When choosing an AEO agency, look at who works with a clear methodology, documentation and measurable goals — not who promises the most. You buy technical competence and an accountable process, not guarantees of position or a mention. Ask them to understand your business first, explain what each recommendation is based on, measure clearly, and leave you access to the accounts. Watch for guarantees of position and opaque methods.
AEO is a field around which many new services, tools, and promises have appeared.
For a business owner, the hard part isn't finding someone who says they "do AEO".
The hard part is telling apart:
- who really understands how search and AI tools work,
- who has a technical and business base,
- who can measure what they do,
- who can document their recommendations,
- and who simply uses new terminology to sell old or vague services.
You don't need to be an expert to work this out. You need to ask the right questions.
This guide is written exactly for that. And you can apply the same criteria to Praion.
What do you really buy from an AEO agency?
You don't buy "first place in ChatGPT".
You don't buy a guaranteed mention in an AI tool's answer.
And you don't buy an organic position on Google that an agency can guarantee.
You buy a process for improving your digital presence.
Depending on the business, that can include:
- research of real searches and needs,
- content strategy,
- improving important pages,
- technical SEO,
- checking whether important pages can be found and indexed,
- sound internal linking of pages,
- technical settings such as canonical and search-engine access control where needed,
- structured data where it genuinely applies,
- local or product presence,
- measurement,
- and connecting the work to prospects, bookings, or sales.
The agency doesn't control a platform's algorithm. But it does control the quality of the work it does for you.
So the right question isn't:
"What position do you guarantee me?"
But:
"What exactly will you do, why do you think it's needed, and how will we judge whether the work has value?"
That's a much more important question.
Before they recommend AEO, do they understand your business?
This is one of the best filters.
Before an agency tells you:
"you need 30 articles"
or:
"you need an AEO audit — a full review"
or:
"we need to change all the schema"
it should first understand your business.
For example:
- What do you sell?
- To whom?
- Which products or services genuinely matter to you?
- How do customers find you today?
- What's the path from search to sale?
- Which markets or areas interest you?
- What data do you already have?
- What do you consider a real business outcome?
If they propose a solution before asking these questions, there's a problem.
Because AEO isn't a goal in itself. It's part of a wider journey:
discovery → understanding → evaluation → action → business outcome.
An agency that isn't interested in the last part can hardly decide correctly about the first.
What to ask: methodology, technical base, and real examples
Ask them to explain clearly how they work. Not only which tools they use.
1. Methodology
Ask: how they find the audience's real questions, how they decide which pages come first, how they tell a real content gap from just another chance to write an article, and how they connect content to a business priority. The answer should be explainable without fifteen technical terms.
2. Technical SEO
You don't need to know all the terms below. But your partner needs to understand them and be able to explain what problem each one solves.
An agency working seriously with Search and AEO should understand at least the basic technical topics:
- crawlability of pages,
- indexability,
- settings that control search-engine access,
- canonical,
- redirects,
- internal linking of pages,
- sitemap,
- correct rendering and access to content when the site relies on JavaScript,
- technical accessibility,
- Search Console,
- and structured data where it's genuinely useful.
And schema?
Schema is useful. But it isn't "the secret of AEO". There's no special schema that secures an appearance in AI-generated answers.
Structured data has value when it: describes real content correctly, uses the appropriate structured-data types, and supports search features where they exist.
So the right agency doesn't sell you "schema so the AI understands you". It explains where it has real use and where it doesn't.
3. Real examples
Ask to see work. Not just client logos.
Ask for: a page, an article, a review, an example of a site's organisation, a measurement system, or a real project. Then ask:
"Why did you do it this way?"
The quality of work shows far more in the reasoning behind the decision than in the final image they present.
How do they measure performance?
Measurement is perhaps where real work is most distinguished from an impressive presentation of AEO.
It isn't enough to tell you:
"we track whether ChatGPT mentions you".
Measurement should be split into levels.
Search performance
For example: impressions, clicks, searches, landing pages, organic traffic, and changes in real organic presence. Separately, the technical review checks whether Google can find and serve the important pages normally in Search.
Presence in Google's AI features
Where there's real data directly from the platform itself, it should be used.
Search Console's separate Generative AI performance reports provide first-party Google data for specific generative AI features. The Search report includes impressions from AI Overviews and AI Mode, while there is a separate report for generative AI features in Discover. Google says these insights have been rolled out worldwide since 31 August 2026. If the Search report doesn't appear for a property, the site usually hasn't received enough generative AI impressions yet or has been excluded from those features.
And one more thing has to be clear:
This measures specific Google features. It isn't an overall indicator of presence across all AI tools.
It doesn't tell you how you appear in ChatGPT, Perplexity, or the range of AI tools.
