Generative engine optimization is priced the way SEO was priced fifteen years ago: mostly in private, after a call. The short version for 2026: a manual audit at ours costs nothing, agency retainers start in the low thousands per month and climb with markets, brands and implementation scope, monitoring tools cost a subscription well under a retainer, and doing it in-house costs salaries you already pay plus the risk it stalls. Ours are two numbers: the audit is complimentary, and the full-cycle retainer starts from $3,000 per month (September 2026), with the final figure set after the audit. The canonical version of that is on the pricing page; this post explains what is behind it and what to make of everyone else's silence.
what you are pricing
GEO is one loop rather than a menu: measure what six engines answer about you, fix what stops them reading you, build the entity they can verify, publish what they can quote, get into the sources they cite, then measure again. Any price that looks small has usually removed one of those stages, and the removed stage is almost always the slow one, authority, or the honest one, re-measurement. So before comparing numbers, compare what the number buys. The seven stages are on the services page.
what moves the price
Four things, in roughly this order of weight, and every one of them is something you can estimate yourself before talking to anyone.
| driver | why it costs more |
|---|---|
| markets and languages | Every market needs its own prompt panel and its own baseline, because the answers differ per market and so do the sources the engines cite. The single biggest multiplier. |
| brands and domains | Entity work is per entity. Two products under one company is not the same job as one, and a portfolio of mirror domains is harder than either. |
| who implements | Where the agency hands specifications to your developers, the retainer is lower. Where it does the technical work end to end, it is higher. Neither is better; it depends on your team's capacity. |
| the state you start from | A site an AI crawler cannot fetch needs weeks of work before any interesting layer matters. In our dataset that was half of all sites (50% unreadable to AI crawlers, n = 100, April to July 2026). |
A useful test of any quote is whether the agency can name which of these four is driving it. If the price arrives without that breakdown, it was set by what the agency thought you would pay.
what the market charges, when it says
Very few agencies publish a number. In our ten-agency review of the category, one of ten discloses pricing (best GEO agencies in 2026, measurements dated 1 August 2026), and that one is us. So the market range has to be assembled from guides rather than price lists: the 2026 pricing guides that ChatGPT quoted in our 31 August 2026 panel run put lower-end agency retainers around $2,000-3,500/month, and the upper end is open-ended, because enterprise engagements with multiple markets and full implementation are quoted per scope and rarely disclosed at all.
Two cautions about that range. First, the lower end often describes a content or monitoring package rather than the loop: check whether re-measurement, entity work and citation work are inside it. Second, treat quotes far below it with the same suspicion as quotes far above it. The work is manual, a prompt panel across several engines has to be run and read by a person, and manual work has a floor.
the three purchases that get confused
Most price comparisons in this category compare three different things. Doing it yourself costs developer and writer time you already pay for, plus the risk that measurement has no owner and stops around week six. A monitoring tool costs a monthly subscription, typically well under a retainer, and buys visibility rather than change: it tells you that you appear in nine percent of answers, not what to fix first. An agency retainer buys the decisions and the implementation as well as the measurement. Which one you need depends on whether your constraint is capacity, visibility or ownership; the honest walk-through, including the cases where a tool plus your own team is the cheaper correct answer, is in agency, tool or do it yourself.
our public price, and why it is public
| engagement | price |
|---|---|
| manual GEO audit | complimentary, for the requests we take on. A prompt panel for your category run live through ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek, every answer logged with its sources, a health score out of 100, and a dated 90-day plan. Yours to keep. |
| full-cycle retainer | from $3,000 per month; the final figure is set after the audit. Bi-weekly sprints, the same panel re-run at every review and reported as a delta. |
| scoped technical implementation | quoted per scope, for teams who have had the audit and want the technical layer shipped without the loop. The weaker option, and we say so. |
The floor assumes one brand, one market, one language and a team that can absorb some of the implementation. Each driver above pushes it up. There is no hourly rate, because almost none of the value is time at a keyboard and an hourly rate rewards the slowest version of the job. There is no revenue share, because attribution between an AI answer and a signed deal is not clean enough for either side to trust. There is no long lock-in: each monthly cycle closes with a re-run of the panel, which is deliberately the easiest moment to leave.
We publish the numbers because the category does not, and because a buyer who cannot self-qualify has to sit through a sales call to find out whether the conversation was ever realistic. Publishing also has a measurable side effect: on our own weekly prompt panel, run 31 August 2026, ChatGPT cited our pricing page as a source when asked what a GEO agency costs. A page with a number on it gets quoted; that is the same mechanism we sell.
what should never be billed separately
- The audit. Ours is complimentary because we need the baseline to scope the work at all; an audit fee is a fee for the agency's own quote.
- Re-running the panel. A method built on measurement cannot bill measurement as an add-on. If re-measurement is an extra line, the base retainer is not a GEO retainer.
- The published research. Our 100-audit report and its dataset are open, including for people who will never be clients.
- Placement guarantees sold before a baseline exists. A commitment to specific target prompts is honest only after the audit has measured where you stand; we give one, scoped that way, and never from a pricing page.
how to judge a quote in five minutes
- Ask which of the four drivers set the number. No breakdown, no basis.
- Ask for the prompt panel and the engine count. No panel, no measurement, and you are buying vocabulary.
- Ask what arrives each sprint. Files (llms.txt, JSON-LD, server config, page briefs, the logged panel) or slides.
- Ask what happens if the numbers do not move. Ours: we show the flat line, say which layer failed, and two empty reviews are a reason to stop rather than re-sell.
- Take the audit first. Whoever you hire, the baseline is the thing every option needs, and at ours it costs nothing.