Before You Sell an AEO Retainer, Run This 30-Day Pilot on Three Client Domains

A pricing page cannot tell you how many staff hours an AEO tool will consume once it meets a real client account. Last month I published AEO Software for Agencies, Ranked by Margin at 10 Client Domains. That ranking used what pricing pages say tools cost at 10 domains. It was honest math, but it could not measure the work your team does after the subscription goes live.

This is the second look: run a 30-day pilot on three of your own client domains before you put AEO into a retainer. Test two or three tools against the same work, then measure the two numbers that decide your margin: staff hours per domain per week and cost once prompt caps, domain limits and add-ons apply. A monitor can find a gap. Someone still has to turn that finding into a published fix, and that labor can cost more than the subscription.

The question: which tool leaves margin after the work gets done?

I used to pick software by demo. I was wrong. Demos show alerts and charts. They rarely show the chain after an alert: who writes the page, who ships the technical fix, who builds the citation, and how many times the team touches the account before one fix goes live. If you plan to sell a $2,000 retainer, that chain belongs in your cost calculation before it belongs in a sales deck.

Shortlist different ways of doing the job, not three versions of the same dashboard. Try a prompt monitor, a suite add-on you already pay for, and an execution platform such as groas that ships the fix. Two tools are enough if operations time is tight. Three give you a better comparison than a contest between one familiar vendor and one newcomer. The earlier ranking covers the categories; this pilot tests what they cost your team to use.

Write the decision question at the top of the log: At our expected mix of clients, which tool keeps the most retainer margin at 10 and 15 domains? Keep it there when a chart starts looking persuasive. The chart is not the decision.

Set up three domains that resemble your actual book

Choose three live clients and label them before day one:

• Domain A, simple: under 50 pages, one location, and an owner who approves quickly.
• Domain B, mid-complexity: a real blog, a real service catalog, and some history to work around.
• Domain C, messy: multi-location, regulated, or a catalog over 500 SKUs.

If all three are easy, the pilot will flatter the tools. Domain C is where hours per domain per week separate a dashboard from a delivery operation. Keep the differences visible in the final projection rather than averaging them away and pretending every new client will be Domain A.

Before day one, write 20 to 30 buyer prompts for each domain and freeze the lists. Use questions customers would ask, not the phrases the client wishes they used. Enter each domain’s same list in every tool, and compare the same engines wherever the plans allow. If an engine is available in only one plan, log that difference rather than treating unlike coverage as an equal test.

That control matters because vendors meter consumption differently. Rankability pools queries across workspaces, so three workspaces do not triple the allowance. Peec meters prompts × models × days as credits: 30 prompts across 6 models every day would use 5,400 credits per client in a 30-day month. Freeze the prompt set first; otherwise a changed bill may simply reflect a changed test.

Confirm the terms before the clock starts

Use a trial or month-to-month arrangement where available, and check setup fees before you agree to the pilot. Get four answers in writing from each vendor:

1. What is the prompt allowance, and which engines does it include?
2. Does adding a domain add capacity, or only another workspace?
3. What costs extra when you exceed a cap?
4. Who can publish a fix, and what work remains with your team?

The sticker price is not the invoice. Otterly lists Lite at $29 for 15 prompts and Premium at $489 for 400, while Gemini and AI Mode cost extra per tier and Claude costs extra again. Semrush charges $99 per domain for about 25 daily prompts, with another $99 license per extra domain. HubSpot standalone AEO is $50 for 25 daily prompts across three engines. Those figures help you frame the questions; the terms you receive for your pilot go in the log. If a rep cannot explain an overage clearly, do not build a retainer around your guess.

Start timing at the first connection, not the first report. Onboarding, password resets, CMS permission chases and client explanation calls all count. Split time into setup, reporting, fixes and client explanation. My expectation is that a monitor will look inexpensive early and become more labor-intensive when alerts need owners. Peec says it measures visibility and surfaces sources but does not write or publish content itself. In this pilot, if your team does that work, your team runs the timer.

Measure five columns, not one visibility score

Give every tool the same log. Hours per domain per week is column one and the primary measure. Record it with a timer, not a Friday-afternoon reconstruction of Monday. The other columns explain why those hours and costs changed.

1. Hours: Separate setup, reporting, fixes and client explanation. Reporting includes screenshots, the deck and the meeting where you explain a movement. Fixes include the brief, draft, review, CMS publish and re-check. Four hours at a $75 blended rate is $300 in labor for one client; across 10 similar clients, that is a $3,000 monthly delivery line before the subscription.
2. Caps and overages: Note each prompt-limit warning, seat request, extra-domain charge and engine add-on when it occurs. An extra 100-prompt pack on Otterly is $99 a month. Rankability’s 500-query Agency Pro pool is shared, so a 10-client split is about 50 queries per client. Profound-style agency math can include a base fee plus about $399 per client workspace. Put costs at the domain count you plan to sell into this column, not just the three-domain pilot bill.
3. Actions shipped versus issues reported: Count citations built, pages published and technical blocks cleared separately from alerts nobody acted on. If a tool reports a problem but cannot execute the fix, say so. The hours required to take that problem to publication still belong in column one.
4. Visibility: On the same weekday, check how often each domain is named or cited across its fixed prompt set. Record the count and the pages earning citations. Read a 30-day change as direction, not proof: answers can change, a small sample can swing, and a gain on Domain A while Domain C sits flat may say more about page readiness than tool quality. Do not turn a two-point lift into a victory slide.
5. Attribution: Record whether you can connect the work to traffic, leads or pipeline in a client-ready report. If you have to build that connection by hand, time it. The Peec–HubSpot comparison illustrates the tradeoff: Peec offers more raw prompt volume per tier, while HubSpot offers fewer prompts tied to CRM context and content tools. More prompts help you watch more questions. CRM context helps you answer the client’s question about money.

