AEO agency operations are the processes an agency uses to get a client's brand cited or recommended in AI-generated answers. The stages look like SEO stages, but the unit of work changes from a keyword and a ranking to a question and an answer. That change reaches intake, measurement, reporting and capacity.
A common shortcut is to bolt AEO onto the SEO workflow and report it in the SEO template. That fails quietly, because the SEO template has nowhere to put an answer that changes from one run to the next. I held the Head of SEO Operations and AEO Strategy role at W3whiz and managed the engineers building AI-search tooling at Search Atlas. For the SEO model, start with SEO agency operations.
How AEO agency operations start: prompt sets, not keywords
AEO intake collects questions, not keywords. The input that drives everything is a prompt set: the list of questions a buyer would put to an answer engine about the client's category, problems and brand. Building it is a joint task with the client, because they know the questions sales hears and you know which ones produce answers worth being in.
A prompt set should group questions by intent, such as category questions, comparison questions, problem questions and brand questions. It should also record which answer engines matter to this client. A B2B software client and a local services client will not care about the same surfaces. Intake also needs a baseline: for each prompt, which sources are cited today, which brands are named, and whether the client appears.
Scope by prompt group
AEO scope is defined by prompt groups and content assets, not by traffic targets. You agree which prompt groups you are working on, which pages or new assets serve them, and which fixes sit outside your reach. The last part matters. If an answer engine cites a competitor's comparison page and a review site, you cannot edit either. Scope has to say what the retainer covers and what it can only influence.
A scope that promises citations sets up a renewal problem you cannot solve.
Measuring by sampling
You measure AEO by sampling answers repeatedly and recording what appears. Answers from the same prompt can differ between runs and between engines such as ChatGPT, Gemini, Perplexity and Google AI Overviews, so a single check proves little. A sampling protocol fixes the prompts, the engines, the number of runs and the recording format, so that month-over-month comparisons mean something.
| Measure | What it tells you | Limit |
|---|---|---|
| Citation presence | Whether the client's pages appear as sources for a prompt | Varies between runs, so report a rate across runs, not a yes or no |
| Brand mention | Whether the brand is named in the answer text | Can occur without a link, and a link can occur without a mention |
| Answer accuracy | Whether what the engine says about the client is correct | Needs a human reading each answer |
| Competitor presence | Who is cited or named instead | Only as complete as your prompt set |
Say plainly in reports that these measures are directional. Nothing in AEO matches a stable rank position, and tooling that tracks AI answers is still maturing. A client who is told that early trusts the report more later.
Deliverables
AEO delivery is a small set of repeatable content and technical assets tied to prompt groups. The deliverables differ from SEO delivery in kind.
- Answer-first pages. Pages that open with a direct, self-contained answer to a specific question, then support it.
- Source-page upgrades. Revising existing pages that already get cited or should be, so claims are specific and attributable.
- Structured data. Markup that states what the page is and who it is about, validated before release.
- Fact consistency. The brand's core facts stated the same way across the site, so an engine does not synthesize a contradiction.
- Gap briefs. A short brief for each prompt where the client is absent, naming what a citable answer would need.
Each asset has an owner and a QA step. The QA step checks that the opening sentence answers the question on its own, that claims are checkable, and that the markup validates. The checklists are in AEO agency SOPs.
What goes in an AEO report
An AEO report shows prompt-level movement, what was shipped, and what is being tried next. Lead with a table of prompt groups and the sampled citation and mention rate for each, compared with baseline. Follow with the assets shipped against those groups. Close with the gaps still open and the next change to test.
Do not fold AEO numbers into the SEO traffic chart. They answer a different question, and a dip in one does not mean a dip in the other. Where answer engines send visitors, you can look at that referral traffic, but treat it as a supporting signal and not the headline.
Where AEO delivery breaks
- Unsampled claims. Someone checks one prompt once and reports a win or a loss. The next run says the opposite.
- Promised outcomes. The scope implies citations. The engine changes behavior. The client holds you to the promise.
- Monitoring hours nobody planned. Sampling is recurring work. If it is not in capacity, it is done by whoever has a spare hour.
- Everything treated as content. Fact inconsistency and missing markup get ignored because the team only produces articles.
Monitoring is a capacity line of its own
AEO capacity has a recurring monitoring load that SEO does not. Each client needs a sampling run per cycle across the agreed prompts and engines, plus a human read of accuracy. Multiply that by the client count before you promise a start date. This is where the hours go missing, and it is why the number on a utilization report can look healthy while the monitoring is being done off the clock.
Size it the same bottom-up way as SEO: hours per task type, summed per person. Then add monitoring as its own task type. The principles in the utilization benchmark guide apply, but the available-hours trap is sharper, because new work types tend to get absorbed instead of booked.
Checking the model before you sell it
The fastest check before adding AEO to a roster: do prompt sets, a sampling protocol and a separate report section exist? If any is missing, the work is being done on the SEO model and will misreport. That is the kind of gap I look for in an AEO agency utilization consultant engagement. For the wider framework, see the AI search optimization guide, and for the entity side of the problem, GEO agency operations.
FAQ
What are AEO agency operations?
AEO agency operations are the processes an agency uses to deliver answer engine optimization: building prompt sets, creating answer-first assets, sampling answers across the major answer engines, and reporting on citation and mention rates. They differ from SEO operations because the unit of work is a question, not a keyword.
How do you report AEO results to clients?
Report sampled citation and brand mention rates per prompt group against a baseline, then list assets shipped and gaps still open. State that the measures are directional, because answers vary between runs and tooling is immature. Keep AEO numbers separate from the SEO traffic chart.
Can an SEO team deliver AEO without changes?
Rarely. The SEO team's skills transfer, but the workflow does not. AEO needs prompt sets at intake, a sampling protocol for measurement, answer-first QA and a recurring monitoring task in the capacity plan. Without those, results get reported on the wrong template.
Should an agency promise AI citations?
Promising citations is a mistake. Answer engines change behavior without notice, and an agency cannot edit the third-party pages that are often cited. Promise the work: prompt sets, shipped assets, sampling and reporting. Describe outcomes as directional so the scope holds up at renewal.