Skip to content
Learn · AEO fundamentals

What an AEO agency actually does

An AEO agency is a services firm hired to make a business more likely to be named inside AI-generated answers, and the work splits into two jobs with opposite fixes: getting your pages into the pool of sources an engine retrieves for a question, and getting selected once you are in that pool.

Vignette: a name being marked inside a generated answer. Illustration, not measurement: no figure appears in the loop.
In short

Key takeaways

  • Four different business models are sold under the one label: retained agency, platform plus a strategist, self-serve software with paid execution credits, and research-led enterprise consulting. They are priced between roughly $99 a month and figures the enterprise providers do not publish at all.
  • Being retrievable and being cited are different problems. Authority work moves the first, structural rewriting moves the second, and a single blended visibility number cannot tell you which one you are buying.
  • The most common alternative to hiring an AEO agency is not another AEO agency. It is doing nothing, or assuming the marketing agency you already pay is covering it.
  • We checked two large local marketing incumbents live on 2026-08-15. Neither mentions AEO, AI search, LLM visibility, answer engines or citations anywhere on its homepage.
  • No provider can promise a citation. Answers vary between runs on the same question, so the honest unit of delivery is a rate over repeated measurement, not a screenshot.

The work, stated concretely

Strip the category language away and an AEO engagement is two jobs stacked on top of each other. The first is getting your pages into the set of sources an engine will consider when someone asks a question in your category. That set is governed by ordinary authority signals: how you rank, whether the host looks like a real business, whether other sources corroborate what you say about yourself. The second job is getting picked once you are in the set. That is governed by structure: whether the specific sentence that answers the question exists as a clean passage, whether your page is the shape the engine tends to quote for that question, whether your claim agrees with the consensus of the other sources it is reading beside you.

The two jobs have opposite fixes and a good agency will tell you which one it diagnosed. If you are absent from the pool entirely, rewriting your paragraphs is wasted money and the work is links, entity clarity and rankings. If you are in the pool and losing the selection most of the time, buying more links is wasted money and the work is restructuring pages so the answer is extractable. A brand sitting at a middling citation rate can be in either state, and reporting one fused number without saying which is how a provider sells the expensive wrong quarter of work.

Underneath that, the task list is unglamorous and mostly recognisable: fix crawler access, clean up entity markup and third-party references so machines agree on who you are, write answer-first cornerstone and comparison pages, earn corroboration from sources the engine already trusts, then measure again against the baseline you started from. There is no AI-specific lever that bypasses any of it. What is genuinely new is the measurement layer, because the thing you are trying to move is a probabilistic output rather than a ranked list.

An answer enginenot captured
Providers in this space range from self-serve tracking software to retained agencies. Provider A publishes monthly tiers, Provider B sells a platform your team operates, and Provider C does not publish pricing.
prompt
“which agency should I hire to get my company mentioned by AI”
captured
not captured

illustrative, not a capture An answer like this sorts on what providers publish about themselves, because that is what is readable. It cannot see delivery quality, engine coverage or whether any of the numbers behind a provider's claims would survive being checked. Those are the things you have to ask for. Written to show the shape of an engine answer. It has no session, no capture date and no sample size, so it is not evidence about any category, including yours.

Four business models sold under one label

We audited six AEO and GEO providers in August 2026: aeoengine.ai, Breaking B2B, GrowthX, daydream, Graphite and CrowdReply. Three of the six were given full site inventories built from their own sitemaps. Everything in this section is what those providers publish about themselves, captured on 2026-08-15, and none of it has been verified by us beyond the fact that they say it.

The retained agency model is the familiar one. Breaking B2B is a senior-led boutique run by one operator, Sam Dunning, and it publishes a tier list of $4,000, $5,500, $10,000 and $20,000 and above per month, with a stated average client at roughly $7,000 a month, on 90-day sprints with no lock-in. Its site is built on proof density rather than surface area: 222 URLs from its sitemap, of which 137 are podcast episodes, 47 are blog posts and 19 are named case studies. aeoengine.ai runs the same lane at a much lower price, publishing $1,597 and $2,997 monthly tiers plus a custom enterprise option on 90-day rolling terms, and describes its delivery as a human strategist working alongside a large set of execution agents. Its footprint is the opposite bet: 777 URLs discovered in our crawl, including roughly 45 free tools and 381 blog posts.