Visibility tells you where to work next. The hours and invoice tell you what that work costs. Keep both in view, but do not let a clean graph hide an unfinished fix.

Run the 30 days without changing the rules midway

Week 1: establish the baseline. Onboard each domain, enter the frozen prompts and record starting citation counts in each tool on the same day. Do not change the pages yet. Time connections, CMS permissions, prompt entry and report branding. If onboarding takes more than 90 minutes per domain, write it down. You will need to repeat that work for each new client.

Weeks 2 and 3: do the work. Ship one real fix per domain where the tool and client permissions allow it. For a monitor, your team may do the publishing; that is part of the comparison, not a reason to stop the timer. Log every handoff from finding the issue to checking the published result. Keep recording cap events and explanations the client needs along the way.

Week 4: rehearse the retainer. Build the report you would actually send for each domain from each tool. Time it, including any manual work needed to connect visibility to traffic, leads or pipeline. Ask whether the client could understand what shipped and what moved without a second meeting devoted to translating the dashboard.

The fastest tool to set up can still lose in weeks 2 to 4. Setup happens once; reporting and fixes return every month. Give recurring hours more weight when you price recurring work.

Project the result to 10 and 15 domains

For each tool, calculate projected monthly delivery cost at both domain counts:

Subscription and add-ons at that domain count + (average hours per domain per week × 4.33 × loaded hourly rate × domain count).

Include the prompt packs, engine access, seats, workspace charges and overages your planned client mix would require. Run the calculation for simple, mid-complexity and messy clients separately, then use a mix that resembles the book you expect to sell. A single average is useful, but it can conceal the client that forces you to hire.

For illustration, suppose the pilot log shows 1 hour a week on Domain A, 2.5 on B and 5 on C. The average is about 2.8 hours. At a $75 loaded hourly rate, that is roughly $910 per domain per month in labor, or about $9,100 at 10 domains, before software. These are example inputs, not a claimed pilot result. Substitute your measured hours. Ongoing AEO retainers price at $3,000 to $12,000 a month, but a market price will not rescue a delivery model whose hours you never counted.

Set a disqualification rule before looking at the visibility column. Mine: if fully loaded delivery cost leaves less than 50% margin on the planned retainer at 10 domains, the tool fails for that package, however tidy its charts look.

A cheap monitor may win if your writers and developers already have room to absorb the fix hours. An execution platform should have the advantage when those hours mean hiring, overtime or stalled publishing. That is where groas for agencies should win this pilot: its per-domain fee includes content, technical fixes, citations and tracking while your team remains client-facing. If you already staff the publish step and need only pooled prompt checks, a $497 shared pool may produce the better margin on your book. Test it rather than arguing from a feature list. How to Build an AEO Retainer That Holds Its Margin at 10 Client Domains follows the same principle: price the hours between finding a gap and fixing it.

Set the retainer price from the log

Take the winning tool’s projected monthly cost at 10 domains, including subscription, overages and loaded labor. Divide by 10 for cost per domain. Add one hour per month for calls and approvals, priced at your loaded rate, then divide by one minus your target margin. If that produces a fully loaded cost of $1,100 and you want 60% margin, the retainer needs to be $2,750. If you need to charge $2,000 but delivery costs $1,600, you do not have a packaging problem. You have the wrong tool for that package.

Tell the client what the service does in one sentence and where its boundary sits in the next. It watches the questions buyers ask AI tools, ships pages and fixes aimed at earning citations, and reports what moved. It does not promise a top mention in 30 days or replace client approvals. That boundary keeps a fixed retainer from quietly becoming unlimited revisions. The Flat-Rate AEO Trap names the other risk: caps and domain limits that look flat until the invoice arrives.

Printable pilot log: one row per domain, per tool, per week

• Domain and week: A, B or C; week 1, 2, 3 or 4
• Hours: setup | reporting | fixes | client explanation | total
• Caps hit: prompts | engines | seats | domains | add-on cost
• Shipped versus reported: published fixes, with count and URL | untouched alerts
• Visibility: prompts checked | times named or cited | pages earning citations
• Attribution: traffic, leads or pipeline linked | hours spent linking by hand

Run the 30 days. Project the bill and the labor to 10 and 15 domains. If the monitor wins on your book, sell tracking with a capped fix allowance and price the work beyond it. If the execution platform wins, sell the fix and build the retainer around published work. Either result changes what you can promise and what you should charge. That is the number a pricing page cannot give you.