Agency RFP questions for AI search optimization (2025 guide)

TL;DR: Most agencies selling "AI SEO" can't tell you what percentage of their clients' brand mentions show up in ChatGPT, Gemini, or Perplexity, or how they'd move that number. This guide gives you 37 specific RFP questions, what a good answer sounds like, and what a red-flag answer sounds like, so you can separate real expertise from rebranded keyword work.

Why do you need a separate RFP process for AI search?

Traditional SEO RFPs ask about backlink velocity, technical audits, and rank tracking on Google. Those questions are still fine for traditional search. They're nearly useless for judging whether an agency can get your brand cited by ChatGPT, Gemini, Claude, or Perplexity.

The two disciplines share almost no measurement infrastructure. Classical SEO measures position 1 through 10 on a results page. AI search visibility measures whether a large language model includes your brand name, your product, or your point of view when a user asks a relevant question. Different signal sources. Different levers. Different reporting.

A 2024 study by Seer Interactive found that AI-generated answers in search draw from a much narrower set of cited sources than traditional organic rankings, which makes the competitive dynamic more like winning a single zero-sum slot than climbing a long tail of keywords [1]. A separate BrightEdge analysis (Q4 2024) found that roughly 41% of Google searches already returned a generative AI overview instead of a traditional blue-link list [2]. That share is climbing fast.

The agency you hire needs to understand how LLMs decide what to cite, how to measure citation frequency, and how to structure your content and authority signals to get your brand into that narrow cited set. A standard SEO audit won't tell you any of that. Your RFP has to ask directly.

You also need to vet whether the agency's claimed AI search work is real. This market is roughly 18 months old (as of mid-2025), and most "GEO" or "AEO" offerings are SEO deliverables with a new cover page. The questions below make that gap visible fast.

What credentials and track record should you ask for?

Start with provenance. Ask the agency to show you, in writing, three clients for whom they measured AI search citation frequency before and after their work. Not traffic. Not rankings. Citations, meaning the brand name appearing in a ChatGPT, Perplexity, Gemini, or Claude response when a target query is run.

If they can't show that, ask why. The honest answer might be "this is new and we have limited before/after data, but here is our methodology and here is what we've tracked so far." That's acceptable. The red flag is a vague claim that their SEO work "feeds" AI results, with no citation tracking data behind it.

Specific questions to put in your RFP:

1. "Can you share anonymized examples of AI search citation frequency reports you've produced for clients, including the methodology for how citations were sampled?"

2. "What percentage of your current client engagements include an AI visibility measurement component?"

3. "Which AI platforms do you actively monitor, and at what sampling frequency?"

4. "Have any of your team members published research, spoken at conferences, or contributed to public methodology on generative engine optimization?" (Check the answer against LinkedIn and conference records before accepting it.)

5. "Name the three tools you primarily use to measure AI citation frequency and explain why you chose them over alternatives."

The tools question is a good litmus test. Mature agencies right now use some mix of platforms built specifically for AI visibility tracking, plus their own manual query sampling protocols. If the answer is "we use Semrush and Ahrefs," those tools track traditional search rankings and were never designed to measure LLM citation frequency. You can read more about what good tracking infrastructure looks like in our overview of AI SEO tools and AI visibility tools.

6. "How do you separate correlation from causation when reporting that a content change improved AI citation rates?" This one is hard. Nobody has a clean controlled experiment. But an honest, methodologically careful answer is far more reassuring than a confident claim.

What does good AI search strategy look like and how do you evaluate an agency's approach?

A credible AI search strategy has at least four parts: entity authority building, content structure optimization, source reputation development, and ongoing citation monitoring. Ask the agency to walk you through each.

Entity authority is about making sure AI models have a clear, consistent, well-sourced understanding of who you are, what you do, and what claims you can credibly make. That means structured data (Organization, Product, FAQ, HowTo schemas), Wikipedia and Wikidata presence where applicable, and consistent brand signals across high-authority third-party sources. Ask: "How do you audit a brand's entity footprint before you start, and what signals do you prioritize for improvement?"

Content structure for AI search differs a lot from traditional SEO writing. LLMs retrieve passages, not pages. A tight answer block that directly addresses a question gets pulled into a citation more often than a 2,000-word keyword-dense article. Stanford HAI's 2024 research on retrieval-augmented generation found that models preferentially surface content matching the syntactic shape of a direct answer to a query, favoring declarative sentences over descriptive prose [3]. Ask: "Can you show us an example of content you restructured or created specifically to improve AI citation rates, and explain what structural choices you made?"

