The choice between an AEO specialist, your existing SEO agency and building in-house should be decided by which stage is actually failing for you, not by which category sounds most current. If crawlers cannot reach your site, any competent technical SEO fixes it cheaper than a specialist will. If your problem is that nobody independent writes about you, none of the three fixes it quickly and only one of them will admit that.
Most comparisons of this decision are written by one of the three parties. This one is too, so the useful thing we can offer is the cases where we are the wrong answer, stated specifically enough to act on.
What each option is actually good at
| Option | Genuinely good at | Weak at | Typical failure |
|---|---|---|---|
| SEO agency | Crawlability, rendering, indexation, structured data | Measuring answers, third-party work | Reports rankings because that is what the tooling produces |
| AEO specialist | Diagnosis across engines, measurement design, entity work | Scale content production, and sometimes basic technical depth | Sells a programme before establishing which stage is failing |
| In-house | Product knowledge, speed of iteration, owning the baseline | Comparative context, knowing what normal looks like | Nobody owns it, so the prompt set is run twice and abandoned |
The question that actually decides it
Before comparing suppliers, work out which stage is failing, because it moves the answer more than any difference between vendors. This takes an afternoon and costs nothing.
| If the failing stage is | The right buy is | Because |
|---|---|---|
| Crawler access or rendering | Your existing SEO agency, or an engineer | It is ordinary technical work and specialists charge a premium for it |
| Entity resolution | A specialist, briefly | It is unfamiliar to most SEO practices and is a short, bounded project |
| Passage extractability | In-house editing, with a specialist setting the standard | Your writers know the product. They need the rule, not the labour |
| Third-party corroboration | Nobody, in the short term | It is a PR and product marketing problem wearing a technical costume |
| You do not know | A diagnostic only | Any programme bought before this is scoped against a guess |
Buy the diagnosis before the programme.
A diagnostic produces a named binding constraint, a baseline you keep, and a prioritised list. It also frequently ends with the answer that your existing agency can fix this, which is a cheaper outcome than the alternative.
Ours is free and you keep the output either way. Book an audit call and put us alongside two others if you want a genuine comparison.
What each one costs, and what drives it
No credible benchmark survey exists yet, so treat this as observation of the market rather than published data. The useful part is what moves the number.
An SEO agency extending into this work usually charges within its existing rate, because most of what it will do is work it already does. A specialist charges a premium justified by cross-category context and measurement design, which is real value if you need it and dead weight if your problem is a robots.txt line. In-house costs an engineer for a fortnight and then a recurring marketing commitment that is small but must not lapse.
What should not drive price is the number of pages on your site. Extractability is a property of the twenty paragraphs that matter, so per-page pricing across three hundred pages is a large invoice attached to a small effect.
Three scenarios, and what we would actually advise
Series A, thin coverage, small team. Do not hire anybody. Run the free checks, fix whatever access problem turns up, rewrite the five pages that matter, and put a recurring calendar entry on a ten-question prompt set. Your constraint is that nobody writes about you yet, and no retainer accelerates that as efficiently as your founders talking to people.
Established, good SEO, invisible in AI. This is where a specialist genuinely earns money, and usually briefly. The likely findings are an entity that does not resolve and content written as continuous argument. Both are bounded projects rather than programmes.
Well documented, quoted often, never recommended. Your problem is corroboration and possibly the product view itself. An agency can map which sources shape your category and which criticisms recur. Nobody can rewrite what the market thinks in a quarter, and any proposal implying otherwise is selling comfort.
What the buying data suggests about all three
Gartner surveyed 645 B2B buyers between August and September 2025 and found they weighed an average of seven information sources during a purchase, that 45 percent used generative AI, and that 69 percent turn to sales representatives to validate what AI told them.
Two implications for this decision. Visibility work feeds a shortlist that gets verified elsewhere, so it is upstream of your sales process rather than a replacement for it. And because buyers weigh seven sources, a supplier who only improves one of them is doing a fraction of the job regardless of which category they belong to.