Visits from other platforms
Real traffic from platforms like ChatGPT, Perplexity, Gemini, or other services can be examined where it actually appears in the traffic data. For ChatGPT specifically, OpenAI adds utm_source=chatgpt.com to referral URLs, so clicks to the website can be identified in analytics. This measures visits generated by clicks; it isn't a complete record of every instance in which ChatGPT may have used or mentioned a source.
Business outcome
The most important part: prospects, phone calls, bookings, sign-ups, sales, revenue, or any other action that has real value for the business.
And monitoring specific questions in AI tools?
There are tools, often known as prompt trackers, that monitor how AI answers specific questions or prompts. They can be useful as exploratory tools. But care is needed.
An AI tool's answer can vary: from moment to moment, by user, by location, by phrasing, by model, or by platform.
So:
Prompt monitoring isn't equivalent to a stable ranking system.
If an agency presents a table of 20 specific AI-tool questions as if it were the equivalent of organic ranking on Google, ask them to explain exactly what it measures and what the limits of that indicator are.
How we connect AEO to business outcome overall, we develop in how to measure AEO ROI.
Ask them to explain what each technical recommendation is based on
In a field that changes this fast, this is one of the most important questions.
Ask:
"What is this recommendation based on?"
The answer can be: a platform's official documentation, real data from your own site, a controlled test, Search Console, measurement data, or documented professional experience from similar projects.
Not every decision needs to be written on an official page. But it should clearly distinguish:
what we know, what we've measured, and what is a hypothesis we want to test.
Google Search Essentials and anti-spam policies
For practices concerning Google Search, also ask:
"Is what you're proposing compatible with Google Search Essentials and the anti-spam policies?"
This matters especially when a proposal rests on: scaled content production — with AI or without it — created mainly to manipulate rankings and without substantial value for users, links created mainly to manipulate rankings, pages created mainly to lead the user elsewhere, or other aggressive techniques presented as "AEO".
The fact that a method is new or automated doesn't mean it's outside the basic rules of Search. A serious agency should be able to explain not only why it believes a practice will help, but also whether it stays compatible with the platform's rules.
This also matters when third-party AEO or SEO tools are used. A tool can be useful, but Google explicitly says third-party tools don't have access to its internal ranking data or internal AI systems. Their metrics and predictions should be treated as data from that tool, not as official ranking data.
Access and account control
There's one more criterion that's often ignored until a problem arises:
who controls the accounts and what access you give the agency.
A serious agency should ask for the access rights it needs for the specific work — and be able to explain why it needs them. There's no reason to give full admin access everywhere from day one if it isn't necessary.
Keep your business's core accounts and critical digital assets under your control, such as:
- domain and website hosting,
- Search Console,
- Analytics,
- Google Business Profile,
- CMS,
- advertising accounts,
- and other critical business systems.
The agency can have the access it needs without gaining exclusive control of these accounts.
Also ask from the start: who creates the new accounts, which company email they belong to, which access rights are needed, and what happens to access if the collaboration ends. A sound collaboration shouldn't create technical lock-in.
Warning signs: what to avoid
1. "Guaranteed first place"
"First place in ChatGPT." "Guaranteed AI citations." "Sure appearance in 30 days."
Organic appearance or mention in search results and AI tools' answers isn't something an agency can guarantee. A serious agency can tell you:
"This we control. This we don't. This we'll do. And this is how we'll measure it."
2. "Schema will get you into the AI"
Structured data is a useful SEO tool. It isn't a special road to AI tools' answers. If the whole AEO proposal revolves around schema markup, a large part of the real picture is missing.
3. Vague results reporting
"You'll see improvement."
That isn't a measurement plan. You need to know: which metrics they'll track, from which tools, the baseline, how often measurement happens, and how the results connect to a business outcome.
4. They ask you almost nothing about the business
If, before understanding your product, customer, market, margins, priorities, sales process, and which actions you consider success, they've already decided what you need, that's a serious warning sign.
5. "Secret method"
An agency doesn't need to reveal every internal process. But it should be able to explain the reasoning of the work. If everything ends in "it's our internal method" or "we can't tell you how it works", you don't have enough information to evaluate what you're buying.
6. Excessive fear
"If you don't start now, you're finished." "In six months there won't be SEO." "All your competitors are already ahead."
These aren't technical arguments. They're pressure.
7. Every third-tool metric presented as truth
A metric estimating how often a business appears in AI answers can be useful. But it remains a metric of that specific tool. It isn't an internal ranking indicator from Google, ChatGPT, or another platform and shouldn't be presented as one.
8. They propose practices that conflict with Search Essentials or anti-spam policies
If the agency bases the strategy on mass low-value content, links created to manipulate rankings, or other practices that conflict with the platform's rules, the risk is yours.
9. They ask for full access everywhere without explaining why
"Send us all the passwords" isn't sound access management. Ask specifically: which account they need, which access level, for which task, and for how long.