Source reputation development means building the third-party citation ecosystem that LLMs use as a trust signal. It's adjacent to traditional link building but not the same thing. LLMs pay attention to which domains discuss your brand, more than which domains link to your domain. Ask: "What is your strategy for getting a client's brand mentioned in the publications and databases that LLMs are trained on and retrieve from?"

Ongoing citation monitoring is the only way to know if any of the above is working. See the RFP questions in the measurement section below.

Red flags in strategy pitches:

• An agency that promises to "optimize for AI" purely by adding FAQ schema and calling it done. Schema helps. It's one input among many.
• Claims that "ranking #1 on Google guarantees AI citation." Google's own AI Overviews do not exclusively cite #1 ranked pages; Google's guidance indicates that AI Overview sources span many ranking positions [4].
• No mention of entity-level signals at all. If they're only talking about keywords and links, they're selling you SEO.

What to weight in an AI search agency evaluation

| | | |---|---| | Citation measurement methodology | 30% | | Strategy differentiation from SEO | 20% | | Content approach for AI retrieval | 20% | | Technical AI search knowledge | 15% | | Honesty about uncertainty | 15% |

Source: Spawned editorial framework, 2025

What specific RFP questions cover measurement and reporting?

Measurement is where most agencies fall apart, because there's no universal standard yet for AI search visibility reporting. That's your chance to stress-test them.

Here are the measurement questions worth including verbatim in your RFP:

7. "Define 'AI search visibility' as you would report it to us, including the formula or methodology."

8. "What is your sampling methodology for checking brand citations? How many queries, on which platforms, at what frequency, and with what controls for prompt variation?"

This one matters a lot. A single phrasing of a question can produce different results than a slightly different phrasing of the same question in the same LLM. Sound measurement requires prompt variation, not one canonical query. If the agency doesn't mention this, push on it.

9. "What baseline will you establish before starting work, and how long do you typically need to establish a statistically meaningful baseline?"

10. "How do you measure share of voice in AI responses, meaning how often our brand appears relative to named competitors?"

11. "Can you show us what a monthly AI search report looks like from one of your current clients?" (Redacted for confidentiality is fine.)

12. "What KPIs do you commit to reporting, and which of those are you willing to put in the contract as measurable deliverables?"

Be skeptical of any agency that commits to specific citation frequency numbers before auditing your brand's current position. Good agencies establish a baseline first, set targets against that baseline, then commit to a direction and methodology, not a guaranteed number. The space is too new for guaranteed outcomes to be credible.

For context on what mature AI search visibility reporting looks like at the KPI level, the AI search visibility metrics and KPIs framework is a useful benchmark to share with agencies and compare their proposed approach against [5].

The table below shows what an honest agency can and cannot reasonably commit to reporting, given the current state of tooling.

How does AI search optimization differ from traditional SEO, and what questions reveal whether an agency understands the difference?

Test this directly in the RFP with a few pointed questions.

13. "Explain the difference between optimizing for Google's traditional organic rankings versus optimizing for citation in a ChatGPT or Gemini response. What is the same, what is different, and which of your team's skills transfer and which do not?"

A strong answer will acknowledge that entity signals, content structure, and authority markers matter in both, then explain the specific differences: LLMs don't crawl on a fixed schedule the way Googlebot does, citation depends partly on training data and partly on real-time retrieval, and there's no equivalent to a SERP position to track incrementally.

14. "How do your content recommendations differ for a brand trying to improve AI citation versus a brand trying to improve traditional rankings?"

15. "Google AI Overviews, Bing Copilot, ChatGPT search, Perplexity, and Claude all have different retrieval architectures. How do you adjust strategy across platforms?"

This question reveals whether the agency knows these platforms are not identical. ChatGPT's browsing feature retrieves content in near-real time, while the base model's knowledge has a training cutoff. Perplexity indexes content more aggressively and shows citations more openly. Google AI Overviews pull from the Google index but apply their own weighting. A one-size approach across all four is a methodological problem.

For a deeper look at how generative engine optimization differs from classic SEO at a technical level, the concepts there map directly onto what you should expect an agency to articulate without prompting.

16. "What role does structured data and schema markup play in your AI optimization work, and which schema types do you prioritize?"

Specific schema types (FAQ, HowTo, Article, Organization, Product) are the right answer. A generic "we use schema" is not.

What questions should you ask about content strategy for AI search?

Content is the core deliverable for most AI search engagements, and the approach should look materially different from keyword-driven content marketing.

17. "Walk us through how you identify which questions or queries we should be trying to get our brand cited in. What is your query discovery methodology for AI search specifically?"

The best agencies here describe a process that analyzes conversational query patterns (more than keyword volumes), tests those queries across AI platforms to see what's currently cited, and identifies gaps where the brand could plausibly be included but isn't.

