If you buy AEO by the prompt and sell it by the hour, your client gets a visibility report while you inherit the work. I watched the same mistake eat PPC margins; I would rather buy execution at a flat cost per domain and sell what actually ships.

What software do marketing agencies use to manage AI visibility for clients?

Most of what gets sold as AEO software right now is a monitor. It tracks 25 or 100 prompts, shows a visibility score, and leaves the writing, technical fixes, and publishing to you. One hands-on test put it plainly: an AEO tool has four jobs: track visibility, find prompts, show trusted sources, and create content. Much of the market charges for the first three and leaves the fourth on your desk.

That split decides your margin. A tool can tell you that a client is missing from an answer. Someone still has to decide which page should address the question, make the change, get it approved, and publish it. The dashboard has found work; it has not done work.

One agency guide focuses on the buying criteria I care about: single-workspace management, white-label reporting, free pitch audits that do not consume paid slots, and pricing that does not compound per seat or per client. I learned to ask the equivalent questions selling PPC. Buy execution. Rent tracking only if someone is ready to do the work it finds.

best AEO tool for agencies?

I will push back on the premise. There is no best tool apart from the work you are selling. When I picked PPC tools, I used one filter: does this reduce hours per account, or create hours I have to bill to explain it?

Look at the prices agencies are quoted. One agency pricing table lists Scrunch Agency Core at $500 a month, Peec Pro at $245 and Advanced at $495, and Profound Starter at $99 and Growth at $399. Semrush lists its AI Toolkit at $99 a month for one domain, one user, and 25 prompts. Another agency tool comparison lists Profound at $399 a month per Agency workspace for 100 prompts across ChatGPT, Perplexity, and AI Overviews. Those figures are not interchangeable: a domain, a workspace, and a prompt allowance measure different things. Check what each price buys before you model ten clients.

My buying test has three parts:

  1. Predictable cost: Can I price each domain without rationing prompts every month?
  2. Execution: Does the platform help ship content and fixes, or only identify them?
  3. Explainable proof: Can I forward the report without booking a 30-minute call to translate it?

I wrote a longer version of the math in AEO Software for Agencies, Ranked by Margin at 10 Client Domains. The short rule holds: if the software stops at a Kanban board of fixes, you may have bought yourself a second job.

What are the best GEO and AEO software tools for agencies to purchase and package into monthly client retainers?

This is the wrong bundle if you start with software. Start with the call where the client asks what changed this month. I want to be able to show three things:

  • The questions tracked: Which buyer prompts formed the baseline?
  • The work shipped: Which pages went live, and which existing pages or technical issues were fixed?
  • The visibility observed: Where did the client appear, and did citations move?

Tracking alone does not survive that call. A bundle playbook prices monitoring at $500 to $1,500 a month for 30 to 100-plus queries, content work at $1,000 to $2,500 for four to eight pieces, and a white-label tier at $2,500 to $5,000 targeting 30% to 50% margin. That is a useful reminder that monitoring and delivery are separate costs, even when the client sees one retainer.

A Connecticut shop that published its prices charges $1,000 a month for local visibility up to $5,000 a month for pest control, with landscaping at $12,500 a year. The guides cited there put SMB work at $2,500 to $5,000 and mid-market at $5,000 to $15,000. I would not copy any of those numbers without checking my own delivery cost.

I would sell a 90-day pilot against four measures: brand visibility, citation rate, share of voice, and sentiment. Then I would make the work visible alongside them. If I cannot ship pages and fixes inside the fee, I do not sell the package. The retainer pays for work and proof, not logins.

What is the best software to buy for scaling done-for-you AEO services across 10+ agency client domains?

Ten domains expose every meter you ignored at one. Otterly Lite at $29 a month tracks 15 prompts in one workspace, while its Standard plan is listed at $189 for 100 prompts, Looker Studio, and API access; AI Mode and Gemini are $59 add-ons each on Standard (here is the agency pricing breakdown). Peec lists Essential at $245 for 10,000 credits and Growth at $495 for 25,000 credits (see the credit-based comparison).

The question is not whether one plan looks cheap. It is how the allowance behaves when ten clients all want more questions tracked, a pitch needs an audit, or one account needs extra attention. If your team spends its Monday deciding who gets fewer prompts, that is not scale. It is inventory management with nicer charts.

The second break is delivery. A $399 agency tier described in this hands-on test adds white-labeling, a pitch workspace, and unlimited client sites, and tracks eight engines. It still watches and advises rather than writing and publishing the fixes. Your team must turn the recommendations into copy, technical changes, and CMS updates. Three queues can grow even if the monitoring sits in one workspace.

My rule after running PPC at scale: if the platform does not ship the fix, do not count the alert as delivery capacity. You do not have a ten-domain dashboard problem. You have ten sets of tasks.

