AI Visibility Audit: What It Includes and Costs (UK)

TL;DR

An AI visibility audit is a structured, evidenced test of what assistants such as ChatGPT, Gemini and Perplexity actually say when you put your customers' real questions to them. A good one covers several engines, repeats the questions on different days, keeps a dated record of every answer, checks accuracy and the technical plumbing, and finishes with a prioritised fix plan and a baseline score you can measure against. UK prices run from a couple of hundred pounds to several thousand, so judge an audit by what is inside it, not by the figure on the invoice.

Most UK business owners have never actually seen what ChatGPT, Gemini or Perplexity say about them. Not because the answers are hidden, but because nobody has sat down and looked in an organised way. An AI visibility audit is that structured look. In this guide I will walk you through what a real audit contains, section by section, what it ought to cost over here, the warning signs that tell you a report is repackaged SEO rather than genuine work, and five checks you can run yourself in the next 20 minutes without paying a soul.

What is an AI visibility audit?

Strip it back and an AI visibility audit is one thing: a structured test of what AI assistants say when your customers' questions are put to them, with the evidence attached. You take a fixed set of the questions real customers ask, run them across engines such as ChatGPT, Google's AI results, Perplexity, Gemini, Claude and Copilot, record what comes back, and turn the pattern into a plan you can act on in order.

You will hear it called an AEO audit (answer engine optimisation) or a GEO audit (generative engine optimisation). The label barely matters. The job underneath is always the same: work out whether AI assistants mention you, whether they recommend you, whether they describe you accurately, and whether they can even reach your website to begin with.

It is easy to underrate what is at stake here. These assistants do not hand you ten blue links to scroll through. They name a few businesses and quietly drop everyone else, and the gap between being named and being ignored is brutal.

1.2%

of local business locations get named by ChatGPT when it answers (SOCi Local Visibility Index, 2026)

The same study found Gemini recommending 11% of locations and Perplexity 7.4%, while those same brands surfaced in Google's local three-pack 35.9% of the time (SOCi Local Visibility Index, 2026). Read that gap carefully: ranking well on Google does not carry you into AI answers, and that is precisely why AI visibility needs a test of its own. For the wider picture, see our UK guide to AI search optimisation.

What a proper audit includes

Eight things separate a proper AI visibility audit from a thin one: a question set built from real customer language, the same questions run across several engines more than once, a dated record of every answer, a citation map covering you and your rivals, an accuracy check, technical checks on crawler access and indexation, a prioritised fix plan with effort estimates, and a baseline score you can re-test against down the line. Take any one of these away and you are left with a lesser thing.

A fixed question set built from real customer language. The questions should come out of enquiry emails, reviews, sales calls and quote requests, not a keyword tool. "Who is the best emergency electrician in Leeds that answers at weekends" is an audit question. "Electrician Leeds" is not.
The same questions across multiple engines, more than once. AI answers wobble from one session to the next, and from one day to the next. A single run only tells you what one engine happened to say on one occasion. Repeating the questions is how you tell a consistent finding apart from noise.
Dated evidence, traceable. Every claim in the report should trace back to a dated record of the exact answer the engine gave. If you cannot see the raw output for yourself, there is nothing else in the report you can actually verify.
A citation map. Where you get named or cited, where your competitors do, and which sources the engines keep leaning on, whether that is directories, review platforms or press coverage. This is what shows you where the fixes actually need to land.
An accuracy check. What the AI gets plain wrong about you: an old address, a service you dropped years ago, the wrong prices, a mix-up with a similarly named firm, or details it has simply invented.
Technical checks. Whether AI crawlers can reach your site at all, through robots.txt and your CDN or firewall settings, whether Bing has you indexed, and whether your schema markup is present and correct. Plenty of sites block AI crawlers without knowing it, usually through a default CDN setting nobody remembers switching on.
A prioritised fix plan with effort estimates. Specific actions, in the order you should tackle them, with an honest read on the hours each one takes. Not vague "phases".
A baseline score. A repeatable measure, so that when you re-run the same question set a few months later you can see whether the needle moved instead of guessing at it.

Five free checks you can run in 20 minutes

You do not have to pay anyone to get a rough sense of where you stand. Five checks, using tools already sitting on your desk, will tell you whether AI assistants can reach your site, whether they mention you, and what they say when they do. Block out 20 minutes and write down everything you find as you go.

Look up your robots.txt. Go to yourdomain.co.uk/robots.txt and search the page for GPTBot, ClaudeBot, PerplexityBot and Google-Extended. A "Disallow: /" under any of those is you turning that assistant's crawler away at the door.
Ask ChatGPT your top customer question. Open a fresh chat and type the question a new customer would, something like "best independent accountant for small limited companies in Bristol". Note who gets named, and whether you are anywhere in the list.
Ask about your business by name. "What do you know about [Business Name] in [Town]?" Nothing catches inaccuracies faster: wrong opening hours, a phone number from two premises ago, services you stopped offering years back. It is the quickest way to hear what ChatGPT actually tells people about you.
Run site:yourdomain.co.uk on Bing. A number of assistants pull their web results from Bing's index. If Bing surfaces only a handful of your pages, or none at all, that hole feeds straight through into what the AI can say.
Read your Google Business Profile as a stranger. Google's AI results and Gemini both lean on it hard. Look over the categories, description, photos, reviews and Q&A as though you had never heard of the company. If an AI summarised this profile cold, would it get you right?