18. "How do you structure content so that an AI model is more likely to lift a passage from it? Can you show us a before/after example?"

19. "Do you write content for retrieval-augmented generation (RAG) optimization specifically, and if so, how does that differ from your standard content approach?"

RAG is the retrieval mechanism that many AI products (including some versions of ChatGPT search, Perplexity, and Google AI Overviews) use to pull live web content into their responses. Content optimized for RAG retrieval tends to be built around direct declarative answers, short factual paragraphs, and clear entity labels. If the agency doesn't know what RAG is, that's telling.

20. "How do you handle E-E-A-T signals in content written for AI search?"

Google's Search Quality Rater Guidelines define E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) as a quality framework, and Google's documentation ties it to how AI Overviews are sourced [6]. A good agency will discuss author credentials, first-hand experience signals, and sourcing practices, more than keyword optimization.

21. "What is your approach to content freshness for AI search? Does publishing date affect citation probability, and how do you manage content updates?"

Here, honest agencies will say the data is thin. Training cutoffs mean older content may simply not be in a model's training set regardless of quality. For real-time retrieval systems, recency does matter. An agency that gives you a confident universal answer here is probably oversimplifying.

What technical and on-site questions should be in the RFP?

Technical work for AI search overlaps with technical SEO but has its own priorities. These questions separate agencies that have actually worked through the technical requirements from those reusing their SEO audit checklist.

22. "How do you audit and improve a site's structured data for AI search visibility, and which validators and testing tools do you use?"

Google's Rich Results Test and Schema.org's validator are the right tools to name. The agency should also mention that structured data has to be accurate and match the page content, more than merely valid.

23. "Does your technical audit evaluate how our content is indexed by third-party AI crawlers, such as those from Anthropic or OpenAI, and how do we check robots.txt settings that may be blocking them?"

This is increasingly relevant. Several brands have accidentally blocked AI crawlers while leaving Googlebot permitted, which affects whether their content is in the retrieval pool at all. OpenAI's crawler is GPTBot; Anthropic's is ClaudeBot. A current agency will know these and know how to check for them.

24. "How do you approach internal linking and site architecture for AI search? Does page depth affect citation probability?"

25. "What is your strategy for optimizing our brand's presence on third-party platforms, like Wikipedia, Wikidata, LinkedIn, Crunchbase, and industry databases, as signals for AI entity recognition?"

This is one of the more underrated technical levers. LLMs build entity graphs from structured external sources, and a brand with an accurate, well-maintained Wikipedia article or Wikidata entry holds a real advantage in entity recognition over one without.

What questions should you ask about team, process, and who actually does the work?

Agency pitches often feature senior strategists who won't touch your account once the contract is signed. These questions surface that.

26. "Who specifically will work on our account, what are their individual backgrounds in AI search, and can we meet them before signing?"

27. "What percentage of the actual execution work is done in-house versus outsourced to contractors or white-label providers?"

White-label AI SEO is already a big business. An agency may be reselling another firm's deliverables with no real internal expertise. That's not automatically disqualifying if they're transparent and can show you the actual team, but hidden white-labeling is a problem.

28. "How do you stay current on changes to AI search systems? GPT-4o, Gemini updates, Perplexity ranking changes, and Google AI Overview changes all affect strategy. What is your process for tracking these?"

An agency that only follows the major SEO newsletters will be months behind on AI search developments. Look for mention of reading AI company research blogs, tracking model cards and release notes, monitoring changes in LLM behavior through their own sampling, and joining communities where practitioners share current observations.

29. "How large is your AI search practice relative to your total agency, and how many of your clients are currently receiving AI search services?"

Small AI search teams inside large traditional SEO agencies often mean AI search is a secondary offering with secondary resources. That's fine if they're honest about it and the deliverables are scoped accordingly. It's a problem if they pitch it as a core competency.

30. "What does your onboarding process look like, and how long from contract signing to delivery of a baseline citation audit?"

A reasonable answer is four to eight weeks for a thorough baseline across multiple platforms and a meaningful sample of queries. Faster than two weeks usually means the baseline isn't thorough. Longer than twelve weeks before any deliverable suggests process problems.

What contract, pricing, and scope questions belong in the RFP?

AI search engagements are priced inconsistently right now because the market is immature. Monthly retainers for meaningful AI search work at established agencies currently run roughly $3,000 to $15,000 per month for mid-market brands, based on publicly shared pricing from a handful of specialist firms, though the range is wide and the market is not yet transparent [7]. Project-based audits tend to run $5,000 to $25,000 depending on scope. Be skeptical of either extreme: $500/month "AI SEO" is almost certainly rebranded keyword stuffing, and $50,000/month for a startup with no existing authority base is hard to justify.

