Choosing an AEO agency comes down to one test: can they open a laptop, run a query you picked, and show you an AI answer that names a client, then explain what they changed to earn it? Everything else, the framework diagram, the tool logos, the 40-slide deck, is downstream of that. The 12 questions below are ordered so the cheapest disqualifiers come first.
The market makes this harder than it should be. Answer engine optimization in 2026 sits roughly where SEO sat in 2005: every agency claims the capability, and a much smaller number have built it. Our own breakdown of what an AEO agency actually delivers covers the full scope; this guide is about vetting whoever you're about to sign with. Your job in the first call is to find out which one you're talking to, in under thirty minutes.
What does an AEO agency actually do?
An AEO agency works to get your brand cited inside AI-generated answers (ChatGPT, Perplexity, Google AI Overviews, Claude, Gemini, Copilot) rather than ranked as a blue link. The work splits across four workstreams: entity trust, on-site answer assets, off-site corpus seeding, and citation measurement. An agency that only does the second one is an SEO agency with a new service page. If you want the complete map of how those workstreams fit together, we've laid it out in the AEO Master System, the 10-pillar answer engine stack we run engagements against.
The commercial case rests on two numbers. Ahrefs' most recent measurement found that the presence of an AI Overview now correlates with a 58% lower average click-through rate for the top-ranking page, up from 34.5% when it first ran the study eight months earlier. The traffic is being absorbed at the top of the SERP. Meanwhile Seer Interactive's case study on ChatGPT referral traffic recorded visitors arriving from ChatGPT converting at 15.9% against 1.76% for Google organic. Less traffic, better traffic.
| Workstream | What it covers | Who usually skips it |
|---|---|---|
| Entity trust | Knowledge graph presence, Wikidata, consistent profiles, 30+ corroborated mentions | Content-only shops |
| Answer assets | Answer-first pages, question-shaped headings, schema, extractable chunks | Nobody, this is the commodity layer |
| Corpus seeding | Reddit, G2/Capterra, third-party "best X" listicles, YouTube, PR | Almost everyone |
| Measurement | Prompt set, share of model, citation tracking, sentiment | Agencies reporting impressions |
Which 12 questions should you ask an AEO agency?
Ask these 12 questions in four blocks, in this order: proof, method, measurement, commercials. Proof first is deliberate. Roughly a third of agencies will fail question one, and there is no reason to spend forty minutes on methodology with a vendor who cannot demonstrate a single citation. Send the list in writing before the call so the answers are prepared rather than improvised.
Proof (questions 1 to 3)
1. Can you run a query I choose, right now, and show me a client named in the answer? Good answer: they screen-share, run it live across two engines, and the client appears. Red flag: a case study PDF, a screenshot, or a request to "send that over after the call." Live beats archived because AI answers are probabilistic and screenshots age badly.
2. What was citation rate at month one versus month six, on which prompt set? Good answer: a dashboard with a named prompt set, a start figure, an end figure, and an explanation of what changed technically. Red flag: "improved AI presence," "significant lift," or any sentence describing a result without a number attached to it.
3. Which client can I call, and were they cited across multiple engines or just one? Good answer: a reference, offered without hesitation, plus honesty about which engines worked and which did not. Single-engine wins are common and fine. ChatGPT, Perplexity, and Google AI Overviews use different retrieval logic, a difference we walk through in AEO vs GEO vs SEO. Claiming all six went well is the red flag.
Method (questions 4 to 7)
4. What percentage of this engagement happens off our website? This is the question that separates the field, and almost nobody asks it. Citations depend heavily on third-party corroboration: Reddit threads, review platforms, listicles, editorial mentions, because models cannot verify truth and default to counting agreement across trusted sources. That's the same dynamic behind why G2 reviews influence what ChatGPT recommends. An agency proposing 100% on-site work is selling you schema and calling it AEO.
5. How does your approach differ across ChatGPT, Perplexity, and Google AI Overviews? Good answer: specifics. Perplexity weights freshness heavily and cites densely. Google AI Overviews remain partly tied to organic ranking, though loosening fast: Ahrefs found 76% of AI Overview citations also ranked top 10 in mid-2025, falling to around 38% less than a year later. Red flag: one undifferentiated "AI optimization" process.
