Vetting an AEO agency in 2026 comes down to one demand: proof of citations, not promises of them. A real agency shows you sample citation tracking reports from ChatGPT and Perplexity, names its data sources on the first call (Bing Webmaster Tools AI Performance, Google Search Console, platforms like Profound or Otterly), and prices in the market’s honest range of $2,000 to $10,000 per month. Everything else is theater.
The theater is everywhere right now. AI search became the loudest gold rush in marketing after Profound raised roughly $155 million and hit a $1 billion valuation in early 2026, and Search Engine Land’s analysis of 6.77 million sessions showed ChatGPT commanding 92 percent of trackable AI referral traffic. Every SEO shop in your inbox rebranded overnight, which means most “AEO agencies” pitching you have never earned a citation for anyone. Published pricing guides from Digital Elevator and Stackmatix agree on what legitimate work costs, but price alone will not protect you. These nine questions will. Ask them in order, and listen for the weak answer, because you will hear it a lot. If you are still building a shortlist, start with a scan of the best AEO agencies in 2026 and then put every name on it through this list.
1. Can you show me a citation tracking report from a current client?
This is the fastest filter, because an agency doing real AEO work generates these reports every month and can redact one in five minutes. No sample means no clients, no tracking, or both.
The strong answer: a redacted monthly report showing the exact prompts tested, the dates, the engines (ChatGPT, Perplexity, Gemini, Google AI Mode), and citation share against a baseline, exported from Profound, Otterly, Peec AI, or a documented in house pipeline. You should be able to type the same prompts into the same engines and roughly reproduce the results.
The weak answer: “We have a proprietary dashboard, we can show you after kickoff.” Or a single screenshot of one lucky ChatGPT answer with no date and no prompt list. If you want to understand what real tracking output looks like before the calls, review the current AI visibility tracking tools and ask which one produced their sample.
Do not walk into these calls blind. Run the free AI visibility audit first so you already know which engines cite you and your competitors before any agency frames the story for you.
2. Which prompts will you track for my business, and how did you pick them?
Prompt selection is the AEO version of keyword research, and it determines whether the whole engagement measures anything real. A weak prompt list makes any agency look good; a strong one is uncomfortable on day one.
The strong answer: a documented list of 50 to 200 prompts built from your sales calls, intake forms, People Also Ask data, and keyword research in Semrush or Ahrefs, mapped to buying stages, with a baseline showing who gets cited today on every one. For a law firm that means prompts like “best construction defect lawyer in Charleston,” not “what is a lawyer.”
The weak answer: “We monitor your brand across AI.” Brand monitoring only tells you about people who already know you. The money is in category prompts where the engine recommends someone, and a vendor who cannot explain how they choose those prompts is guessing.
3. What exactly ships in the first 90 days?
A real AEO agency can name deliverables and put weeks next to them, because the early work is concrete: audit, schema, entities, content, baseline report. Vague scope in the proposal becomes vague work in the retainer.
The strong answer: something close to this sequence. Weeks 1 to 2: baseline citation audit and prompt list. Weeks 2 to 6: schema buildout and technical fixes, including AI crawler access for GPTBot and PerplexityBot. Weeks 3 to 8: entity cleanup across directories and profiles. Weeks 4 to 12: first batch of answer shaped content and third party placements. Day 30, 60, and 90: citation reports against the baseline.
The weak answer: “Ongoing optimization tailored to your goals.” That sentence has billed millions of dollars and shipped nothing. Ask them to name five artifacts you will physically receive by day 45; agencies that do the work answer in seconds, and agencies that do not start talking about strategy.
4. How will you handle schema, and can I see markup you have shipped?
Schema is table stakes, so this question tests two failure modes at once: agencies that skip it, and agencies for whom it is the entire product. Both fail, in opposite directions.
The strong answer: JSON-LD covering Organization, Service (or LegalService and MedicalBusiness for professional niches), FAQPage, and review markup, validated in Google’s Rich Results Test, with live client URLs you can inspect yourself. Crucially, a strong agency volunteers that schema alone does not earn citations; it makes you machine readable so the rest of the work can pay off.
The weak answer: either “schema is included” with no live examples, or a pitch where schema markup is the primary deliverable. Guides from Optimist and AI Advantage Agency both flag the same pattern: if the main deliverable is markup, you are buying a slice of technical SEO wearing an AEO costume at four times the price.
5. Where does your citation data come from?
Data provenance separates measurement from invention. AI answers vary by session, location, and phrasing, so an honest agency triangulates several sources and tells you the error bars; a weak one sells you a single unverifiable number.
The strong answer: a named stack. Bing Webmaster Tools AI Performance data for how your pages surface in Copilot and Microsoft’s AI experiences. Google Search Console for AI Mode and AI Overviews era click patterns. GA4 referral segments for chatgpt.com and perplexity.ai traffic. A tracking platform, whether that is Profound, Otterly, or Peec AI, running your prompt list on a schedule. Each source has gaps, and a good agency will say so unprompted.
The weak answer: “our proprietary AI model estimates your AI impressions.” Estimated impressions with no methodology are the vanity metric of this cycle: easy to inflate, impossible to act on, and disconnected from any pipeline number you care about.
6. What does your entity cleanup process include?
Entity cleanup is the unglamorous work that makes models confident about who you are, and most rebranded shops have never done it. If the person on the call cannot define an entity problem, the engagement will never touch one.