The platform-plus-strategist model is GrowthX, which publishes a from-$6,000 monthly price and positions explicitly against agencies, on the argument that agency dependency is the thing the buyer should be escaping. The shape of its site tells you it means it: of 144 URLs, 65 are product documentation and 58 are long-form essays, against three blog posts and five customer stories. Your team runs the product, with a strategist in the loop.

The self-serve software model is CrowdReply, which publishes subscription tiers of $99, $299 and from $499 a month, and sells execution as per-unit credits on top: individual social comments, posts and placements bought like inventory. One of its features posts through established persona accounts, which is undisclosed paid placement, and that is a platform terms and disclosure question a buyer should ask about before it appears on their brand.

The research-led enterprise model is daydream and Graphite. Neither publishes a price, so we do not state one. Graphite's public position is a research team publishing original work on AI answers and a fixed prompt benchmark re-run each quarter; daydream's is proprietary agents paired with dedicated experts. Both are aimed at companies with a marketing team already in place.

The question that decides whether you need one at all

The realistic alternative to hiring an AEO agency is rarely a different AEO agency. It is waiting, on the reasonable belief that AI answers are not yet deciding anything in your category. Or it is assuming the marketing agency you already pay has this covered, because they now use the word AI in their deck.

That second assumption is checkable in about four minutes, and we checked it. On 2026-08-15 we fetched the homepages of two large marketing incumbents serving local service businesses. Blue Corona's homepage lists roughly eighteen services, including web design, hosting, paid media, SEO, Google Business Profile, call answering, call tracking and a reporting dashboard it describes as carrying more than 300 KPIs, which is their figure, not ours. It mentions AEO, AI search, LLM visibility, answer engines, ChatGPT and citations nowhere. Its only AI is a chat widget, which is a conversion tool. Scorpion's homepage splits its offering into technology, advertising and marketing, with no AI-search offering on it.

So there is a large population of businesses already spending real money on marketing every month, already measurement-literate, already convinced organic discovery is worth paying for, and with no measurement of AI visibility at all, because the provider they pay does not sell it. Whether that gap matters yet depends entirely on whether AI answers name businesses in your category. In some categories they do not, and if that is your situation the correct purchase is nothing.

The first useful question is therefore not which agency. It is whether an engine, asked the question your buyers actually ask, names companies at all, and if so which ones. That is a measurement, it takes a few hours of instrument time, and any provider unwilling to answer it before you sign is telling you something.

What an AEO agency cannot do for you

It cannot promise a citation. Ask an engine the same question twice and it can name a different set of companies, so any honest commitment is expressed as a rate over repeated runs, not as an outcome on a given day. A provider showing you a screenshot of one answer is showing you one sample of a noisy process.

It cannot prove you are absent. Our instruments can prove an AI Overview was present for a query. When one comes back empty, that reading is consistent with a genuine absence, an unexpanded asynchronous block and a rate limit, and those are indistinguishable in the output. Absence is a hypothesis that needs more than one instrument before it becomes a finding.

It cannot cover every engine, and the ones that say they do are describing a wish. We measure ChatGPT and Gemini through a vendor that renders those consumer surfaces, and Google AI Overviews through search results with the overview expanded. We do not measure Claude at all, because no ground-truth path to the consumer product exists for it and the model-level API is a different surface that no buyer uses. We do not measure Perplexity either, because there is no compliant path to its consumer product and its API answers as a model rather than as the product a buyer uses. Those gaps are real for everyone working in this category, whether or not they publish them.

And it cannot turn a citation into revenue on its own. The chain runs from citation to demand to revenue, and the last step belongs to your conversion layer. An engagement that moves your citation rate into a site that cannot convert the resulting visits has moved a metric and nothing else. That is worth knowing before you buy, not after.

Where Caldrin sits, and who we are wrong for

We are a senior-led practice that measures with our own instrument and publishes the gaps in it. The front door is a free AI visibility report, anchored at a struck-through $1,500 and delivered on a call rather than as an auto-generated PDF; the work after it is scoped per engagement rather than sold from a tier list, because two businesses with the same revenue rarely have the same problem. Five of the six providers we audited in August 2026 gate or charge for the initial assessment; CrowdReply is the exception, and only because it is self-serve software with no assessment to gate.