31. "Break down your proposed scope into specific deliverables with delivery timelines. What exactly are we paying for each month?"

32. "What are the contract term requirements, and what triggers would allow early termination without penalty?"

Given how new this market is, avoid getting locked into a twelve-month contract with no performance-based off-ramps unless the agency has a very strong track record.

33. "Do you offer any performance-based pricing component tied to measurable AI citation frequency improvements?"

Few agencies will fully commit to performance pricing in AI search right now, and that's defensible given the measurement complexity. But an agency willing to tie some part of its fee to measured outcomes is putting skin in the game.

34. "Who owns the content, structured data, and other deliverables you produce for us? Does ownership transfer to us on contract termination?"

35. "What reporting cadence do you commit to, and how does reporting get delivered, including dashboard access, written reports, or both?"

Tools like Spawned's AI visibility platform and others in the AI visibility tool category give brands a way to verify agency-reported numbers independently, which is worth asking whether the agency will support or resist [8].

What questions test whether an agency understands the future direction of AI search?

This market moves fast. An agency whose thinking stopped at what AI search looked like in 2023 will be behind before the engagement is finished.

36. "How do you expect AI search behavior to change over the next twelve months, and how does that affect the strategy you're recommending to us?"

There's no single correct answer, but a thoughtful one will mention the expansion of multimodal search (images and video in AI responses), the deeper integration of AI answers into Google's core search product, and the growing role of real-time retrieval versus training-data-based answers. A blank stare or a generic "AI is growing fast" is not a strategy.

37. "Are you familiar with the current research on what factors predict AI citation, and can you point us to specific studies or papers that inform your methodology?"

Published research on this is thin but growing. The Columbia Journalism Review published an analysis of news citation patterns in AI search in 2024 [9]. Microsoft Research has published on retrieval-augmented generation quality [10]. A few academic marketing journals have begun publishing on brand visibility in generative AI. An agency that cites nothing is not reading the literature.

The AI search space is genuinely evolving, and Google AI search in particular is changing fast enough that any agency should be able to name two or three specific changes from the past six months that affected their clients' visibility. If they can't, their AI search practice is not an active one.

How do you score and compare agencies after you receive their RFP responses?

Build a simple rubric before you review responses. Score each agency on five dimensions: measurement maturity, strategy depth, content methodology, technical knowledge, and honesty about uncertainty.

Measurement maturity carries the most weight because it's the line between knowing whether the work is performing and guessing. An agency with no citation frequency measurement infrastructure cannot tell you if they're doing anything at all.

Honesty about uncertainty is easy to underweight and shouldn't be. This field is genuinely new, and any agency claiming a decade of proven AI search results is lying, because that's longer than the tools have existed in their current form. Agencies that admit what they don't know, while explaining how they navigate it, make more trustworthy partners than those projecting false confidence.

A useful scoring table for comparing agencies:

| Evaluation Dimension | Weight | What a strong score looks like | |---|---|---| | Citation measurement methodology | 30% | Documented sampling process, specific tools named, platform coverage | | Strategy differentiation from SEO | 20% | Clear articulation of what's different and why | | Content approach for AI retrieval | 20% | RAG awareness, structure examples, E-E-A-T integration | | Technical AI search knowledge | 15% | Crawler awareness, schema specificity, entity signals | | Honesty about uncertainty | 15% | Acknowledges limits, shows baseline-first approach |

After scoring, have at least one finalist run a paid pilot. A six-to-eight-week paid pilot covering baseline audit plus initial recommendations, scoped at a fixed fee, lets you see actual work product and responsiveness before you commit to a longer retainer. This is standard practice in mature performance marketing procurement, and it's reasonable to ask for here too.

For brands that want to run their own AI visibility checks as a sanity layer on top of agency reporting, BrandRank.ai visibility insights analysis covers what independent monitoring looks like in practice [11]. The AI search visibility metrics and KPIs framework is also worth sharing with any finalist agency and asking them to react to it.

Sources

1. Seer Interactive, AI Search Citation Analysis, 2024
2. BrightEdge, Generative AI Research Report Q4 2024
3. Stanford HAI, Research on Retrieval-Augmented Generation, 2024
4. Google Search Central, How AI features and your website work
5. Spawned, AI search visibility metrics and KPIs framework
6. Google, Search Quality Rater Guidelines
7. Industry pricing observations from publicly available specialist agency rate cards, 2025
8. Spawned, AI visibility tool overview
9. Columbia Journalism Review, AI search news citation analysis, 2024
10. Microsoft Research, Retrieval-Augmented Generation quality research
11. Spawned, BrandRank.ai visibility insights analysis
12. OpenAI, GPTBot web crawler documentation