How Do You Choose an AEO Agency in 2026?

If you are shopping for an AEO agency, you have probably already opened four tabs of ranked lists and noticed they disagree about everything except their own author. That is worth two minutes before you book any call.

We sell the service on this page, so read the next section with that in mind. The check we ran is one anybody can repeat in ten minutes, and the numbers underneath it are ours.

What does an AEO agency do?

An AEO agency gets your brand named inside AI answers from ChatGPT, Google AI Overviews, AI Mode, Perplexity and Gemini. It measures how often each engine cites you against named competitors, rebuilds pages into passages a model can lift, earns mentions on the third-party sources those engines pull from, and re-runs the measurement weekly.

That is the whole job. Everything else on an agency site is a description of how they staff it.

Why every AEO agency shortlist names its own publisher first

We read the ranked lists sitting on page one for this query on 30 August 2026. There is a pattern in them, and it is not hidden so much as unmentioned.

The three lists on page one all place their publisher in the top slot

Minuttia's list of ten, published in December 2025 and updated on 5 August 2026, opens with Minuttia. NoGood's ranking, published 25 April 2026, scores NoGood at 96.3 out of 100 and puts it first. Pepper's list of eleven, published 26 June 2026, leads with Pepper at $8,500 a month.

Three lists, three publishers, three top slots. None of the three flags the placement.

The ranking you are reading was written by a company that appears in it.

This is not fraud and it is not unusual. Ranked lists are a content format, agencies are the only people motivated to build them, and the format has no referee. It is still the most important fact about the page you are standing on.

The scores are unaudited and the criteria cannot be failed

NoGood's 96.3 is a real number built from six criteria including thought leadership, reputation and fit for the future. Nothing about those is measurable by a reader, and no agency has ever scored badly on its own rubric.

The client outcomes carry the same problem. "358% increase in AI Overview appearances in five months" is a real claim from a real list, with no baseline, no prompt set and no engine named. A percentage without a denominator is a shape, not a result.

AI engines read these lists, which compounds the bias

Here is the part that matters for the thing you are actually buying. When a buyer asks ChatGPT which AEO agency is best, these are among the pages it reads. The self-placement does not stop at the page, it propagates into the answer.

We wrote about this failure mode in why self-promotional listicles underperform in AI citations. The short version is that engines reward the list format and cannot audit the ranking inside it.

What the shortlist measures:

• •How many agencies the publisher could describe
• •Which of them have a public case study
• •How the publisher scores against its own criteria

What your decision actually turns on:

• •Whether anyone has measured your citation share this week
• •Which engines your buyers use, and whether those engines cite sources at all
• •How much of your category's source pool sits on domains you do not own
• •Who runs the loop in month seven, when it stops being interesting

Every list answers the first set. The second set decides whether the retainer does anything.

5 things an AEO agency can move

These are the levers that exist. A provider who cannot describe all five as concrete artifacts is selling an SEO retainer with a new word on the cover.

Lever #1: A model can only quote a passage it can cleanly extract

Models lift self-contained passages of roughly 40 to 60 words, not whole pages. Rebuilding priority pages into answer blocks aimed at specific buyer prompts is the highest-yield on-page work available, and we cover the mechanics in why passages beat pages.

Lever #2: Most of your citation pool is not your own domain

In our corpus, four of the twelve most-cited domains were brand-owned sites. The other eight were places you have to earn your way into. Reddit alone drew 14,698 citations and appeared in 13.6% of all answers.

Most of your citation pool sits on domains you do not own and cannot edit.

An agency that only touches your own site is working a third of the problem and billing for all of it.

Lever #3: The engines cite sources at very different rates

Our 63-day study found Google AI Mode cited a source in 97.4% of its answers, ChatGPT with web search in 92.5%, and Gemini in 79.1%. Those gaps decide where a citation is even available to win.

An agency that reports one blended "AI visibility" score is flattening three answers that disagree. Ask which engines its measurement actually covers, by name.

Lever #4: A model can name you and still describe you wrong

If an engine calls you a budget tool when you sell enterprise, that framing is now part of your pitch on every call you are not in. Correcting it is entity and source work, and publishing more posts does not do it.

Lever #5: The weekly loop is where in-house programs die

The work is measurement, passage repair, off-page placement and re-measurement, run every week. Teams lose this on cadence rather than on strategy, usually in the first quarter that gets busy. We compare the two paths in GEO in-house versus agency.

2 things no AEO agency can move

The honest constraints are missing from every list we read, and they are the reason engagements get judged unfairly in both directions.

Constraint #1: An incumbent gives up a category slowly, or never

Our concluded CITE Index study ran 500 buyer prompts nightly through ChatGPT, Gemini and Google AI Mode for 63 days, collecting 90,132 answers across ten categories.

Across that window the category leader changed on only 18.7% of day pairs, and in four of the ten categories it never changed once. The average leader held 78.2% of its own category's answers on the days it led. The full corpus sits in the final report.

Displacement is slow and absence is permanent. Those are different problems and they cost different amounts.

If a competitor owns your category's answers today, an agency can put you in the answer alongside them long before it takes the top slot off them. Any pitch that promises otherwise inside a quarter is arguing with the data.

Constraint #2: Nobody can guarantee you a citation count

Nobody controls what a model retrieves on a given week. Sielinski's convergence work, a July 2026 preprint, found rankings needed between 33 and 94 answers to settle across 30 platform-topic combinations, and three never settled at all.

A guarantee on this surface is either a misunderstanding or a sales tactic. The honest commitment is a prompt set, an engine list, a threshold and a re-measurement cadence, all in writing.

What an AEO agency costs in 2026

Published ranges cluster tightly once you strip the labels. 310 Creative's August 2026 pricing breakdown and the tiers inside Pepper's list land in roughly the same place.