The clearest thing we can show you today is a before-picture rather than a win. For 603 Basement Solutions, a client who has cleared us to use their name, we ran a 488-point local grid across 8 service keywords and 61 towns in New Hampshire, Massachusetts and Maine. On 2026-08-09 they held a local pack position at 10 of those 488 points, which is 2.0 percent. One competitor held 39.3 percent of the same grid. That is a diagnosis with its denominator attached, and it is the shape of what a first engagement produces.

Now the part most provider pages leave out. We have no revenue-attributed win yet. We hold every client citation share figure we have measured, because they are single-sample readings with roughly 25 points of margin and our own rule is that nothing travels until it has run at least three times. We have no B2B SaaS case studies. We publish no guarantee, while aeoengine and Breaking B2B both publish one. We do not have the procurement furniture a large buyer expects; GrowthX publishes seven legal pages including an MSA, a DPA and a subprocessor list, and we do not. And we will not run undisclosed persona posting, which removes one of the faster ways to manufacture the appearance of being recommended.

If you need a vendor who will hand you a case study from your exact vertical with revenue attached, we are the wrong call and you should say so on the first one. If you want to know what is actually true about your visibility before anyone sells you a plan, that is the part we are built for, and it costs nothing to find out.

Questions

Questions, answered plainly.

What does an AEO agency actually do?

Two separable jobs. First, getting your pages into the set of sources an engine considers for a question, which is moved by rankings, links, entity clarity and third-party corroboration. Second, getting selected once you are in that set, which is moved by page structure and whether the answering sentence exists as a clean, quotable passage. A good engagement says which of the two it diagnosed, because the fixes are different and a fused visibility number hides the distinction.

How much does an AEO agency cost?

It depends which model you buy, and the published range is wide. Among the providers we audited in August 2026, the self-serve software tiers start at $99 a month, the lower-priced managed tiers were published at $1,597 and $2,997 a month, and the senior-led B2B boutique published $5,500 to $40,000 and above per month across two tier sets, with narrower-scope plans from $4,000. Two of the six publish no price at all. Those are the providers' own published figures, captured on 2026-08-15, and they change.

Can an AEO agency guarantee I will be cited by ChatGPT?

No, and a guarantee is a signal to slow down. The same question asked twice can return a different set of company names, so citation is a rate over repeated runs rather than a state you achieve. Any honest commitment is expressed as a measured rate with the number of runs behind it, over a stated window.

My marketing agency says they already handle AI search. Do they?

Check the homepage rather than the sales call. We fetched two large local marketing incumbents live on 2026-08-15. Blue Corona lists roughly eighteen services and mentions AEO, AI search, LLM visibility, answer engines, ChatGPT and citations nowhere; its only AI is a chat widget. Scorpion's homepage splits into technology, advertising and marketing with no AI-search offering. Neither is a bad agency. Neither sells this.

Which AI engines can actually be measured?

Fewer than most coverage claims suggest. We measure ChatGPT and Gemini through a vendor that renders those consumer surfaces, and Google AI Overviews through search results with the overview expanded. We do not measure Claude, because no ground-truth path to the consumer product exists and a model-level API is a different surface. Perplexity is not covered either, because it has no scraper and automating the consumer site is disallowed in its robots.txt. Any provider claiming full coverage of every engine is describing an ambition.

Method

What is measured, and what this page is not.

This is an explainer. It carries no figures, and it is not a reading of your category. The disclosure below states the instrument that produces the numbers the essay refers to, so the distinction is on the page rather than assumed.

instrument
Caul
what was measured
Nothing on this page. Where the essay refers to citation share, that figure is produced separately, per account.
how
A prompt set written once for a category and then frozen, run against every engine in clean sessions, with each answer stored unmodified.
over what window
Reviewed on 2026-08-16. The engines change, so read the essay against the date on the byline.
what this cannot tell you
An explainer is not evidence about your category. Being named is not being recommended, and it is not traffic or revenue. Any figure about your own visibility has to come from a capture of your own category, carrying its sample size and its window.
Keep reading

The rest of the cluster.

All Learn guides · Glossary

See where you stand.

The audit is a real sweep of your category, benchmarked against competitors you name, delivered on a call so the findings get explained rather than emailed. You keep the report and the underlying data whatever you decide afterwards.

Get your free auditWhat is in the report