AEO services: the two jobs sold under one name
A buyer-side guide to answer engine optimization services. What the scope should contain, why most of the work sits on pages you do not own, what the market charges, and the four questions that separate a programme from a monthly report.
Answer engine optimization services bundle two jobs that have nothing in common except the acronym. The first is measurement: send buyer-intent prompts to AI engines on a schedule, record whether you are named and cited, and report which pages get cited instead of you. That job is mostly software and costs what software costs. The second is earning the citation: getting your brand onto the third-party pages those engines already read. That job is editorial, technical and relationship work, it takes months, and nobody can guarantee the outcome. Most retainers price the first and promise the second. Read any scope you are handed with that split in mind and most of the confusion in this market disappears.
The phrase “AEO services” is about eighteen months old and already means four different things depending on who is selling. It can mean a dashboard with a monthly call attached. It can mean a content retainer with an AI-flavoured brief. It can mean digital PR rebadged. It can mean a genuine programme that measures, acts and re-measures. All four quote in the same range, and the deck looks the same.
This page is written from the buyer’s side. It does not rank providers and it names none. It sets out what the work actually consists of, which parts are cheap and which are not, and what to check before you sign.
citeOS is our product. It is a measurement product: it probes AI engines with buyer prompts and scores what comes back. It is built and operated by Emergence Media, a Web3 growth agency, which does sell services, so we are inside the market this page describes rather than above it. Discount accordingly. Nothing on this page ranks a provider, recommends one, or quotes a citeOS tier. If you want the offer rather than the guide, it is on the answer engine optimization services page, separately and clearly labelled.
Two jobs, one label
Split every proposal you receive into these two columns before you read the price. Almost every disagreement between a buyer and an AEO provider six months in is traceable to a scope that blurred them.
| Job one: measurement | Job two: earning the citation | |
|---|---|---|
| What it produces | A dated answer to “where do we stand”, per engine, per prompt | A change in where you stand |
| What it actually is | Software. Scheduled API calls, parsing, storage, a dashboard | Editorial, technical and relationship work on pages you do not control |
| Cost driver | Number of prompts times number of engines times frequency | People, and the cost of earning a place on other people’s pages |
| Time to result | Immediate. The first report is the result | Weeks to months, and it varies by engine |
| Can it be guaranteed? | Yes. Either the report arrives on the agreed prompts or it does not | No. Inputs can be committed, outcomes cannot |
| Buy it separately? | Yes, and usually you should. Tools start around $29 a month | Only worth buying once you know what the measurement says |
The practical consequence: if a proposal is mostly job one at a job two price, you are paying agency rates for a software subscription. If it is mostly job two with no job one attached, you will have no way of telling in month six whether any of it worked. A serious engagement contains both and says which is which.
Where the work actually is
Here is the reason job two is slow and expensive, and the single most useful chart we hold. It is the source mix from our own corpus: 39,948 citation events across 72 audited crypto brands, 17,276 pages and 4,395 source domains, April to August 2026.
Your own domain is 7.78 per cent of it. The other 92.22 per cent of citations land on pages somebody else publishes. That is not an argument against fixing your own site, which is cheap, fast and necessary. It is an argument about proportion: an engagement whose deliverables are all on-site is working on the small end of the problem, however good the on-site work is.
Three things in that chart are worth pulling out because they change what a sensible scope looks like.
- The long tail is 79.09 per cent. Most citations are single pages on domains you have never heard of. You cannot buy your way onto a list of them because there is no list. You get there by being the kind of entity that gets included when someone writes about the category.
- Tier-1 press is 0.36 per cent. In our crypto dataset, a 21-outlet mid-tier basket was cited 6.34 times more often than the tier-1 names, 907 events against 143. A retainer that spends its placement budget on the two mastheads everyone recognises is optimising for the smallest slice on the board.
- Reddit is 2.19 per cent and present for 71 of the 72 brands. Small share, near-universal coverage. That combination makes it a presence problem rather than a volume problem, which is a different kind of work from publishing.
This is crypto data. The category shares will be different in SaaS or healthcare. The own-domain share being small is the part we have seen hold everywhere we have looked, and it is the part that decides how a scope should be weighted.
What a real scope contains
Five items. If any one of them is missing, ask why before you sign.
