What AI Visibility Actually Costs in 2026 — and Why the Same Work Is Quoted at $99 and at $15,000
Ask four vendors what it costs to get your business showing up in AI answers and you will get $99 a month, $2,500 a month, $8,000 once, and "let's set up a call". All four will describe the work in nearly the same words.
That spread is not a market still finding its level. It is four genuinely different products sold under one name, and once you can tell them apart the pricing stops looking irrational. Here is each band, with numbers as published at the time of writing — verify them, because this market repriced twice while I was writing about it.
What does a $29–$99 per month tool buy?
A prompt sampler, and only that. You give it a list of questions, it asks the engines on a schedule, and it reports whether you were mentioned.
As published: Otterly.AI starts around $29/month. Profound publishes $99/month Starter and $399/month Growth, both billed annually. Peec AI runs roughly $95 Starter, $245 Pro, $495 Advanced. What each of those buys against doing the same measurement on your own server is broken down page by page: the Profound alternative and the Peec AI alternative, including the cases where buying the tool is the correct answer.
What you should know before treating that as cheap:
- Prompt allowances are small. 50 tracked prompts on one Starter tier, 150 on Pro. A local service business with six services across four cities exhausts that immediately.
- Engine coverage is a plan feature, not a given. One major tool tracks three engines at once out of six on standard plans, with extra models as paid add-ons, and puts Claude behind Enterprise. My own server log recorded Claude fetching a client site 129 times in eight days while answering real people. A Starter customer would have been shown zero for that engine — not "unmeasured", just zero.
- The output is a sample. Change the prompt list and the number changes. I have watched a frozen metric in one panel fall from 194 to 25 while receiving no new data, because the vendor was recomputing its index. Small month-over-month moves in these tools are the instrument, not your site.
This band is worth buying. It is not worth mistaking for a measurement of reality, and it tells you nothing about whether anyone called you.
What does a one-off audit cost?
Published ranges for a one-off generative-engine-optimization audit run roughly $5,000 to $15,000 at agency grade. Some engagement-grade AI visibility audits are quoted higher.
What separates a good one from an expensive checklist is whether anything was actually measured. Most audits in this category inspect your pages, compare them against a list of best practices, and infer what AI probably thinks of you. That is a review, not a measurement.
Four things can genuinely be measured, and they need four different instruments:
- Whether AI agents can fetch your pages at all — per agent, with the rendered word count each one receives.
- What assistants are already pulling from your site — only visible in your own server log.
- Whether you own anything an engine would quote — an inventory, not an opinion.
- Whether you could tell if AI sent you a customer — the gap in your attribution chain.
My own version is published at from $1,500, includes the logger install so layer two is measured rather than estimated, and leaves the log on your infrastructure. I am not going to pretend that is the market rate; it is below it, deliberately, because the audit is how someone decides whether to work with me.
What does a monthly retainer cost, and what makes it worth it?
Agency retainers in this category are commonly quoted at $2,500 to $10,000 a month, with enterprise engagements well above that. My published tiers are $2,500 and $4,500.
The honest question is not the number, it is what recurs. A retainer earns its keep when three things happen every month:
- New citable pages enter the set. This is the actual lever and almost nobody reports it. In one six-week series, AI responses citing a client climbed from 63 to 2,605 while the pages-cited count went 13 → 14 → 17 → 22 and then stopped dead. Responses saturate against a fixed set of pages. Growth resumes only when new pages become quotable.
- The measurement is rerun, not restated. A monthly report that reproduces last month's screenshot with a new date is not a measurement.
- The attribution chain keeps holding. Tracking breaks silently. A first-touch mark that stopped surviving brand searches three weeks ago will still produce a confident-looking report.
If none of those three is happening, you are paying rent on a dashboard.
Why is the same work quoted at $99 and at $15,000?
Because the price is set by which of these four things is being sold, and the four are rarely named:
| What is really being sold | Typical published price | What it can tell you |
|---|---|---|
| Prompt sampling software | $29–$495 / month | Whether you appeared in its sample |
| A page-level review | $5,000–$15,000 once | What best practice says you should change |
| Content production at volume | $2,500–$10,000 / month | Whether the citable set is growing |
| Measurement and attribution build | project or retainer | Whether any of it produced revenue |
The first and second are what most buyers get when they think they are buying the fourth. That is the whole confusion, and it explains both ends of the spread honestly — nobody is necessarily overcharging, they are selling different things.
What should a local service business actually spend first?
In this order, and the order matters more than the amounts:
- Nothing, until you know whether AI reads your site. Server-side logging is an afternoon of work. It either shows agents pulling your pages or it does not, and both answers change what you should buy next.
- One data page, built from records you already own — job history, costs, failure rates. In two site logs the pages assistants actually pulled into answers were overwhelmingly pages made of numbers; service and location pages were crawled and then never consulted. This costs nothing but writing time.
- The cheapest tool that covers your engines and returns citations rather than just mentions. Check the engine list against what your log shows is actually reading you.
- Attribution, before a retainer. Without it a monthly engagement cannot be judged, and in home services the money moves over the phone where no analytics product can follow it.
- A retainer, if and only if steps one to four showed something worth compounding.
Most businesses do this backwards: retainer first, measurement never.
One exception worth naming: if you buy paid leads from Google Local Services, auditing what you were charged for pays back faster than anything on this list, because it recovers money already spent rather than chasing new demand. That is LSA recovery.
How do you tell whether a price is honest?
Three questions. They work on me too.
"What would a failure look like in your reporting?" If nobody can describe how their instrument would show the work not working, you are not buying a measurement.
"Which engines are in the number, and which are excluded?" Excluded engines get reported as absence. Ask for the list in writing.
"Is the price on your website?" In a category where one-off audits span $5,000 to $15,000 and retainers span $2,500 to $10,000, a vendor who will not publish a floor is asking you to negotiate without a reference point. Mine are on the pricing page, including the parts I think most businesses should not buy from me.
Above these bands the question changes shape. Enterprise platform contracts are quoted at $10,000 to $70,000 a year with nothing published at all, and the spread between two such quotes is usually governance features rather than measurement quality — that arithmetic, and how to test your own quote against the one public reference price in the category.
If the band you land in is a person rather than a subscription, the same three questions work in a hiring conversation — what to ask when hiring an AI SEO specialist puts them in that order, and the terms I work under are written out for the version where that person is me.
If you want to know which band you are actually in before anyone quotes you anything, the four-layer audit is the diagnostic — and the open dataset shows the kind of measurement it produces, on real logs, for free.
I publish my prices because in this market that is the differentiator. Engagements and what they include are on pricing; the measurement instrument is described on the Citation Tracker page. If you want a second opinion on a quote you have been given, let's talk — that costs nothing.
Related: