AEO Agency: What Answer Engine Optimization Actually Buys You

What AEO means, and why the name exists

Answer engine optimization is the work of getting your business named inside the answers that AI assistants compose, rather than in a list of links. An AEO agency sells that work. The acronym is newer than the activity, which is the source of most of the confusion around it.

The name exists because the shape of the result changed. Ten blue links accommodated ten businesses, and being eighth still put you on the page. A composed answer names three or four and stops. There is no eleventh place to occupy and no partial credit for nearly making it.

What did not change is the retrieval step underneath. An assistant asked something current does not consult a private opinion of the web. It runs a search, reads what comes back, and writes from those pages. So whatever decides which pages rank still decides which pages an assistant ever sees. The differences between SEO, GEO and AEO are real but much smaller than the separate vocabularies imply.

The four things a real engagement contains

Strip away the naming and almost every genuine engagement is some mixture of four activities. They differ enormously in cost, in speed, and in how much anyone can promise.

Measurement

A question set that matches how your buyers speak, put to several assistants on a schedule, recording who gets named. Quick to establish and genuinely necessary, because without a baseline nobody can later show that any of the rest worked. It also changes nothing by itself, and should not be priced as though it does.

Technical access

Confirming the AI crawlers can actually read your pages. What your robots.txt permits, whether your content delivery network is challenging them, whether your pages render without JavaScript. This is binary, it blocks everything else when it fails, and it is usually a few hours of work. Presented as a programme, it is a warning sign.

Content

Pages that answer the questions buyers ask, with claims specific enough to be worth quoting. This is where most billable time goes and where quality varies most. Analysis of 871 cited URLs found the pattern is not what most people assume.

Presence you do not own

Being referenced on sites that are not yours. The slowest part, the least controllable, and the one that most decides whether you get retrieved at all.

The proportions between these four are what buyers most often misjudge, usually in the same direction. Technical access feels like the serious, expert part and it is a few hours. Content feels like the commodity part and it is most of the value. Presence elsewhere feels like something you either have or do not, and it is the slowest thing to build and the strongest predictor of whether an assistant ever encounters your pages in the first place. An engagement weighted heavily toward the technical work is usually one that has run out of things to do, because that corner of the job is genuinely small once it is done.

What nobody can promise, whatever the pitch says

A supplier who will say these things plainly is worth more than one who will not.

A position. There are no positions. You are named or you are not, and the same question can produce different names on different days. Any guarantee of ranking in AI answers is describing something that does not exist.

A timeline for the slow half. Crawler access can be fixed this week. Building the authority that gets you retrieved runs in quarters, and depends partly on other people deciding to reference you.

A specific number of citations. Nobody controls how often an assistant is asked about your category, and nobody can count how many real answers named you.

Revenue attribution. Most recommendations produce no click, so the chain from answer to sale is broken at the first link. What you will be shown is a model. It may be a reasonable model, and it is still a model.

None of that makes the work unworthwhile. It makes the promises the place to concentrate when you are choosing.

How AEO agencies actually differ from SEO agencies

Less than the two names suggest, and the honest version of that answer is a good sign rather than a bad one.

Because retrieval runs on the search index, most of what improves AI visibility is work an experienced SEO team already knows how to do. Technical accessibility, topical depth, structured data, and being referenced elsewhere are the same levers under both names.

Three things genuinely are different, and a supplier who names them is telling you something real.

The measurement is new. There is no impression report, so it has to be built deliberately. An engagement with no measurement plan is an engagement that cannot be evaluated.

Extractability matters more. A page that ranks and cannot be quoted cleanly is a page that gets paraphrased without attribution. Structure, specificity and answering the question directly carry more weight than they do for a ranking alone.

The surfaces move on different clocks. Perplexity retrieves live and can change within days. ChatGPT shifts over months. AI Overviews track Google rankings. A plan that treats them as one target will read as failing on two of them for a long while.

What it costs, and what the shapes mean

Pricing in this category is unusually wide, because the label covers everything from a one-off audit to an ongoing content programme. The shape of the price tells you more than the number.

A fixed-price audit is the cheapest honest entry point. You get a picture of where you stand and a list of what to fix. Useful, finite, and it does not fix anything by itself.

A monthly retainer is the common shape for ongoing work. The question to ask is what proportion goes to content production, because that is usually where the value is and it is the line that varies most between suppliers charging the same amount.

