The AEO Delivery Math: Hours, Capacity, and Margin for Agencies

Read every agency playbook on answer engine optimization and you will find the same two things: an org chart with three roles, and a price. What you will not find is the number that decides whether the service makes money.

Superlines' GEO playbook for agencies recommends €1,500 to €3,000 per month as an add-on to an existing SEO retainer, or €5,000 to €10,000 for a standalone audit. Writesonic's agency GEO guide lands in the same band at $1.5K to $3K monthly and $7K to $12K for an audit. Both name the same roles: a lead who owns research and testing, account managers carrying three to ten accounts, and content strategists doing the optimization work.

Both are useful. Neither tells you what delivery costs. So the sequence most agencies follow is: pick a price from a blog post, sell it, then discover in month four what it costs to deliver. That is not a pricing strategy. It is a bet.

This is the cost side. If you have not settled the packaging question yet, pricing and packaging an AEO retainer covers the revenue side, and this piece is what you should run before you commit to any of those numbers.

Why the published capacity numbers mislead you

"One strategist can handle five to ten clients" gets repeated everywhere. Read it carefully. That is a claim about account management capacity, not delivery capacity.

An account manager holding ten accounts is attending ten calls, sending ten reports, and answering ten inboxes. A specialist delivering AEO for ten accounts is running ten prompt sets across four engines, auditing ten sets of citations, briefing and reviewing content for ten brands, and doing ten reconciliations. Those are not the same job and they do not have the same ceiling.

When an agency takes the 5-to-10 number as a delivery ceiling, staffs to it, and then watches quality collapse at client seven, nothing went wrong operationally. The number was answering a different question.

Your real delivery ceiling is not something you can read off a blog. It falls out of your own hours per deliverable, which is the one input nobody publishes and everybody has sitting in their timesheets.

Build the delivery cost model

Do this per deliverable, not per client. Per client is where estimates go to die, because "a client" is not a unit of work.

The table below is a worked structure, not a benchmark. I am not going to tell you that a baseline audit takes 9 hours, because I have not measured your team and anybody who quotes you an industry hours figure has not measured theirs either. Replace every number with your own timesheet data. The structure is the point.

That last row is the one that quietly eats agencies. A prompt set is not a deliverable you build once. Engines change, the client launches a product, a competitor enters the category, and a prompt set that is six months stale is measuring a market that no longer exists. Every playbook tells you to build 10 to 20 core prompts. None tells you they decay.

Multiply your hours by your loaded rate, add tool cost per client seat, and you have delivery cost. Only now can you look at that €1,500 to €3,000 band and tell whether it is generous or suicidal for your shop. For a lot of agencies at four articles a month, it is closer to the second one than they expect.

The margin formula, and the variable that actually moves it

Monthly margin per client = retainer, minus (hours × loaded rate), minus tool cost per seat.

Run it and you find something counterintuitive. The lever with the most travel is not your hourly rate and it is not the retainer. It is content volume, because production plus briefing plus QA is the majority of your variable hours, and it is the line clients instinctively want more of.

Which produces the classic AEO retainer death spiral. Client asks for more articles. You say yes because you want the renewal. Your hours go up, your margin goes down, and citation share does not move much, because at some point the constraint was never article count. Now you are less profitable and losing the account on results.

The defense is to write volume into the scope as a fixed number with a documented price per additional article, and to be able to say out loud why article eleven does not help. Agencies that cannot make that argument end up buying their own renewal with margin.

Find your capacity ceiling, then hire against it

Take your monthly recurring hours per client and divide into a specialist's realistic delivery hours. Realistic means after internal meetings, sales support, and the research time that keeps the practice current. Nobody delivers 40 hours of client work in a 40-hour week, and an AEO specialist reading engine changes is not slacking, that is the job.

The ceiling this produces is usually lower than the number in the playbooks, and it is real, because it is yours.

Three things move it, in order of how much they actually help:

Standardize the prompt set structure. One taxonomy across every client, so a specialist switching accounts is not relearning a filing system. This is the biggest single win and it is free.

Standardize the QA gate. A written checklist turns a senior judgment call into a task a mid-level person can run. Judgment does not scale. Checklists do.

Batch by task, not by client. Test every client's prompt set on the same day. The context-switching cost of AEO work is high because the engines are the same and only the brand changes.

What does not move the ceiling: better tools. Tools compress the testing and reporting rows, which are already your cheapest hours. They do nothing to briefing, production, and QA, which are your most expensive. An agency that buys a platform expecting a capacity jump has misdiagnosed which rows are heavy.

When you hit the ceiling you have two options, and the honest framing is that hiring is a bet on pipeline. A specialist takes months to become productive, and if the fourth new logo does not close, you own the salary anyway. White-label AEO delivery converts that fixed cost into a variable one, at a worse unit rate. Lower margin per client, no capacity risk. Take the worse unit rate while your pipeline is still lumpy. Hire when it is not.

Report the metrics you can move

Search Engine Land's rundown of GEO metrics to track in 2026 lists eight: citation frequency, share of model voice, answer inclusion rate, entity recognition, sentiment, prompt coverage, retrieval success, and conversion influence.

All eight are worth knowing. They are not all worth reporting, and this is a delivery decision, not a measurement one. Every metric on the monthly report is a metric you will be asked to explain, and explanation is billable time you did not scope.

Lead with the two or three you can move with the work in the contract. If the retainer buys four articles and structural fixes, you can move prompt coverage, citation frequency, and retrieval success. You cannot move sentiment, and putting it on a report guarantees a call about a number you have no lever for.

Then reconcile monthly: citations earned against deals influenced. That is the only line that survives a budget review, and it is the reason the FastTrackr AI case study is worth reading before you build your report template. It shows both sides moving together, which is what a client is actually buying.

The month-four flat spot, and how to survive it

Every AEO retainer has one. Months one to three look great because the baseline was near zero and the first structural fixes are the cheapest wins you will ever get. Month four flattens. Citations plateau while your hours stay constant.

Nothing is broken. You harvested the easy gains and the next tier is slower. But it lands exactly when the client is deciding whether to renew, and if you did not tell them it was coming, the flat spot reads as failure.

So put it in the proposal. Name it, say when it arrives, say what happens next. Agencies that pre-commit to the flat spot renew through it. Agencies that discover it live spend month five defending the whole program.

Budget for it too. Month four is when you need retrieval-side work, which is a different cost line from content production. If the client's content cannot be ingested cleanly, no amount of publishing fixes it. The AI Feed Engine handles that layer, and the free llms.txt generator is a fast way to find out whether a prospect has this problem before you scope the retainer. Ten minutes there will change what you quote.

Run the model before the next proposal

The uncomfortable version: most agencies selling AEO today do not know their delivery cost, priced off somebody's blog post, and will find out in month six.

You do not need a perfect model. You need three numbers: hours per deliverable from your own timesheets, your loaded rate, and tool cost per seat. That is an afternoon with data you already have, and it tells you whether your current retainer has margin in it.

Run it before the next proposal, not after. How OnlyAEO works covers the measurement and content engine that sits underneath the delivery model, so you can drop a real per-seat number into the tool line instead of a guess. The service line build-out is the strategic version of this argument if you are still deciding whether to offer AEO at all.

The agencies that build a durable AEO practice are not the ones with the best tactics. They are the ones who know what an hour of delivery costs them and price like it.