Questions worth asking all three
- Which stage do you think is failing for us, and what would you look at to find out?
- What is the cheapest thing we could do without you?
- Which of these problems are you not the right supplier for?
- Who owns the measurement baseline if we stop working together?
- Show me work where the result was flat, and tell me why.
The second and third questions do most of the work. A supplier who cannot name free work you should do first, or a scenario they decline, is describing a package rather than a method.
Red flags specific to each option
Generic warnings are useless, so these are the ones that actually distinguish a good supplier of each type from a poor one.
| Option | The specific red flag | What it usually means |
|---|---|---|
| SEO agency | Describes AEO as just good SEO with no further detail | Has not looked at crawler-level differences between agents |
| SEO agency | Reports AI visibility from their existing rank tracker | The tool produces it, so it gets reported. It measures nothing here |
| AEO specialist | Proposes a programme before reading your logs | Selling a package rather than solving your problem |
| AEO specialist | A visibility score with no denominator | A number nobody can recount or dispute |
| In-house | No named owner and no recurring calendar entry | The prompt set gets run twice and quietly abandoned |
| In-house | Measurement lives in one person’s spreadsheet | It leaves when they do, and the baseline goes with it |
How to run the comparison fairly
If you are genuinely evaluating all three, the comparison only means something if you give them the same input, which almost nobody does.
Write a one-page brief containing your ten buying questions, the result of your own crawler log check, and the answer an assistant currently gives when asked what your company is. Send the identical brief to every supplier and to whoever would own this internally.
Then judge the responses on a single criterion, which is whether they tell you something you did not already know from the brief. A supplier who reflects your own document back at you with a price attached has demonstrated exactly what they will do for twelve months.
That test costs you an afternoon and it separates suppliers far more reliably than any capability deck, because it is the only part of the process where they have to think about your specific situation rather than present their general one.
The honest summary
Most companies asking this question do not yet need a specialist. They need an afternoon of free diagnostics, one technical fix, and a habit of measuring the same ten questions every month.
The minority who genuinely need a specialist usually know why by the time they ask, because they have already done the free work and the answer came back saying the constraint is entity resolution or corroboration. If you have not done that work, the most valuable thing any supplier can tell you is to go and do it first.
One practical note on timing. If you are choosing a supplier while your own diagnostics are incomplete, you are not comparing suppliers, you are comparing sales processes, and the better salesperson wins regardless of who would do better work. An afternoon of free checks changes what you are buying and frequently changes who you should buy it from. It is the cheapest thing in this entire decision and the step most often skipped, usually because somebody senior has already asked for a plan.
Whichever way you go, keep the diagnosis and the delivery separate in your own head. The party who tells you what is wrong has an interest in the answer being something they sell, and that is true of an agency, a specialist and an internal team building a case for headcount. Running the free checks yourself first is not distrust. It is the only way to evaluate the diagnosis you are given, and it takes an afternoon.
If you want the longer version of this decision, the full buyer guide has twelve questions, a scoring sheet and the contract terms worth insisting on.
Related reading
- How ChatGPT picks companies to name. The mechanics underneath, which determine which supplier can actually help.
- Indexed, retrieved, cited. The three stages this decision turns on, and how to tell which one you are failing.
Do the free diagnostics before you talk to any of the three, including us.
An afternoon of checks changes what you are buying and often changes who you should buy it from. It is the only part of this decision that costs nothing and improves every other part of it.
When you want a diagnosis rather than a proposal, book an audit call. We work the constraints in order and say plainly when the answer is that your existing agency can handle this.
Shaban runs answer engine optimisation for B2B SaaS companies at Uncited Brands. He works the unglamorous end of the problem, mostly crawler access, retrieval diagnostics and measurement baselines, and publishes the tests that do not go his way alongside the ones that do.