Published UK prices for a one-off AI search audit run from £0 to £8,500. The paid ones we could verify on 10 August 2026 sit between £500 and £6,600. The spread has very little to do with quality, because the word “audit” carries no definition in this market: two providers can charge £800 and £6,500 for documents that do not resemble each other in length, method or output.
So the number on its own is unusable. What follows is every published UK price we could stand behind, then the specification that makes any of them comparable.
The wider decision this sits inside, including who sells the work and what the tools cost to run yourself, is on buying AI search work.
What is actually published
Method. We started with the ten UK agencies holding the first page of Google for “generative engine optimisation agency”, checked each one’s pages for a published figure, then added UK providers found by searching for published audit pricing. Every figure below was read from the provider’s own page on 10 August 2026 and is that provider’s price for its own service. Estimates of what “the market” charges are excluded, including the several pricing guides written by agencies about their competitors.
| Provider | What it is | Published price |
|---|---|---|
| MarGen | Free AI visibility audit | £0 |
| Tilio | Competitor snapshot | £500 |
| Findable | The Diagnostic, cross-channel | £500 |
| Tilio | AI visibility audit | £800 |
| SEO Strategy | Technical and content audit | £1,200 to £2,400 |
| SEO Strategy | Search visibility audit | £2,400 to £4,200 |
| MarGen | Synaptic audit | £2,950 |
| SEO Strategy | Strategic visibility assessment | £4,200 to £6,600 |
| MarGen | Synaptic enterprise audit | £4,500 to £6,500 |
| MarGen | Enterprise tier | £8,500 and up |
One provider in the set publishes a monthly figure rather than an audit price: Reboot Online states “our pricing starts at £2,500 per month” for its GEO work. Tilio also publishes a four-level monthly ladder from £499, £1,750 and £4,250 a month, with the fourth level priced on enquiry.
A correction to our own earlier writing, since it is the same dataset. Our UK agency comparison said two of ten publish a price. That was true of the ten GEO service pages we checked. It understates the picture: MarGen publishes a full price ladder on a separate pricing page that the service page does not link to. The accurate version is that two of ten put a price on the page a buyer lands on. At least three publish one somewhere on the site. We did not check every page of every domain, so treat the count as a floor.
Why almost nobody publishes one
Because ambiguity is what makes a salesperson necessary.
An audit with a published price has to have a published scope. A published scope can be compared with the one next to it. A quote produced after a discovery call can be sized to the buyer instead of to the work. Both are legitimate businesses. Only one of them lets you shop.
There is a second reason. It is the more interesting one. Nobody has settled what an AI search audit contains. A £500 document and a £6,500 document can both be honest, because one is a measurement and the other is a measurement plus a strategy plus an implementation plan. Until the format standardises, the price band is really a band of definitions.
What has to be in it before the price means anything
Seven things. Ask for them in writing before you pay, whatever the number is.
- The prompt set, plus who wrote it. How many buying questions, of what shapes, agreed before the work starts. If the provider chooses the prompts, the provider chooses the score. Fifty prompts written by someone who has never sold your product is worse than fifteen written with your sales team.
- Repetitions per prompt. These models are probabilistic. The same question asked twice returns different shortlists. One run per prompt is not a measurement, it is an anecdote with a table around it. Three is the working minimum.
- Which platforms, named. ChatGPT, Perplexity, Google AI Overviews, Google AI Mode, Gemini, Copilot. “Across AI search” is not a list. Coverage of one surface at the price of six is the most common way this gets sold badly.
- A citation count against a named index, with a date. A count of how often your domain is cited, sourced to a specific tool and a specific day, so it can be re-run and compared later. A proprietary score out of 100 cannot. Two vendors’ scores do not agree because they are not measuring the same thing.
- The same measurement applied to named competitors. Your own number in isolation tells you nothing. The gap is the finding.
- The cited sources, listed. Which pages the models actually leaned on when they answered. This is the part that tells you what to change. It is also the part most often left out, because collecting it is dull manual work.
- A ranked list of fixes, with the reason for the ranking. Not everything that is wrong. Three to five things, in order, each with what it would cost and who does it. An audit that ends in a hundred-item spreadsheet has moved the hard decision back to you and charged you for the privilege.
If a proposal covers those seven, the price is a fair comparison against another proposal that covers those seven. If it does not, you are comparing two different products that share a name.
What the free ones are
A lead magnet. No dishonesty in that, as long as everyone knows.
A free audit is almost always automated: a handful of prompts, a domain-level score, a PDF, then a call. The tell is in the same market data above. MarGen publishes a free AI visibility audit alongside a £2,950 one. Both are real. The free one is the thing that gets you onto the call where the £2,950 one is sold.
Take the free ones. Run three of them. Then notice that they disagree with each other, which is the most useful thing they will tell you.
What we charge
£500. It is the only fixed public price on this site. It buys the Findable Diagnostic: a cross-channel read led by AI search, covering the AI surfaces, the organic search underneath them, the paid search around them and where the page loses the visit once it arrives. Five working days. It ends in a ranked list of what to fix first with the reason for the ranking. If it does not give you specific prioritised findings tied to revenue, you get the £500 back.
Two things it deliberately is not. It is not the widest document on this page. At £500 against £6,500 it could not be. It is a decision aid, not an encyclopedia. And it is not delivery. Where the fix is ongoing paid search we sell that separately, on published limits and a published monthly price, so a diagnostic never quietly becomes a retainer you did not agree to buy.
The price is a founding price. It rises as the proof accumulates, which is a plain commercial statement rather than a scarcity tactic.
Five questions before you pay anyone
- How many prompts, how many repetitions, on which named platforms? Three numbers and a list. Any provider doing this properly answers in one sentence.
- Which index are you counting citations against, on what date? If the answer is a proprietary score, ask what it is computed from.
- Will you measure my two closest competitors on the same method?
- Will I get the list of sources the models cited? This is the part you act on.
- Does it end with a ranked list or a backlog? Ask how many items. Anything over ten is a report, not a decision.
Before you commission any of it, run the afternoon version yourself. It costs nothing. It takes half a day. It is the only way to know whether the paid version is telling you something you did not already know. If you want the vocabulary first, start with what AI visibility means and how to measure it. If the spend you are auditing is paid search rather than AI visibility, the ten checks a PPC audit runs, in the order they bite is the equivalent page for that market, where the published prices are just as scarce.
Data in this article: every price read from the provider’s own page on 10 August 2026 (tilio.co.uk, margen.net, seostrategy.co.uk, rebootonline.com). Keyword figures from Ahrefs keywords-explorer-overview, GB, the same date: “geo audit” 200 searches a month at difficulty 5, “ai visibility audit” 150 at difficulty 1, “seo audit cost” 150 at difficulty 2, “ai search audit” 60 at difficulty 7. One US provider found during the search is excluded from the table because its prices are in dollars.