Try to find out what it costs to hire someone for AI search optimization and you will run into the same wall every buyer hits. The published guides quote ranges wide enough to park a truck in. The agency pricing pages say “it depends” and offer you a call. The one number you actually wanted, what will this cost me per month, is the one number almost nobody will print.
I find that ridiculous, so here is ours: $3,000 a month, flat. This post explains what that buys, why the rest of the market will not give you a straight number, what genuinely drives the price of this work up or down, and the spending traps to avoid no matter who you hire. If you read to the end you will be able to evaluate any quote you get, including ours.
Why a range is not a price
To be fair to the market, numbers do get published. The big cost guides quote agency retainers running from low four figures to five figures a month depending on business size, and a few firms print tiers or a “starting at” figure. But look closely at what you are being given. A range that spans a car payment to a salary is not information you can budget with. A tier tells you what the cheapest customers pay, not what you will. And “starting at” is a floor, which means the real number is still going to be discovered on a call, calibrated to what you can afford rather than to what the work costs.
That structure is not an accident. A wide range keeps every prospect in play, and moving the specific number onto a call is how the price gets fitted to your budget. It is not a scandal, it is how services get sold. But you should recognize the vagueness as a sales device, not a property of the work itself. The proof is that a flat number is perfectly possible to print, because you are reading one.
There is a second reason, less cynical: a lot of shops genuinely do not know what to charge because the discipline is young and their scope is undefined. If a vendor cannot tell you what they will do in month two, they cannot price month two either. The vagueness of the number usually tracks the vagueness of the deliverables, which is worth treating as a signal.
And there is a reason specific to this field for publishing your price anyway: AI engines answer “how much does X cost” whether or not you have published a number. I wrote about what happens to companies that hide their pricing: the engine fills the silence with a guess. It would be strange to sell the fix for that problem while hiding my own number and letting a machine invent one for me.
What our $3,000 covers
One flat rate, six-month initial term, then month-to-month. Every month it covers six published pieces built off a documented map of your ideal customers and the questions they ask AI engines, the off-site source work that citations actually come from, landing-page optimization so the traffic has somewhere useful to land, link-building outreach, and measurement against a fixed query set so you can see whether any of it is working. The whole process is public on the methodology page, and the full scope is on the offer page.
The reason a flat price is possible is that the scope is fixed. I sell one program at one intensity to one kind of company, B2B SaaS. I do not sell a bigger version to bigger logos. If a flat price sounds impossible to the agency you are talking to, it is because their scope is a negotiation.
What actually drives cost, anywhere
Strip away the packaging and the price of this work, from anyone honest, is driven by four things.
Content volume and depth. Most of the monthly effort is producing pieces engineered to be cited, which is slower than producing pieces engineered to fill a calendar. More pieces, more cost, roughly linearly.
The state of your source layer. Citations come from what credible third parties say about you. If the pages that currently write your AI answer are stale or wrong, someone has to find them and get them corrected, one relationship and one email at a time. A company with a clean source layer is cheaper to move than one with years of accumulated misdescription.
Category competition. Winning “best CRM” costs more than winning a precise question in a niche category, for the same reason it does in traditional search: more incumbents, more corroborated consensus to displace.
Measurement overhead. AI answers are unstable, so honest measurement means repeated runs across engines, which takes real hours. Anyone quoting you a price that includes “AI visibility reporting” for pennies is running one query once and calling it data.
Notice what is not on the list: software licenses, proprietary platforms, secret techniques. If a quote is heavy on those, you are paying for the vendor’s tooling, not your outcome. I wrote separately about when a tool is actually worth paying for, and the answer is later than the invoices suggest.
The four ways this work gets priced
Quotes will reach you in four shapes, and each one optimizes for something different.
A monthly retainer buys continuity, which happens to be what this discipline runs on: citations come from accumulation and refresh, not from a one-time push. This is the shape that fits the work, which is why it is the shape I sell.
A one-time project fits the pieces of the work that genuinely end: a source audit, a schema and entity cleanup, a batch of foundational pages. A project quote for “GEO” as a whole is a category error, because the outcome decays without maintenance and the source layer does not correct itself on a deadline.
Hourly consulting is the teach-your-team option, and for companies going in-house it is the best money in this list: a few hours of an experienced practitioner reviewing your query map and your first pieces beats months of guessing. It fails only when it becomes a standing arrangement, at which point you are paying agency prices without agency accountability.
Tools-only means software subscriptions and your own labor. Cheapest on the invoice, most expensive in hours, and it works exactly as well as your team writes. The trap, covered in the tool decision guide, is that the subscription feels like progress while the writing and outreach it is supposed to measure never quite starts.
What you should not pay for yet
Three line items show up on quotes constantly and deserve suspicion in the first six months.
An audit sold as a product. A real visibility audit takes an afternoon and I published the whole procedure free. Paying four figures for a PDF of findings, with the actual fixes quoted separately, is paying for the menu instead of the meal.
A monitoring platform bundled from day one. Measurement matters, but a tracking dashboard pointed at a program that has not produced anything yet reports zero with great precision.
llms.txt and schema packages priced as if they move citations. They are hygiene. They take hours, not months, and no engine has confirmed the file changes anything. Pay for them as hygiene or do them yourself; never as the headline deliverable.
When the right price is zero
Some companies should not hire anyone yet, including me. If you are pre-revenue, if your category is one where buyers genuinely do not ask AI engines anything yet, or if you have a marketer with writing time and the appetite to run a documented process, start in-house. The method is public. I wrote an honest breakdown of the agency-versus-in-house decision, and the summary is that an agency starts paying for itself at the point where consistency, not knowledge, is your constraint.
What I got wrong
We did not start at one flat price. We launched with founder pricing, raised it, and adjusted it again as we learned what the work actually costs to deliver well. More instructively, I spent time early on quoting scope-by-conversation like everyone else, and watched what it did to the conversations: buyers negotiated the scope instead of interrogating the outcome, and every call was about what could be trimmed rather than what would work. The flat number ended those conversations. The buyers it filters out are the ones who wanted to buy a discount, and the ones it keeps arrive already knowing the price, which means the first call is about their category and their sources instead of my rate card.
That is also the honest disclosure baked into this post: I publish my price partly because it is the right thing for buyers, and partly because it works. Transparency is a competitive move against an industry that will not match it.
The checklist for any quote
Whoever you talk to, make them name the number, the scope it buys per month, where the work happens (on your site or out in the sources that write your AI answer), what they will show you as evidence it is working, and what they refuse to promise. On the evidence question, accept nothing vaguer than repeated citation runs against a fixed query set plus AI referral traffic in your own GA4, in your account, that you keep if you leave. Anyone who guarantees citations on a schedule is guessing on your budget; the honest timeline has a flat stretch in it and a real vendor will tell you that before you sign.
If you want the version where the number is already printed and the method is already public, that is the offer. Either way, do not accept “it depends” as a price. It is not one.