6. What is your entity foundation checklist, and where are we failing it today? Good answer: they have already looked. They can name your knowledge panel status, your sameAs inconsistencies, your review-site gaps, and which schema types actually get cited on your pages today. Red flag: entity work described as something to "get to in phase two." If models cannot resolve who you are, nothing downstream gets cited.
7. Who writes the content, and how does first-hand expertise get into it? Good answer: a named process for extracting your team's expertise: customer interviews, subject-matter expert calls, original data. Red flag: an offshore content pool and a word-count deliverable. The Princeton GEO study, presented at ACM KDD 2024, tested nine content tactics against 10,000 queries and found citing sources, adding statistics, and quoting named experts among the strongest levers, lifting AI citation rates by up to 40% in the paper's own benchmark. Generic content captures almost none of that.
Measurement (questions 8 to 10)
8. What is the prompt set, who chooses it, and how often is it re-run? Good answer: you approve the prompts, they run on a fixed schedule, and raw data comes to you, the same discipline we recommend when you're tracking your own brand mentions in ChatGPT. Red flag: the agency picks the prompts privately. A self-selected prompt set is a scoreboard the vendor controls.
9. Which numbers in your report are inputs you control, and which are outcomes you do not? Good answer: they draw the line themselves. Pages published, schema deployed, mentions placed: inputs. Citation frequency, share of model, sentiment: outcomes influenced by model updates outside anyone's control. Red flag: an agency that guarantees outcomes has either misunderstood the medium or is planning to report on inputs only.
10. How do you connect citations to pipeline? Good answer: AI referral traffic segmented in analytics, self-reported attribution on demo forms, assisted-conversion tracking. Seer's study across 3,119 queries and 42 organizations found brands cited in an AI Overview saw 35% more organic clicks and 91% more paid clicks than when uncited. The effect is measurable if someone bothers to measure it.
Commercials (questions 11 and 12)
11. What does month six look like? Good answer: a concrete description of steady-state work: prompt monitoring, corpus refresh, content updates, per-engine gap plugging. Red flag: an engagement that is 80% audit in month one and vague afterward. Expect 60 to 90 days for initial citation movement and three to six months for meaningful share-of-model growth.
12. If we leave, what do we own? Good answer: content, schema, prompt set, historical tracking data, and any third-party profiles created on your behalf, all yours, exportable. Red flag: schema injected through the agency's own tag manager, or tracking that lives in an account you cannot access. Get this in the contract, not the call.
How should you score the answers?
Score the four blocks by weight, not by count. Proof carries 40% because it is the only unfakeable signal; commercials carry 10% because contract terms are negotiable and expertise is not. An agency that aces method and measurement but cannot produce a live citation has a good process and no track record. That's a discount, not a disqualification, but price it accordingly.
| Block | Questions | Weight | Pass condition |
|---|---|---|---|
| Proof | 1 to 3 | 40% | One live, unrehearsed citation on a query you chose |
| Method | 4 to 7 | 25% | Named off-site plan and real per-engine differences |
| Measurement | 8 to 10 | 25% | Written prompt set plus an explicit input/outcome split |
| Commercials | 11 to 12 | 10% | Concrete month-six scope and full asset ownership |
What should an AEO agency cost in 2026?
Serious AEO retainers run roughly $3,000 to $15,000 per month in 2026, with growth-stage engagements clustering at $3,000 to $8,000 and standalone audits at $1,500 to $5,000. Below about $1,500 per month you are almost certainly buying rebranded SEO. If your budget sits below that floor, our AEO checklist for founders walks through the 40 fixes worth doing yourself before you pay anyone.