The strong answer: an audit of every place machines learn about you: Google Knowledge Panel, Wikidata, your Google Business Profile, industry directories and review platforms (Avvo for lawyers, RealSelf for surgeons, G2 for software), and data aggregators. Then a correction pass: consistent name, address, descriptions, and categories everywhere, sameAs schema connecting your profiles, and outdated or contradictory listings fixed or removed.
The weak answer: “We build citations for you,” meaning bulk directory submissions. That is 2012 local SEO resold under a new acronym. Contradictory data is worse than sparse data for AI retrieval, and adding fifty junk listings makes the contradiction problem bigger, not smaller.
7. How will you earn mentions on the sources AI engines actually cite?
Owned content alone cannot win recommendation prompts, because when someone asks “who is the best X,” engines lean on third party sources: comparison articles, review platforms, forums, and press. An agency with no answer here can only ever get you cited for definitions.
The strong answer: engine specific source targeting. Citation studies from Semrush and Profound show each engine has its own source diet, with Perplexity leaning on review sites and community content like Reddit while ChatGPT favors established publications. A strong agency maintains those source lists for your category and runs digital PR, review generation, and expert commentary placements against them, without buying links.
The weak answer: “We will publish four optimized blog posts per month.” Content matters, but if the entire off site plan is your own blog, you will watch competitors get recommended from listicles and review threads you are not in, forever.
8. What results do you project at 90 days, and what will you refuse to guarantee?
This question is a trap in both directions, and it should be. The right answer contains a projection and a refusal, because anyone who guarantees citations is lying about how AI engines work.
The strong answer: ranges with reasoning. First citation movement in 6 to 12 weeks on lower competition prompts, meaningful share of the tracked prompt list by month four to six, and leads tagged by AI source in your CRM so visibility ties to pipeline. Paired with a flat refusal to guarantee citation counts, since the engines control their own outputs. For the measurement framework a serious agency should follow, see how to measure GEO ROI.
The weak answer: “Cited in ChatGPT within 30 days, guaranteed.” Every credible vetting guide, from Optimist to SCALZ.AI, flags citation guarantees as the loudest red flag in the category. The other weak answer is total refusal to project anything, which usually means they have no track record to project from.
9. Who owns the content, schema, accounts, and data if we part ways?
Ask this last, get it in the contract, and treat a bad answer as disqualifying no matter how good the first eight were. Offboarding terms are where weak agencies claw back control they did not earn.
The strong answer: you own everything. Content lives on your domain, schema is in your codebase, Bing Webmaster Tools and Google Search Console properties are verified under your accounts with the agency added as a user, and tracking platform history exports to you at exit. Thirty day offboarding with a full handover document.
The weak answer: content hosted on domains the agency controls, tracking data locked inside their dashboard, or accounts registered to their email. When the relationship ends, and in a market this young many will, you lose the asset you spent a year paying for. Compare their answer, and the other eight, against a written scope of services so every deliverable and ownership term is on paper before money moves.
Shortlist built and calls booked? Get the free AI visibility audit and bring your own baseline, because the agency that has to react to your data instead of presenting theirs shows you who they really are.
What else should you check before hiring an AEO agency?
How much should an AEO agency cost in 2026?
Expect $2,000 to $5,000 per month for small business scope, $5,000 to $10,000 for mid market, and $10,000 to $30,000 and up for enterprise, with one time audits at $1,500 to $5,000. Those bands come from 2026 pricing guides published by Digital Elevator, Stackmatix, and others. Quotes far below usually mean thin content only scope; far above should come with enterprise deliverables to match.
What monthly deliverables should an AEO agency send?
A citation tracking report with prompts, dates, and engine level results against baseline, shipped schema and technical changes, entity corrections completed, content published, third party placements earned, and Bing Webmaster Tools plus Google Search Console data showing AI surface trends. If the monthly report is a call with slides and no artifacts, the retainer is paying for meetings.
How long before an AEO engagement shows results?
First citation movement typically appears in 6 to 12 weeks on lower competition prompts, because retrieval based engines like Perplexity and ChatGPT search refresh from the live web. Competitive recommendation prompts, the “best lawyer in Dallas” tier, usually take four to six months of accumulated third party mentions. Anything faster than six weeks at scale is luck, not process.
Are citation guarantees ever legitimate?
No. ChatGPT, Perplexity, and Gemini control their own outputs, answers vary by session and phrasing, and no vendor controls the models. A guarantee is either ignorance of how the engines work or a bet that you will not audit the results. The legitimate version is a projection with ranges plus a performance clause tied to shipped deliverables.
Can a traditional SEO agency do AEO?
Sometimes, and the overlap in technical and content foundations genuinely helps, but hold them to the same nine questions with zero discounts. The tell is measurement: an SEO agency doing real AEO shows citation reports and prompt baselines alongside its ranking reports. One that only shows Semrush positions and calls it AI visibility has changed its vocabulary, not its work.
Do I need my own tracking tool on top of the agency?
It is cheap insurance. Otterly starts at $29 per month and Peec AI runs roughly $100 to $220, which buys you an independent check on the agency’s numbers. Good agencies welcome this; some will even provision you a seat in their Profound or Otterly account. An agency that discourages independent tracking is telling you something important.
Nine questions, one pattern: strong agencies answer with artifacts, weak agencies answer with adjectives. You do not need to be an AEO expert to run this playbook; you just need to keep asking for the sample report, the live schema, the prompt list, and the ownership clause until the adjectives run out. In 2026 the difference between the agency that gets you recommended by ChatGPT and the one that bills you for a rebrand is visible in the first thirty minutes, if you ask questions that can only be answered with proof.
Tagged