Executive scorecard
A headline AI Visibility Score out of 100, built from four components: visibility, sentiment, accuracy and competitive standing. Plus a per-engine breakdown, because ChatGPT and Perplexity rarely agree about you.
Flagship · Paid audit
A measured plan for AI search, not a vibe.
Fifty prompts a real prospect in your market would ask, run across ChatGPT, Claude, Gemini and Perplexity, plus Google's AI Overviews on 20 tracked keywords. Mentioned where, ignored where, confused with whom, with the evidence for every claim. Then a prioritised plan, ranked by impact, written in plain English. Delivered within two weeks. £4,000.
Why it exists
The free check tells you whether you have a problem. This tells you what to do about it.
Three prompts is enough to know if AI assistants are missing you. It isn't enough to know why, which competitors are taking your share of voice, where your entity is getting confused with someone else's or what the actual lever is to fix it. The AI Visibility Audit is the full survey: enough prompts to be meaningful, enough engines to mirror real customer behaviour, enough analysis to give you a plan rather than a problem.
What you get
One designed report. Every claim backed by evidence.
The report follows the structure of the audit engine that produces it, and every number in it can be traced back to a verbatim model answer. These are the sections.
Topic-category gaps
Every prompt belongs to a buying conversation. The report ranks them weakest first, so you can see exactly which questions your prospects ask where you simply don't exist.
Google AI Overview surface
Twenty tracked keywords: whether an AI Overview appears, whether your domain is cited as a source and which competitors Google is citing instead. Often the single clearest picture of where you stand in Google's new answer layer.
Risk flags, with evidence
Every case where a model invented facts about you or described a different company under your name, quoted verbatim and checked against a verified brand profile. Often the fastest wins in the entire plan: specific, provable and fixable.
Prompt-level verdicts
Every prompt, every engine's answer, word for word, with a fully judged rationale beside it. The auditable proof behind every score, and the section sceptics read first.
Prioritised action plan
Ranked moves, worst problems first: entity disambiguation, corrections to the record, schema and llms.txt fixes and the content gaps that will move the score most. Written so a marketing manager could brief it and a developer could execute it.
Proof
Real numbers from real reports.
Both examples below are genuine 2026 audits produced by the same engine that will produce yours. The reports are confidential, so the names are removed. The numbers are untouched.
A UK logistics brand · scored 20.6/100
Fifty-five prompts, four engines. One engine never mentioned them: 0 of 55 answers. Another managed 6 of 55. The audit surfaced 30 risk flags: models invented depots, fleet details and annual volumes, repeatedly got the founder's name wrong, and one engine confidently described a completely different company overseas under the brand's name. In Google's AI Overviews the brand's domain was cited for 0 of 20 tracked keywords, while a competitor was already being cited. The action plan ranked eight moves, starting with entity disambiguation and public corrections of the invented facts.
An independent serviced-office brand · scored 49.2/100
A smaller 30-prompt baseline across the same four engines, with 16 risk flags. Asked what the brand was, two engines described a jet-charter business instead. Models invented office locations that don't exist, made up pricing, fabricated client testimonials and flatly denied memberships the brand verifiably holds. Perplexity mentioned them in 24 of 30 answers; the weakest engine in 17. The baseline now re-runs on a schedule, so the remediation shows up as a trend rather than an anecdote.
Every finding like these appears in the delivered report next to the model's verbatim answer and the evidence check that caught it. That's what makes the plan arguable with a board and briefable to a developer. If you'd like to see a full anonymised report before you buy, ask.
How it works
Two weeks. Four steps. No surprises.
The process is deliberately compact. You give me about an hour at the start, an hour for the walkthrough at the end, and the rest happens off your team's plate.
01 · Brief
A 60-minute kickoff
One call to confirm your top customer segments, your real competitors (not the LinkedIn-aspirational ones) and the buying questions your prospects actually ask. Everything downstream is calibrated to this.
02 · Measure
Fifty prompts, five surfaces
I design 50 buyer-intent prompts, run them across ChatGPT, Claude, Gemini and Perplexity, capture the verbatim output and measure Google's AI Overviews across 20 tracked keywords. The discovery prompts never include your brand name, otherwise you've just paid me to confirm you exist. A handful of direct prompts do name you, deliberately, to catch entity confusion when an engine is asked about you outright.
03 · Analyse
Citations, hallucinations, entities
Mentions counted, competitor share calculated, every model claim checked against a verified brand profile, hallucinations and entity confusions surfaced with evidence, schema and llms.txt reviewed. Then a ranked plan, worst problems first.
04 · Walkthrough
A 60-minute readout
A live walkthrough of the report with whoever needs to be in the room: founder, marketing lead, SEO lead, the developer who will end up doing the schema work. Questions answered. Plan handed over. Done.
After the walkthrough
What usually happens next.
Most clients leave the walkthrough knowing which two or three moves will actually shift the score, and whether to run them in-house or with me. There are three natural next steps, and the audit is designed so you can pick any of them, or none, and run the plan yourself.