| Engagement type | 2026 range | What it should cover |
|---|---|---|
| Self-serve tracking tool | $29 to $489/mo | Monitoring only, no execution |
| One-time audit | $1,500 to $5,000 | Baseline share of model, prompt set, entity and technical gaps |
| Growth-stage retainer | $3,000 to $8,000/mo | On-site assets, schema, corpus seeding, monthly citation reporting |
| Enterprise / multi-market | $15,000+/mo | Multi-language scope, heavy off-site authority work |
Two framing notes. First, the premium over classic SEO is real: Ahrefs' survey of 439 agencies found $500 to $1,000 the most common single monthly SEO retainer band, well below where AEO work starts. Second, published ranges span $1,500 to $50,000, which means the number tells you almost nothing until you know how much off-site work is inside it. Ask what changes if the budget moves 30% either way; the answer reveals the real scope.
When should you not hire an AEO agency?
Skip the agency in three situations. If your site is not crawlable or indexed in both Google and Bing, retrieval cannot happen and you are paying a specialist to fix a technical problem. If your positioning is unclear internally, no agency can make a model articulate a differentiation you have not articulated yourself. And if your budget caps below $1,500 per month, a self-serve tracking tool, like the options compared in our AI visibility tools guide, plus your own execution will outperform a thin retainer.
There is a fourth case worth naming honestly: if you already have a strong SEO program and an in-house content team, buying an audit and implementing it yourself is often the better trade, and it's worth reading how that trade-off plays out in practice in our in-house vs GEO agency guide. Answer engine optimization is not a separate department. It is the same infrastructure, aimed at a different surface, which is exactly why an agency that treats AEO as unrelated to SEO should worry you as much as one that treats it as an SEO add-on.
If your product already sits in front of AI-savvy buyers but keeps losing the recommendation to a competitor, that's a narrower, sharper problem than "we need an AEO agency," and it's worth diagnosing directly: see why ChatGPT isn't recommending your product before you scope a full engagement.
For everyone else weighing this decision by region or budget specifically, our AEO agency shortlist for Australian SaaS founders applies this same scorecard to named vendors, and GEO agency pricing broken down by model goes deeper on the retainer-versus-project question above.
A baseline before you sign anything
Don't buy activity. Buy outcomes and a system.
A useful AEO engagement should move through something like: baseline audit, prompt research, competitor reverse engineering, entity foundation, content and citation strategy, execution, measurement, reinforcement. That mirrors the broader SEO/AEO/GEO operating model: establish trust, understand demand, reverse-engineer what the engines currently reward, create extractable assets, build external corroboration, and continuously measure what gets surfaced.
The goal isn't to "rank #1 in ChatGPT." There is no universal position #1 in generative search. The goal is to become a trusted, relevant, repeatedly surfaced entity when your ideal customer asks questions your business can solve.
Before you take any of this to a vendor, it's worth seeing where you actually stand. Run the free AEO and GEO audit to get an 80-point citation-readiness score, then use questions 1 through 12 above to find out whether the agency in front of you can move that score, or just talk about it. If you'd rather walk through the results with someone directly, book a 30-minute scoping call and we'll run your category's real buyer prompts live.
FAQs
What is the difference between an AEO agency and a GEO agency?+–
Most agencies use the terms interchangeably, and the deliverables overlap almost entirely. Where a distinction is drawn, AEO targets being the cited source in any answer surface, while GEO targets how generative models chunk, retrieve, and compose passages. Judge the work, not the label.
How long before an AEO engagement shows results?+–
Expect 60 to 90 days for initial citation movement and three to six months for meaningful share-of-model growth. Schema and entity changes can register within two to four weeks of reindexing. Any agency promising faster is overselling; one that cannot give a timeline has not run enough campaigns.
Can we do AEO in-house instead of hiring an agency?+–
Yes, if you have technical resource for schema, a writer who can produce answer-first content, and someone willing to run a fixed prompt set monthly. The constraint is usually off-site corpus work: review platforms, listicle placements, and community presence, which is slow and relationship-driven.
Should we buy SEO and AEO from the same team?+–
Usually yes. The content, schema, and authority work overlap heavily, so separate vendors bill twice for the same tasks and argue over attribution. The exception is when your incumbent SEO agency fails the first question in this guide: showing you a live citation.