-
The prompt set, written out in full
Not “we track relevant queries”. The actual list, agreed before work starts, and fixed for the duration so the before and after are comparable. Insist that it is weighted toward category questions a buyer would type, not your brand name. A set full of “what is Acme” will flatter almost any brand, because the engine has been handed the answer inside the question.
-
The engine list, named engine by engine
“AI search” is not an engine list. Get the names, and get them per engine rather than as a total, because coverage differs sharply and several tools gate the expensive engines behind higher tiers. If your buyers research on one engine in particular, confirm in writing that it is included in the tier you are being sold.
-
A technical pass on your own site
Crawler access for the AI user agents, schema that resolves your entity, and a
sameAsblock that points at the profiles engines already read. This is the 7.78 per cent, it is the cheapest work in the programme, and it should be finished inside the first month rather than drip-fed across a year. -
A named plan for third-party presence
This is the 92.22 per cent and it is where proposals get vague. What does it commit to? Named outlets or named outlet tiers, a number of placements, a number of published pages, a plan for the directories and registries that resolve your entity. If this section is one sentence long, that is the whole review.
-
Re-measurement on the same prompts, dated
Same prompts, same engines, same method, same reporting format, on a stated cadence. Different prompts in month six is not a result, it is a new baseline. Ask for the raw per-prompt log, not just the score, because a single composite number can move for reasons that have nothing to do with the work.
What the market charges
Reported ranges, not our measurement. Firms writing about this market put startup and essentials retainers at roughly $2,000 to $4,000 a month, mid-market at $4,000 to $10,000, and enterprise at $10,000 to $30,000 or more, per Stackmatix and RevvGrowth; ZipTie gives a typical range of $5,000 to $15,000 a month. Those are third-party published figures rather than a survey we ran, and pricing pages move, so confirm at source before you quote them.
We are deliberately not leading on that number, because on its own it tells you almost nothing. Two retainers at the same price can be completely different products. The question that separates them is the split:
| If the retainer is mostly… | You are really buying | The honest alternative |
|---|---|---|
| Dashboard and a monthly call | A software subscription with a service wrapper | Licence the tool yourself and keep the difference. Entry tools start around $29 a month |
| On-site content and schema | Competent SEO with a new label on it | Worth buying, but it addresses 7.78 per cent of where citations land. Price it as the first month, not the programme |
| Placements and earned presence | The expensive job, and the one that moves the number | This is what the fee should mostly be. Ask what the placement budget is, separately from the fee |
| A guarantee of citations | A claim nobody can support | Walk. See the next section |
How long it takes, and why nobody should quote one number
Timelines differ by engine because the engines refresh differently. Reported honestly, from our own tracking:
- Perplexity re-reads the live web most aggressively of the engines we probe. A new page can show up in answers within days to a few weeks.
- ChatGPT, Gemini, Claude and Google AI Mode typically run longer, in the region of six to eight weeks before a new source starts appearing.
- Durable presence, meaning you hold the position across refreshes rather than catching one, is closer to six months.
The reason a guarantee is worthless is that the source set rotates on its own. On one prompt we tracked, ChatGPT turned over seven of its eight cited sources inside four days, with no action from us on either side. Any provider promising uniform results in thirty days is either quoting Perplexity and calling it “AI search”, or not measuring carefully enough to know.
Before you brief anyone, get the baseline. Which engines name you today, and which pages they cite instead.
Run the free scan →Four questions before you sign
These are the four that reliably separate a programme from a report. They cost nothing to ask and the answers arrive fast.
- Show me the prompt set before I pay. Not a sample, the actual list. A provider who will not show it either has not built one or knows it is brand-weighted.
- Which engines, named, and at which tier? Coverage changes by tier at almost every vendor in the tooling layer underneath, and that gate usually sits one tier above where a buyer lands. Get it in the contract, not the deck.
- What are the scoring weights? If a number is going to be reported to your board every month, you should know what moves it. Very few providers publish this. It is still the right question, because a score whose construction is private cannot be audited when it goes up.
- Show me one dated before and after with a named cause. One is enough. A percentage with no date and no mechanism is a claim, not a case study. Ask which specific action they believe moved it and how they ruled out the alternatives.