Take the plan in-house
You have a developer and a content lead, the plan is clear and the priorities are obvious. Take the report, brief your team, run it. A perfectly good outcome. I'd rather you keep your budget than book follow-on work that doesn't earn its keep.
Remediation Sprint with me
From £4,000, four to six weeks. I implement the prioritised plan: schema, llms.txt, entity disambiguation, content scaffolds, working sessions with your team for the items they own. The gap between "we should do this" and "we have done this" is the bit that so rarely gets bridged and this closes it.
Monthly Re-measurement
£750 a month, billed quarterly, same prompt set, same engines, tracked against your baseline. Your first quarter carries a £1,000 credit from the audit fee. Whichever route you pick for remediation, monthly measurement is how you know whether it's working.
Pricing
Clear price, clear scope.
AI Visibility Audit · £4,000
One brand. Three competitors. 50 prompts across four engines, plus Google's AI Overviews. Designed report, 60-minute walkthrough, delivered within two weeks. Includes a £1,000 credit against your first quarter of re-measurement.
Remediation Sprint · from £4,000
I implement the prioritised plan over four to six weeks. Schema, llms.txt, entity disambiguation, content scaffolds, the actual technical lift. Optional, never assumed.
Monthly Re-measurement · £750/mo
Re-run of your full prompt set every month, billed quarterly at £2,250. Visibility movement against baseline, competitor share shifts, new risk flags. An audit from Q1 is stale by Q3; this is how it stays true.
Custom scopes available for larger brands, multi-region footprints, expanded prompt sets or additional competitors. Quoted in writing before any commitment. Full rates for everything else are on the pricing page.
Best for
Who actually benefits from this.
Established brands
You've been investing in SEO for years and you've noticed AI assistants don't seem to mention you. Either they should, or they shouldn't, and you'd like to know which.
Founders being asked the AI question
The board keeps asking about your "AI strategy". A measured baseline is a much better answer than a deck, and a remediation plan is a much better follow-up than a hire.
Brands with messy entities
Common name, M&A history, similarly-named competitors, several products wearing similar hats. Entity confusion is almost certainly costing you mentions, and the audit will prove it.
FAQ
Fair questions about the AI Visibility Audit.
How is this different from the free check?
The free check is three prompts across three models, designed to take 30 seconds of your time and tell you whether you have a problem worth caring about. The AI Visibility Audit is the full survey: 50 prompts across four engines plus Google's AI Overviews, with competitor citation analysis, hallucination evidence, entity-confusion checks, schema review and a prioritised plan. The free check is diagnosis. This is treatment planning.
Which AI engines are covered?
ChatGPT, Claude, Gemini and Perplexity, at the model tiers your prospects actually use, plus Google's AI Overviews measured across 20 tracked keywords. Model versions are kept current as the vendors ship new ones; the scoring method stays the same, so your numbers remain comparable from one measurement to the next.
How long does it take?
Two weeks from kickoff to walkthrough. Week one is prompt design and measurement, week two is analysis, write-up and the live readout. Faster turnarounds are sometimes possible for an additional fee, but two weeks is the realistic floor for work this thorough.
What do you actually deliver?
A designed HTML report covering an executive scorecard, per-engine breakdown, topic-category gaps, the Google AI Overview surface, risk flags with evidence, prompt-level verdicts with every model answer verbatim and a prioritised action plan. Plus a 60-minute walkthrough with whoever needs to be in the room.
Do I have to implement the plan with you?
No. The audit is designed to be useful regardless of who implements. Many clients hand the plan to their in-house team or existing agency and don't engage me again until the next re-measurement. Many others ask me to lead the remediation sprint. Both are fine, and the audit isn't priced as a loss-leader for the next thing.
What if I just want a single competitor benchmarked or fewer prompts?
The £4,000 price is set deliberately, because it reflects the actual work. Smaller scopes don't reduce the analysis effort enough to justify a different price. If budget is the issue, the free LLM visibility check is yours, and a single AI Visibility Audit done properly tends to be worth more than two cut-down ones done cheaply.
How do I book and pay?
Pay by card with the "Book the AI Visibility Audit" button and I'll be in touch within one working day to set the kickoff call. If your finance team prefers an invoice, or there's anything unusual about your scope, book an intro call instead and I'll quote it in writing. No 45-minute discovery-call hostage situation, promise.
Why isn't Microsoft Copilot included?
Because Microsoft doesn't offer public access to it. Consumer Copilot has no public API: the official Copilot APIs are enterprise tools that report on your own Microsoft 365 tenant, not the public answer surface, and the only way to capture what Copilot tells the public is scraping, which is brittle and against Microsoft's terms. Rather than sell data I can't stand behind, the audit measures the engines that can be queried properly. Copilot also runs on the same OpenAI model family as ChatGPT, with Bing grounding on top, so ChatGPT coverage is the closest honest proxy. The moment Microsoft opens a legitimate route, Copilot joins the prompt set.
Make your brand the answer.
Book the audit, or check the free one first.
The AI Visibility Audit is the right call if you already know AI search matters to your business and you want a measured plan. The free check is the right call if you're not sure yet. Both are real options.