Anything priced on citations delivered deserves care. Nobody controls how often the category is asked about, and the counting method is a choice made inside the supplier's own tooling.

The most useful question is not the total. It is how many hours go to each of the four components, and what the first ninety days are meant to produce. A supplier who cannot break that down is selling a category rather than a service. The same test applies whatever the label, and it is the one used in what AI search optimization services include and what an LLM SEO engagement contains.

When you do not need an agency at all

Three situations where hiring is the wrong move, and saying so costs nothing.

You have never checked whether crawlers can reach you. This is free, takes an afternoon, and is the single most common reason a business is invisible to assistants. If this is broken, nothing else anyone does will register. Fix it first and then decide whether you still have a problem.

You have no baseline. Start measuring before you start spending. Six weeks of a fixed question set costs an hour a week and tells you whether you have a visibility problem, a demand problem, or no problem. Agencies who begin with measurement are doing this correctly. You can also do it yourself, and it is worth doing yourself once so you understand what the numbers mean.

Nobody is asking about your category. If assistants are rarely asked the questions your business would answer, visibility inside those answers is not the constraint. That is worth ten minutes of checking before it becomes a quarter of spending.

If none of those apply and you already publish reasonable depth, then the case for help is real, and it is usually a content and authority case rather than a technical one.

There is a fourth case worth naming, because it is the one people are least willing to say out loud. Some businesses are invisible to assistants for the same reason they are invisible in ordinary search: the site is thin, the pages repeat each other, and there is nothing on it that anybody would want to quote. No amount of answer engine work fixes that, because the problem is not how the material is presented to assistants, it is that the material does not exist yet. An agency worth hiring will tell you this in the first conversation rather than the fourth invoice.

How to brief an agency so the work is checkable

Most disappointment here traces back to a brief that could not be failed. Four things make the work checkable.

Give them your question set, or ask for theirs. Everything is measured against it. If you never see it, you cannot tell what any number means.

Ask which of the four components the money buys. Measurement, technical access, content, and presence elsewhere. Hours against each. This one question separates most serious suppliers from most unserious ones.

Agree what ninety days should look like. Not a number of citations, which nobody controls. A list of things that will exist by then: the baseline running, access confirmed, a stated number of pages published, a stated number of external references pursued.

Ask what they will tell you if it is not working. The answer reveals whether the measurement is real. A supplier whose reporting cannot show a bad quarter has not built reporting.

What good reporting looks like

This is where engagements fail quietly. The work may be fine and the reporting makes it impossible to tell, so the relationship ends on a feeling rather than a finding.

It shows the question set. Every number is relative to what was asked. A report that gives you a score without the questions behind it is asking you to trust an instrument you are not allowed to inspect. You should be able to read the list and recognise your own buyers in it.

It gives a denominator. Named in eight answers is not a measurement. Named in eight of forty is. Without the second number there is no way to tell improvement from a longer test.

It separates the surfaces. Perplexity, ChatGPT and AI Overviews move on different timescales, so an average across all three hides the one thing you most want to see, which is whether the fast surface responded to the work while the slow one has not moved yet. Averaged together, real early progress looks like nothing.

It puts what changed next to what happened. The pages published, the access fixed, the references earned, on the same timeline as the visibility figures. This is the only way anyone can later argue that the work caused the result, and it is the part most reports omit.

It is capable of showing a bad quarter. Ask directly what a report would look like if nothing improved. If the answer is vague, the reporting has been built to reassure rather than to inform, and you will not find out that something is not working until you have paid for a year of it.

Whether it is worth doing at all yet

It depends on one thing more than any other, and it is not the size of your budget.

If your category is one people ask assistants about, and your competitors are being named while you are not, the gap is real and it compounds. Each answer that names three rivals and not you is a purchase decision that happened without you in the room, and none of it shows up as a lost click because there was never a click to lose.

If your category is not one people bring to assistants, or if the questions they ask are ones your business does not answer, then this is not where your next quarter should go, whatever the vocabulary around it suggests.

The honest middle case, which is most businesses, is that some of your buying questions reach assistants and more will over time. There, the sensible move is neither a large engagement nor nothing. It is to fix crawler access, establish a baseline you control, and publish real depth on the questions you can genuinely answer better than anyone else. That is unglamorous and it is what actually moves.

None of this needs an agency to begin. It needs somebody to start measuring, so that in six weeks there is something to read.