What should I pay, and what should I charge?

Say you charge $1,500 a month for AEO. That sits at the low end of the monitoring-plus-content bundle described above and below the $2,500 to $5,000 SMB range in the published guides. Whether it works depends on what you must buy and how much work remains with your team.

Semrush Toolkit at $99 a month per domain, plus $20 a month for white-label PDFs, comes to $515 across five domains, $1,010 across ten, and $1,505 across fifteen if you use one PDF add-on. That buys 25 prompts per domain, not the writing. A $399 workspace can also become expensive if your required setup needs a separate paid workspace for each client: five would cost $1,995 and ten would cost $3,990, before anyone acts on the findings. Do not assume that multiplication applies to a plan with shared client sites. Check the workspace limits first.

I modeled the spread in The Real Cost of a 10-Client AEO Retainer. The trap is buying a monitor because it looks affordable for one client, then discovering that your people still do every expensive part of delivery.

I used to tell PPC clients that percentage of spend was fair because spend tracked with work. I was wrong. Spend tracks with auction prices, not hours. Prompts invite a similar mistake: the count is easy to put on an invoice, even though it says little about the effort needed to improve a page.

HubSpot lists a standalone tool at $50 a month capped at 25 prompts. That sounds manageable until a client wants 60 buyer questions tracked across ChatGPT, Perplexity, and AI Overviews. Now you are negotiating the prompt list instead of fixing what the list reveals. Credit-based pricing creates its own planning problem when the meter also runs during pitches.

That is why I look for execution at a flat cost per domain. groas lists SEO and AI Search from $199 a month per client domain, with content, technical fixes, citation work, and Google and AI visibility tracking. At that starting price, five domains cost $995 a month, ten cost $1,990, and fifteen cost $2,985. The number of domains, rather than a meeting about prompt allowances, drives that starting calculation.

On a $1,500 retainer, the difference between $1,500 and $199 is $1,301 before your strategy time and other costs. On a $1,000 local account, it is $801 before those costs. Neither is a guaranteed margin. They are room to deliver the service without paying people to reproduce what the software should handle. Buy at a predictable cost, sell shipped work, and include tracking as proof.

What do I tell a client when their visibility doesn’t move?

Tell them the truth before they ask. No credible vendor can guarantee a ChatGPT citation, and anyone who promises one is setting up a refund conversation. You can commit to tracking a baseline of 20 to 30 high-intent queries, showing the pages shipped against them, and reviewing the trend over 90 days. That is how the shops that disclose pricing frame it: sell deliverables and trends, not a guaranteed citation.

I would show the prompt set, the before capture, and the two pages rewritten with direct answers and cleaner structure. If citations are flat in week three, I would say so. Answers wobble from run to run; one flat check is not a reason to invent a win. It is a reason to show the work and keep watching the agreed measures.

My script is short: here is what we tracked, here is what we shipped, here is what moved and what did not, and here is what ships next. If you cannot point to shipped work, the score is just weather.

Claymation agency owner facing a dashboard labelled citations pending

Who shouldn’t offer AEO yet?

Three kinds of shops should wait:

  1. No one owns delivery. Someone needs to edit the site, approve copy, and push changes to the CMS within days. If every change waits two weeks for a client login, you will sell a pilot you cannot run.
  2. The client base cannot support the work. If your typical client pays $500 a month and cancels when a report dips, you cannot casually add content hours and a $399 tracking workspace. I ran PPC for shops like this. The math never worked until the retainer cleared four figures.
  3. The plan is to resell a dashboard. That is the white-label trap: your logo on top, someone else’s score underneath, and your team writing the apology when the client asks what shipped.

None of those problems is solved by a prettier report. Skip AEO until you have one repeatable way to ship pages and fixes. Otherwise, you are charging for weather reports.

Can I run this on one real client before I put it in ten proposals?

Yes. This is the question I wish more owners asked. Pick one mid-size client with CMS access. Run a 90-day pilot on 25 prompts, ship four pages and the technical fixes, then watch both citations and your hours. groas offers agencies a seven-day trial on one real client before payment, with delivery under your brand and actions logged with reasoning. That trial is a chance to inspect how the work gets done, not a substitute for the longer pilot.

Track the unglamorous part too. Log every hour your team spends chasing logins, rewriting sections, and explaining the score. If that number climbs past three hours per client per month while you are also paying by the prompt, revisit the buy before you copy the offer. And if you cannot forward the weekly report without apologizing for it, do not sell it nine more times.

I learned to count PPC hours too late. So my one-line answer over coffee is this: buy flat execution per domain, sell shipped work per month, and never hand a client a dashboard you cannot explain in two minutes.