Band	Monthly	What it realistically buys
Productized	$500 to $2,500	Schema, some content restructuring, monitoring on one engine. Usually Google only.
Mid-market managed	$4,000 to $12,000	Multi-engine measurement, ongoing passage work, structured data at scale, citation tracking.
Enterprise	$15,000 and up	Dedicated team, digital PR for off-page citations, entity architecture, pipeline attribution.

310 Creative's own warning is the useful one: a quote under $1,500 that claims full AEO coverage is usually traditional SEO with the label changed. Our own read of the delivery models sits in the AI visibility pricing guide.

The number inside the managed band tracks two things and no others: how many prompts and engines you watch, and how much earned-media work your category needs. If a quote moves and neither of those moved, ask what changed.

Build your own AEO agency shortlist in 5 steps

The diagnostic half is done. This is the sequence we give buyers evaluating anyone in this category, including us.

Step 1: Run your ten highest-intent prompts by hand first

Open ChatGPT, Gemini, Google AI Mode, Perplexity and Copilot and ask the same ten shortlist-stage questions your buyers ask. Record where you appear, which competitor appears instead, and every domain cited.

Two hours of this tells you which engines carry your category and whether your problem is absence or displacement. Walk into every call already knowing.

Step 2: Ask each agency for a redacted citation baseline from past work

A firm that measures citations for a living has baselines on hand. Ask to see one with the client name removed.

If what comes back is a rankings export or a traffic chart, they are measuring the old surface. This question ends more conversations than any other on the list.

Step 3: Make them name the third-party sources they would target for you

Because most of your citation pool is not your domain, a provider who only edits your site is quoting for a fraction of the job. A real operator names the communities, review platforms and publications that feed answers in your category.

Vagueness here is the most common tell in the category. "We will build authority" means nobody looked at your source pool.

Step 4: Get the promise written as a prompt set, not a percentage

The commitment that can be held to is specific: these 25 prompts, these four engines, this citation-share threshold, re-measured weekly, reviewed at day 90. A percentage lift with no denominator cannot be failed or passed.

The vetting questions we use for the adjacent GEO label transfer cleanly and sit in how to vet a GEO agency.

Step 5: Check who wrote every list you used to build the shortlist

Go back to the pages that produced your candidate names and find the publisher. If the publisher is on the list, keep the names and drop the ordering.

The names on those lists are mostly real firms. The ranking is the part that was authored by a competitor.

Where an AEO agency fits against the alternatives

Option	What it constrains	Right when
Self-serve tracking tool	Fixes nothing. You get a reading and a queue you still have to work	Someone in-house owns the weekly loop and only needs instrumentation.
Consultant or fixed project	Ends when the invoice clears, before the first drift	You need a baseline and a plan, and you have a team to execute it.
Managed AEO agency	Costs a retainer. Cannot promise a citation count	Nobody owns the loop, and your source pool sits mostly off your domain.
Your existing SEO agency	Usually one engine and a rankings frame	They can show you a citation baseline. Ask before you assume.

The service scope under the neighbouring label is broken down in answer engine optimization services, and the difference between the two acronyms is mostly naming, which we settle in AEO versus GEO. If the internal loop is what you are missing rather than the reading, a managed AEO service is the thing that fills it.

FAQ

What does an AEO agency do?

An AEO agency gets your brand cited inside AI answers from ChatGPT, Google AI Overviews, AI Mode, Perplexity and Gemini. The work is a loop rather than a project: measure citation share against named competitors, map the buyer prompts that decide deals, rebuild pages into 40 to 60 word passages a model can quote, deploy schema and entity fixes, earn mentions on the third-party sources engines pull from, then re-measure weekly and fix what drifted.

How much does an AEO agency cost?

Published 2026 ranges run from roughly $500 to $2,500 a month for productized work covering schema and single-engine monitoring, $4,000 to $12,000 for a mid-market managed program with multi-engine measurement and ongoing content work, and $15,000 and up for enterprise engagements that include digital PR and attribution. Quotes below about $1,500 claiming full coverage are usually traditional SEO relabelled.

What is the difference between an AEO agency and a GEO agency?

In practice, none. AEO agency, GEO agency, AI SEO agency and generative engine optimization agency all describe getting your brand named by AI answer engines. The acronyms came from different corners of the industry settling on different words in the same eighteen months. Judge the firm on the loop it runs and the measurement it can show you, not the label on its homepage.

Are AEO services worth it for a small company?

They are worth it when you can name a specific gap and cannot close it internally. If your buyers use one engine, your source pool is small and someone on your team can run a weekly prompt test, a tracking tool and internal discipline will get you most of the way. If eight of the twelve domains feeding your category's answers are ones you cannot edit, the off-page half is the work, and that is the half in-house teams almost never sustain.

Do I need an AEO consultant or an AEO agency?

A consultant is the right buy when you need a baseline, a prompt set and a plan your own team will execute. An agency is the right buy when nobody in-house will own the weekly loop after the first busy quarter. The failure mode for consultants is a good plan that never ships. The failure mode for agencies is paying a retainer for a loop you could have run yourself.

The bottom line

The AEO agency market is two years old and it has no referee, so the firms compete by publishing rankings of themselves. Every list on page one for this query does it, and none of them says so.

Keep the names and throw away the order. Then go get the two facts no list contains: whether your problem is absence or displacement, and how much of your category's source pool sits on domains you will never own.

Run your ten highest-intent prompts across five engines this afternoon. That single exercise tells you more about what to buy than every ranked list in the category, and it costs two hours. An AI visibility audit does the same thing at more depth if you want the source-pool map alongside it.

A shortlist you built from your own prompts beats any list of ten names, including this one's author.