A fifth, if the engagement is crypto: ask which outlets they would place in, and why those. If the answer is the four mastheads everyone names, hold up the 6.34 times figure above and ask them to work through it.
When you do not need a service at all
Worth saying plainly, since we are inside this market. For a lot of brands the first month of any competent engagement is work you can do yourself in a week, and buying it as a retainer is the expensive way to get it.
- Your site blocks AI crawlers. Fix
robots.txt. No service required. - You have no entity schema. Publish Organization schema with a
sameAsblock pointing at the profiles that already exist for you. An afternoon. - You are missing from the registries engines read. Claim the profiles. In our corpus, SourceForge was cited for 55 of 72 brands, Slashdot for 54, CoinGecko for 55. Those are forms, not campaigns.
- Nobody has written a page about your category with you honestly in it. Write one. Being the subject of a page earned 6.22 citations on average in our corpus, against 3.78 for being ranked in a list and 3.39 for being absent.
- You have never measured. Do that first. Entry tools start around $29 a month, and several run a free scan.
Buy a service when the remaining work is genuinely earned presence on other people’s pages, because that is the part that does not compress into an afternoon and does not scale without people.
How this page was built
The source-mix chart, the 6.34 times mid-tier figure, the subject-of-page numbers and the registry counts all come from one dataset: 39,948 citation events across 72 audited crypto brands, 17,276 pages and 4,395 source domains, April to August 2026. It is our own corpus, it is crypto-only, and the category shares will differ in other industries. Method and weights are on the methodology page.
The timeline figures come from our own per-engine tracking on citeos.io and client domains over the same period, reported as ranges because that is what they are. The pricing ranges are third-party, attributed above, and not ours. The four questions are the same four we apply to tools in the AEO tools comparison, because they turn out to separate good from bad on both sides of the market.
Nothing here is modelled. Where a number is an estimate or a reported range rather than a count, it says so in the sentence.
Common questions
What are AEO services?
Paid work aimed at getting a brand named and cited by AI answer engines. The label covers two different jobs: measurement, which is scheduled probing of engines with buyer prompts, and earning the citation, which is getting onto the third-party pages those engines already read. Most contracts price the first and promise the second.
How much do AEO services cost?
Reported ranges put startup retainers at roughly $2,000 to $4,000 a month, mid-market at $4,000 to $10,000 and enterprise at $10,000 to $30,000 or more, per Stackmatix and RevvGrowth, with ZipTie giving $5,000 to $15,000 as typical. Those are third-party figures, not our survey. The useful question is what share of the fee is measurement versus earned presence.
How long does AEO take to work?
By engine. Perplexity can reflect a new page within days to weeks. ChatGPT, Gemini, Claude and Google AI Mode typically take six to eight weeks. Durable presence rather than a single hit is closer to six months. One number across all engines is a sign nobody measured.
Can an agency guarantee AI citations?
No. Nobody controls what an engine cites, and the source set rotates without anyone acting: on one prompt we tracked, ChatGPT replaced seven of its eight sources in four days. Inputs can be committed, outcomes cannot.
What should be in the scope of work?
The full prompt set, the engine list named per engine, a technical pass on your own site, a named plan for third-party presence, and dated re-measurement on the same prompts. Five items. Missing any of them, ask why.
Is AEO just SEO with a new name?
They overlap and the on-site work is close to identical. The difference is where the result is decided. Classic SEO optimises a page to rank; answer engine optimization has to account for the fact that most citations land on pages you do not own, 92.22 per cent of them in our corpus. That changes the weighting of a plan rather than the underlying craft. More on that in AEO versus SEO.
Do I need an AEO service at all?
Often not yet. Fix crawler access, publish entity schema, claim the registry profiles, get one honest category page live, and measure. Buy a service when what is left is earned presence on other people’s pages.
Related articles
AEO tools, and what each tier tracks
Six tools read on the same day, compared on engine access at the tier you would actually buy.
Sagar SaxenaComparison
AEO vs SEO: what actually changes
Where the two overlap, where they do not, and which of your existing habits still apply.
Sagar SaxenaExplainer
How AI picks its crypto sources
The corpus behind the 92.22 per cent, and what it says about where to spend.
Sagar SaxenaStudy