The right questions in the first conversation
If you want to quickly tell a serious agency apart, ask:
- "What do you need to learn about my business first, before proposing a strategy?"
- "How do you find my audience's real questions?"
- "How do you decide which content or page comes first?"
- "How do you check that the core pages can be found, indexed, and work correctly technically?"
- "When do you use structured data, and when do you consider it unnecessary?"
- "When you propose a technical change, how can I know what documentation, data, or test it's based on?"
- "Are the practices you propose compatible with Search Essentials and the anti-spam policies?"
- "What exactly will you deliver in the first months?"
- "Which first signs of progress or business indicators do you expect we'll be able to assess?"
- "What would be too early to judge?"
- "Which of the data you show me comes directly from Google or another platform, and which is estimated by third-party tools?"
- "What access do you need to our accounts, and why?"
- "Who keeps control of the accounts, data, and core digital assets?"
- "Show me a real example and explain the decisions behind it."
- "What can't you guarantee?"
The last question remains one of the best. But the first is just as important. If they aren't interested in understanding your business, it's hard to believe they can plan for it correctly.
What are the common mistakes — on your side?
The first mistake is buying the biggest promise. The impressive promise isn't necessarily the best proposal.
The second is choosing an agency only from a results table. Tools help. They don't replace strategy, technical judgement, and understanding of the business.
The third is looking only at content production. "8 articles a month" isn't a strategy on its own. The question is: why these articles?
The fourth is ignoring technical SEO. Even very good content can have problems if the important pages: aren't crawled correctly, aren't indexed, have wrong settings for which version of the page is the main one, are cut off from the site's internal linking, or have other technical obstacles.
The fifth is treating schema as proof of serious AEO knowledge. A serious agency knows how structured data is used correctly. But it also knows when not to oversell it.
The sixth is not agreeing in advance what success means. If success for you is more suitable prospects and for the agency more article impressions, there's a problem before the collaboration even starts.
The seventh is giving control of the business's core accounts to an outside partner without reason. The agency should have the access it needs. The business should keep control of its critical accounts, data, and digital assets.
How much should it cost?
There's no single right price for an AEO or SEO/AEO collaboration.
The cost depends on: the size of the website, the sector, the number of services or products, the technical state, the existing presence, content production, the measurement system, and how much real work is needed each month.
So don't evaluate only the amount. Evaluate the scope of work. Ask: how much research it includes, what the deliverables are, what technical work is included, what content will be created or improved, what will be measured, what results reports you'll receive, and what's left out.
A very cheap package that mass-produces weak content can cost more in lost time. But an expensive package isn't automatically better either.
For your own case, evaluate the actual scope of work and what the collaboration includes. The general service structure is shown on the pricing page.
And there's one more possibility: sometimes the right solution isn't working with an agency at all. Depending on the team, the available budget, and the needs, in-house, an agency, or a hybrid (mixed) model may fit. We analyse it in in-house vs agency vs AI.
How do you decide?
If you had to keep only eight criteria, keep these:
1. Do they understand the business?
They don't sell you a solution before understanding the problem.
2. Do they have a clear methodology?
They can explain what they do and why.
3. Do they have a real base in technical SEO?
They don't limit AEO to content production and schema.
4. Do they measure correctly?
They separate search data, data from Google's AI features, visits from other platforms, exploratory indicators, and real business outcome.
5. Can they document their recommendations?
They know how to separate official guidance, real data, tests, and hypotheses.
6. Do they respect Search Essentials and anti-spam policies?
They don't rebrand aggressive or manipulative techniques as "AEO innovation".
7. Do they manage access rights and account control correctly?
They ask only for the access they need and leave control of critical digital assets with the business.
8. Are they honest about the limits?
They don't guarantee things they don't control.
If an agency passes these eight filters, it's worth continuing the conversation.
Where Praion comes in
Praion doesn't treat AEO as a standalone marketing package. We place it within AI Visibility, connected to the website, the technical search foundation, content and measurement.
Apply the same criteria from this article to us. Ask us:
- what we understood about your business,
- what our methodology is,
- which technical changes we consider important,
- what we base each proposal on,
- whether the proposed practices are compatible with the platforms' rules,
- what we'll measure,
- which indicators we consider genuinely useful,
- what access we need,
- and what we can't guarantee.
We won't promise that you'll appear or be mentioned in a specific way in an AI tool's answer. But we can show you: what we found, what we propose, why we propose it, what we'll implement, and how we'll assess progress.
See our approach.
If you don't yet have a clear picture of AEO, start with what AEO is.
No serious organic AEO or SEO strategy can guarantee a specific appearance, mention, recommendation, or ranking in Google or in AI tools. Choosing the right partner rests on business understanding, methodology, technical competence, transparency, sound access management, and how they measure the real value of the work.