The $2,000 AEO Retainer: What It Costs Your Agency to Deliver

A $2,000 AEO retainer can cost $1,770 to $2,860 a month to deliver. That leaves $230 in profit at the low end or an $860 loss at the high end, before onboarding or churn. Most agencies start their quote with the software invoice. Software is the smallest line.

I used to do the same thing. I would see a $99 tracker for 25 prompts and a $29 monitor for 15 prompts, add a couple of articles, and tell myself I had plenty of margin. Then the client needed schema fixed without giving me developer access, approval took two weeks, and the monthly report ate half my junior’s day. What the deck calls managed AI visibility, I call seven jobs that still need doing.

Here is the month, line by line.

The $2,000 retainer: a steady month, not onboarding

Assume one domain in month three: two optimised articles, one citation placement, weekly prompt checks, one monthly report and one client call. This is thin AEO scope. Managed AEO retainers run $1,500 to $25,000 a month, with entry programmes at $1,000 to $3,000. The point here is not to price every possible programme. It is to see what happens to a modest one when every delivery job gets a line.

These are planning estimates, not seven invoices you will receive on the first of the month. In particular, the reporting allocation depends on which tools you already pay for. Do not count the same subscription in both tracking and reporting. Even if you remove an overlapping tool charge, the labour is where the quote gets uncomfortable.

Tracking: the cheap line with expensive caps

Tracking has a pricing page, so it tends to become the anchor for the whole retainer. Otterly’s $29 monthly plan covers 15 prompts. Peec’s $95 plan covers 50 prompts across three engines. Semrush charges $99 for one domain and 25 prompts. Ahrefs Brand Radar starts at $199 on top of a base plan, per index.

I allow $99 to $199 per client for this line. The driver is not the login; it is coverage. An answer can change with the phrasing, location or run, so a single prompt checked once gives you little to work with. More prompts, engines and checks can push you beyond an entry plan. If a client wants 100 prompts checked weekly across ChatGPT, Perplexity, AI Overviews and Copilot, do not quote the 25-prompt tier.

Technical fixes: the developer still gets paid

AEO work meets the client’s CMS eventually. Schema needs changing, internal links need cleaning up, and pages need to be accessible to crawlers. Sometimes you can ship those changes. Sometimes you submit a ticket and watch it age.

US freelance developer rates range from $60 to $100 an hour at mid level and $100 to $175 at senior level. I allow two to four hours of technical work in a steady month, or $180 to $400 in the model. The exact work will move around; pretending it never happens will not make it disappear.

The dev queue or an edge-proxy setup

These are different costs, and the proposal should say which one it assumes:

1. Use the client’s dev queue. You may avoid a direct developer invoice, but changes can wait while the retainer clock keeps running. I used to tell clients that was fine. For work that depends on a fix shipping before a page gets cited, it often is not.
2. Set up an edge proxy. You may be able to ship some schema and page changes without that queue, but setup and testing require specialist time. The cited senior and specialist rates run from $100 to $250 an hour.

The setup belongs in a one-off budget, not quietly inside the recurring two-to-four-hour allowance. Choose the route before you promise a delivery date.

Content: writing is only the first pass

Agency-grade articles run $150 to $800 each, with publishing and links extra. For two edited AEO articles, I put $600 to $800 in the monthly model before either one touches the CMS. Content volume and off-site authority are major recurring cost drivers because a page has to answer directly, read cleanly and survive a subject-matter review.

Then it reaches the client. Legal softens a claim. The subject-matter expert rewrites an answer. Nobody owns the final approval. That delay does not show up on a writer’s rate card, but the rewrite and coordination still consume time. Allow for two approval rounds per article, not a fantasy in which the first draft sails straight through.

Publishing: approved is not live

Publishing means getting into the CMS, formatting the page, adding internal links and checking what the live theme did to your work. A broken staging link or a theme that strips FAQ schema turns a quick upload into another round of fixes.

I allow $75 to $150 a month for one to two hours of junior time across the publishing cycle. If the agency cannot publish directly, a screen-share session with the client’s web person adds friction too. The practical distinction is simple: the content line pays for approved words; the publishing line gets those words onto a working page.

Citations: one placement is not an authority strategy

Other sites mentioning the client matter, but a $2,000 retainer does not buy much room to pursue them. A link insertion averages $179, while a vendor guest post averages $461. Digital PR can run $1,250 to $1,500 per unique link, with typical campaigns at $5,000 to $10,000.