If those checks turn something up, our guide on getting recommended by ChatGPT in the UK walks through what to do about each one.

What should an AI visibility audit cost in the UK?

For a one-off audit in the UK, expect anything from a couple of hundred pounds to several thousand. Published US prices run the same span, from $99 automated reports up to $5,000 and beyond for a full agency audit. The number on the invoice tells you almost nothing on its own. What tells you something is exactly what you get for it.

This is a young market and the range really is that wide. Over in the US, automated self-serve reports go for as little as $99 (Optimum Web, 2026), while agency audits typically start around $500 for a basic automated scan and climb to $5,000 or more for proper hands-on manual analysis (Loved by AI, 2026). Ongoing AEO retainers sit higher still, with starter packages from roughly $2,500 a month (AI Advantage Agency, 2026). UK pricing follows the same shape, just in pounds.

Which is why you should compare contents, not numbers. Put the same questions to any provider: how many engines, how many questions, do you run them more than once, do I get a dated record of the exact answers, are the technical checks in there, and is the fix plan written for my site or lifted from a generic checklist. An audit that fires 10 questions once through a single engine is not a budget version of the real thing. It is a different product wearing the same name.

One principle I will not bend on: a fair one-off audit for a small business should never make you sign a retainer just to read your own results. You should be able to buy the audit, receive the whole thing, and then decide what comes next knowing exactly where you stand. That is how ours works. One fixed published price, everything above included, and the report is yours to keep whether or not you ever spend another pound with us.

Red flags when buying an audit

Five things should send you looking elsewhere: guaranteed placements in AI answers, llms.txt sold as a headline deliverable, schema-only packages, a report that never shows you what the engines actually said, and pressure to sign a long contract before you have seen a single result. Every one of them is the mark of a provider selling process instead of evidence.

Guaranteed placements. Nobody can promise that ChatGPT or Gemini will recommend you. The answers are generated fresh every time, they shift from run to run, and no agency has a back door into any model. What an honest provider offers is better odds and a way to measure the change, not a guarantee.

llms.txt as a headline deliverable. In June 2026 Google updated its Search Central documentation to say llms.txt files are not needed for Search, AI features included, and that they neither help nor harm your visibility (Google Search Central, 2026). Google's John Mueller has likened the file to the long-abandoned keywords meta tag (Search Engine Journal, 2026). It takes a few minutes to create and does no harm, so add one if you like. Just do not pay anyone for it.

Schema-only packages. Ahrefs tracked 1,885 pages that added schema markup and watched AI citations barely twitch: about +2.2% on ChatGPT and +2.4% on Google's AI Mode, close enough to zero to be random noise, next to a statistically significant 4.6% drop in AI Overviews (Ahrefs, 2026). Schema earns its place in the technical checks of an audit. Sold on its own, it is not a fix plan.

No sight of the answers. If a report hands you a score, a grade or a tidy traffic-light dashboard but never shows a dated record of what the engines actually said, you are being asked to take the whole thing on faith.

Contract pressure. An audit that only makes sense as the thin end of a 12-month agreement is a sales tool dressed as a diagnosis. Take the audit, read it properly, then decide in your own time.

Audit or monitoring tool: which do you need?

Monitoring platforms like Profound and Peec watch AI answers continuously, which suits brands with a standing budget to point at it. Peec's published plans open at around €89 a month and climb steeply as you add coverage (Peec, 2026). A one-off audit is answering a different question for a small business, and a simpler one: where do you stand right now, and what should you fix first.

Those tools are built for teams who already know their baseline and want to watch it move week by week, and Profound in particular is priced and featured for enterprise marketing departments. For most UK small businesses the sensible order runs the other way round. Get the one-off audit, work through the prioritised fixes, then re-test. The day AI channels start sending you real enquiries is the day a monitoring subscription begins to pay for itself, and not much before.

Questions people ask

These are the questions UK business owners put to us most often when they are weighing up an audit. Each answer below is written to stand on its own, so you can act on it whether you end up buying an audit from us, buying one from someone else, or running the checks yourself.

How long does an AI visibility audit take?

Most one-off audits take one to two weeks from kick-off to delivery. Repeat runs are the main reason: AI answers vary between sessions, so a credible audit asks the same questions on different days rather than in a single sitting. A same-day report has usually run each question only once.

Is an AI visibility audit the same as an SEO audit?

No. An SEO audit examines how your site performs in traditional search rankings. An AI visibility audit tests what assistants such as ChatGPT and Gemini actually say and recommend in conversational answers. The two overlap on technical foundations, but strong Google rankings do not reliably translate into AI mentions.

Can I do an AI visibility audit myself?

You can run a useful DIY version with the five free checks in this guide, and for some businesses that is enough to surface the obvious problems. A paid audit adds scale and rigour: more engines, more questions, repeat runs, dated evidence and a prioritised plan, which is hard to replicate by hand.

How often should you re-test AI visibility?

Re-testing every three to six months is a sensible rhythm for most small businesses. AI models and the sources they draw on change frequently, so a baseline score from your first audit only stays meaningful if you re-run the same question set later. Re-test sooner after significant fixes, such as unblocking AI crawlers.

Sources

Want the audit done properly?

One fixed price. Six engines, repeat runs, a dated record of every answer, technical checks and a prioritised plan. No retainer required to see your results.

Get started

What Is an AI Visibility Audit? What It Includes and What It Should Cost