How We Price a GEO Retainer in 2026: Scope, Tooling Cost per Client, and What Stays Out
How we structure a GEO retainer: what the monthly fee covers, what Promptwatch agency plans cost per client at 5, 15, and 40 clients, and which work we quote separately.
Other agencies starting out in GEO tend to ask us two pricing questions. What goes in the retainer, and how much of it disappears into software? We won't publish our own fee. It depends on markets, languages, and how much content a client needs, and a number without that context would mislead more than it helps.
The structure is more useful anyway. It's the part that stops a retainer from turning into unpaid content production by month four. Below is how we scope a GEO retainer, how we work out tooling cost per client, and which work we keep out of the monthly number on purpose.
What the retainer covers
A GEO retainer pays for a running program. Ours has four parts.
The prompt set. We design it in the first month and review it every quarter. That covers the research (Search Console queries, sales and support language, Prompt Explorer suggestions), the volume and difficulty check on every candidate, personas where buyer types differ, and the locations the client sells in. The quarterly review is in scope. A rebuild because the client opened a new market is not, and we'll come back to that.
The monthly report. A scope block, visibility and share of voice against the named competitors, citations by page and domain, crawler activity, AI-referred visits, a dated log of what we shipped, and next actions with owners. Writing the interpretation takes longer than pulling the data. The fee should reflect that.
Fix and content work, inside a fixed allowance. Each month we ship an agreed amount from the action queue: page refreshes driven by content gap analysis, fixes from the weekly site audit, crawl errors, internal links, offsite outreach. We set the allowance in hours or pages at the start. When the queue outgrows it, the client decides what waits. Nobody gets unscoped work for free, and we can't quietly under-deliver either.
Crawler log monitoring. Once logs are connected, reading them is retainer work. Which AI crawlers fetched which pages, which requests failed, and whether a page we changed got fetched again before it started earning citations. The ongoing reading is in. The initial connection depends on the client's stack, so it gets its own treatment below.
Tooling cost per client
We run retainer clients in one Promptwatch agency workspace. All three self-serve agency plans include unlimited projects, unlimited prompts, unlimited pitch projects, and 10 team seats. What changes between tiers is the size of the shared pool: responses, visitor events, crawler logs, and Agent Credits.
That makes the per-client arithmetic easy. Here it is at 5, 15, and 40 clients, roughly the roster sizes where an agency would sit on each tier.
| Plan | Monthly price | Clients | Tool cost per client | Responses per client |
|---|---|---|---|---|
| Kick-off | $199 | 5 | $39.80 | 2,000 |
| Growth | $399 | 15 | $26.60 | about 1,667 |
| Scale | $799 | 40 | about $19.98 | 1,625 |
Annual billing is 12 months for the price of 10. That brings the same per-client figures down to roughly $33.17, $22.17, and $16.65.
Two things come out of this table that the pricing page doesn't spell out.
The cost per client falls as the roster grows, but so do responses per client. The pool grows more slowly than the roster. So the reason to move up a tier is almost never price per client. It's running out of responses.
We use a yardstick for that. Every Promptwatch brand plan is sized at 120 responses per tracked prompt per month: 6,000 for 50 prompts on Essential, 18,000 for 150 on Professional, 42,000 for 350 on Business. At 2,000 responses, a client gets about what a brand plan would allot to 16 or 17 prompts. At 1,625, about 13. That isn't how responses are actually spent, since consumption depends on how many models and locations each prompt runs on. It's a sanity check before we promise a client a prompt set the pool can't carry.
The even split is also fiction. A multi-market client with city tracking can burn through several times the responses of a single-market client. We budget the pool per client at onboarding, write that allocation into the scope, and check usage before adding prompts. If the mix pushes the pool to its limit, we move up a tier rather than thin everyone's coverage.
Per response, the agency tiers differ more than people expect. Per 1,000 responses, Kick-off works out at $19.90, Growth at $15.96, and Scale at about $12.29. At the bottom tier you're not buying cheap responses. You're buying unlimited projects, ten seats, and state and city tracking plus Ads Radar on every client, both of which only start at Business ($579) on the brand side.
We fold tooling into the retainer instead of invoicing it separately. At any realistic roster size it's a small slice of the fee, and an itemized $26.60 line invites a conversation about the wrong thing.
What we bill separately
Some work is lumpy, one-off, or open-ended. Put it in the monthly fee and you either overcharge a client in quiet months or eat the cost in busy ones. These get their own quote.
One-off setup. The first month carries more work than any month after it: access requests, prompt research, competitor setup, the baseline report, analytics checks. We quote that as a setup fee. We don't do long-term contracts, so spreading setup over a twelve-month term the client could leave after three would be dishonest pricing.
Content production. Net-new pages are the most variable cost in GEO. One gap needs a short answer block added to an existing page. Another needs a new comparison page built from original material. The retainer covers refreshes and fixes inside the allowance. New articles are quoted per piece or as a monthly content add-on.
Plan choice shows up here too. Promptwatch's agency plans don't include AEO articles or the automated Content Agent. On those plans our team writes the content, and we publish to WordPress through Promptwatch (WordPress publishing is available to every organization) or through the client's own CMS, always with a review step.
CDN log setup that needs the client's engineers. Crawler log connection takes minutes on some stacks and a small project on others. A site behind Cloudflare can use Logpush. WordPress sites can use the official plugin. Vercel and Netlify use log drains. AWS CloudFront goes through standard logging v2 via Kinesis Data Firehose, Akamai through DataStream 2, and some setups need a custom HTTP endpoint. The simple connections sit inside setup. Anything that needs time from the client's infrastructure team gets quoted once, up front.
Prompt set rebuilds. A new market, a new language, or a new product line means a new research cycle. We price it like the original setup.
When one client is better on a brand plan
The agency workspace is our default, not a rule. Three situations push a client onto its own brand plan, billed to the client.
An ecommerce client that needs Shopping Insights. Tracking of ChatGPT shopping carousels and Alexa product results lives in Shopping Insights, which starts on Business at $579 a month. It isn't on the agency tiers.
A client that wants scheduled publishing. The automated Content Agent, with its publishing calendar, monthly budget, publish windows, and review inbox, is listed for Business and Enterprise. Business also includes 10 AEO articles a month. If the client wants that pipeline on their own site, they need the plan.
A client that wants to own the account. Some in-house teams want the data, billing, and logins in their name, so nothing has to move if they ever change agencies. We then work inside their organization as invited users with project-level roles.
There's also the agency with exactly one GEO client at country level. Compare Kick-off ($199, 10,000 responses, 10M crawler logs, 10 seats) with Professional ($245, 18,000 responses, 25M crawler logs, 5 AEO articles, 2 seats). Professional buys more volume per dollar. Kick-off buys city tracking, Ads Radar, and room for client number two.
Where we'd start
Build the retainer around the pool, then the work. Estimate each client's response needs at onboarding, pick the agency tier that covers the roster with headroom, and write every allocation into the scope. Price the monthly fee around the program, and quote setup, net-new content, and awkward log connections separately.
We run this on Promptwatch because the shared agency pool is what makes per-client budgeting possible. One workspace, unlimited client projects, and Agent Analytics crawler logs plus visitor analytics on every client, so each monthly report can explain a movement instead of only recording it. If you want to compare notes on retainer structure, or hand the program to us, write to hello@1001seomedia.com.