The model buys one $179 to $461 placement. That is before outreach time, and a cheap insertion does not guarantee a useful mention. One placement may meet the deliverable. It does not amount to a sustained authority push. If the sales deck promises the latter, this budget is too thin to support it.

Reporting: a CSV is not the answer the client bought

Prompt trackers collect runs. Someone still has to deduplicate prompts, check citations, explain what changed and name the next fixes. I allow two hours for a disciplined monthly report. AI answers can differ between runs, so an unexplained visibility swing needs investigation, not a brighter chart colour.

The model assigns $287 to $450 per client to reporting tools and labour. The underlying analytics and dashboard price ranges are shared-tool costs, not automatically a fresh bill for every domain. Allocate them across clients, and remove any charge already counted under tracking. Keep the human hours. If reporting regularly takes more than two hours, the prompt set or the reporting promise is too large for this retainer.

Account time: calls, approvals and the dip nobody predicted

Someone answers Slack, runs the call, chases the article approval and explains why visibility dipped. A junior handling four to five clients costs $300 to $400 per client per month; high-touch coverage runs $500 to $800. I use $350 to $400 for standard account time. Do not call that free because the person is already on payroll.

Churn sits outside the steady-month table. The cited replacement cost is $2,000 to $3,000, against an average client life of 14 to 18 months. At a $230 best-case monthly margin, replacing a lost client can consume many months of profit. The proposal should not confuse revenue with money available to absorb that loss.

Hidden costs: recurring rework versus one-off setup

The table describes a steady month. Two other bills need their own place in the model.

Recurring: a citation changes, and the page needs rework

An engine cites your comparison table one week and another source the next. You inspect the change, revise the passage, republish and check again. I would reserve one three-to-five-hour rework cycle per quarter per client. At mid-level freelance rates of $60 to $100 an hour, that is roughly $80 to $170 a month when spread across the quarter.

This reserve is not included in the $1,770 to $2,860 table. If your existing technical or content allowance already covers that cycle, do not add it again. If it does not, the apparently profitable end of the range gets thinner.

One-off: onboarding before anything goes live

Month one is not month three. Baseline prompts, an audit, CMS access, GA and Search Console, an entity check and a content plan can take eight to twelve hours before you ship anything citable. At the junior and mid-level rates used here, I would budget $600 to $1,000 for unbilled setup if the agency charges the same $2,000 from day one.

You can charge onboarding separately or recover it over the relationship. Just do not present a steady-month margin as though setup never happened. A short-lived client gives you less time to earn that cost back.

What changes at 5, 10 and 15 domains

Multiply the thin monthly scope and the exposure becomes easier to see:

Each domain still needs its own articles, approvals, fixes and report. A tracker may offer an agency tier, but a volume price on software does not give the team back its writing or client-call hours. Labour scales far more stubbornly than the login bill.

Cheap tools versus flat-rate execution

Here is the comparison agencies tend to leave out. A low-priced tracker can beat a flat execution fee as a software purchase. It does not necessarily beat that fee as a delivered service.

groas lists SEO and AI Search from $199 a month per client domain, with content, technical fixes, citations, visibility tracking and branded weekly reports in the fee. For this comparison, I replace the DIY delivery lines with that starting fee and retain $350 to $400 per client for agency account time. I do not pretend the client stops calling because another provider executes the work.

That is a starting-price comparison, not a claim that every client’s scope fits the starting fee. Confirm what the domain needs before replacing your cost lines. The worked 10-client version goes further into overhead; this table sticks to the delivery model above.

When I would keep the DIY stack

I would keep it when the agency already has writers and a developer on salary, handles fewer than three AEO clients, and sells custom strategy by the hour rather than a packaged outcome. In that setup, a cheap tracker alongside existing staff can make sense. I would still count the staff hours before calling the work profitable.

For a packaged retainer, the shift from white-label SEO to AEO pricing exposes the weakness of a per-deliverable menu: writing, technical work and authority building do not vanish because the tracker costs $99. A flat per-domain execution fee earns its place when it replaces those jobs, not when someone compares it with software alone.

Price the labour before the login

Start with the work the client will expect every month. Then add tracking, shared tools, setup and a reserve for rework. Put approval rounds and publishing access in writing. Otherwise the $2,000 retainer looks healthy only until somebody has to deliver it.

The cheapest mistake is quoting the $99 tracking tier as though it represents the cost of the service. The most expensive mistake is selling full AEO scope for $2,000 while staffing it like a small software add-on. I have made the first mistake. The table is why I would not make the second.