shows where the brand is named, we create the articles, videos and listings you approve, and we review together what worked.
• Research first. We ask ChatGPT and Claude the buyer's question, read every page they cite in full, and turn each search they ran into a section, for buyer personas researched from your market.
• One brand record. Your facts, voice, buyers, competitors and the claims you can and cannot make sit in one record, the Context Graph.
• Checks before launch. Every piece passes a claim check, a humanizer pass, a link validator, an independent score of 85+ from a separate scorer, and human approval.
• Readable site. A website health check makes sure every AI engine can read the site.

How OnlyAEO works covers the full service. Our terms note that AI models change continuously and are outside our control. First measurable results usually take at least 60 days, and after that, visibility compounds as more pages, videos and mentions are picked up.

For Q21, our published case study shows a New York edibles retailer going from 1.5% AI visibility (#50) to 25% visibility, #4 of 374 brands. It is a single retailer, so weigh it that way.

Copy the full question list

Paste these questions into your RFP, then add your decision prompts, weights and deadline.

Q1. Can we see the full prompt set you would track?
Q2. Which AI engines do you track, by official API or scraping?
Q3. How often do you run each prompt, and how do you report variation?
Q4. What formula drives your headline number, and will it stay fixed?
Q5. Can you send a redacted client report with per-engine results?
Q6. Beyond traffic, what will our monthly report measure?
Q7. How do you report declines, and can we rerun your prompts?
Q8. Which AI crawlers should our robots.txt allow or block?
Q9. How do you verify AI crawlers can reach our pages?
Q10. What structured data do you recommend, and why?
Q11. Which technical changes will you implement yourselves?
Q12. How do you decide what to publish?
Q13. What research happens before a page is written?
Q14. How do you check every claim, number and link?
Q15. How do you handle regulated claims and wrong AI answers about us?
Q16. How do you keep our voice and facts consistent?
Q17. What work happens off our site, and where?
Q18. Do you guarantee results?
Q19. What happens if nothing is moving after the first 90 days?
Q20. Do gains persist if we stop working with you?
Q21. Which prompts is a current client cited for, and how has that changed?
Q22. Can we speak with a client reference in our industry?
Q23. Tell us about an engagement that went wrong and what you changed.
Q24. Who exactly will do the work?
Q25. What system access do you need, and how are credentials handled?
Q26. Do you work with our competitors, and what is your conflict policy?
Q27. How is pricing structured, and what is out of scope?
Q28. Will the contract fix report dates, a named lead, methodology and data export on exit?

Frequently asked questions

How long should agencies get to respond to an AEO RFP?

Long enough to assemble evidence such as a prompt set and a redacted report. A very short window favors pre-canned answers.

Should the RFP require a working session?

Yes, for the shortlist, with the people who would run your account in the room, not only the pitch team.

What should we do when an agency will not share its prompt set?

Disqualify it. Without the list you cannot rerun the numbers or tell whether a gain came from better work or a changed prompt set.