AI Visibility Audit: How We Measure Your Brand in AI Answers
A published, repeatable method for checking whether ChatGPT, Gemini, Perplexity and Google AI answers mention, cite and recommend your business, and what to fix first.
- Free audit or in-depth audit
- Same method both times
- Honest about limits
Who the AI visibility audit is for
The audit suits UK businesses whose customers research before they buy and might ask an assistant to compare options: professional and B2B services, local and regional firms, and specialist retailers.
- You want a baseline before paying for any AEO & GEO work.
- You suspect assistants name your competitors instead of you.
- You think an assistant may be describing your business wrongly.
- You are comparing agencies and want to see how each one measures.
If your customers rarely research online, the audit will say so.
How we build the prompt set
The audit is only as good as the questions it asks. We write them from how your customers really ask.
Start from real customer questions
We take your services, your customer types and the questions they raise before buying, using sales calls, enquiries, reviews and search suggestions you can share.
Expand each topic
One head term becomes many follow-up questions: comparisons, prices, suitability, local variants and what to check before hiring.
Tag by customer type and stage
Each prompt is labelled by who is asking and whether they are exploring, comparing or ready to choose.
Fix the set
The final list is frozen so later checks use identical wording. Its size depends on the audit or plan, from about 25 prompts up to 120.
Neutral testing, so the result is fair
AI answers are shaped by account history, location and settings. To get a fair baseline, we remove as much of that as we can.
- Logged-out sessions with no history or saved memory, so the assistant is not tailoring answers to us.
- A stated location, such as Manchester, set the same way for every run.
- Web search switched on where the assistant offers it, because that is how most customers meet it.
- Repeated runs of each prompt, because the same question can return different answers.
- Date, platform and result logged for every run.
One run is a baseline, not proof. Repeat runs show which results are stable and which are chance.
What we measure: mentions, citations and share of voice
These three measures are often confused. We report them separately.
| Measure | Plain definition | Why it matters |
|---|---|---|
| Mention | The assistant names your business in its answer, with or without a link. | Customers can act on a name even without a click. |
| Citation | The assistant links to one of your pages as a source for its answer. | It shows which of your pages the assistant trusts and can send visitors. |
| Share of voice | Of the prompts checked, the share of answers that recommend you compared with your competitors. | It shows how you stand against the businesses customers are choosing between. |
We also record whether what the assistant says about you is accurate, because a wrong mention can cost you business.
Competitor orbit and Open, Contested or Saturated grading
Knowing who is recommended instead of you decides what is worth fixing first.
- Competitor orbit
- We sort the businesses that appear in answers into core competitors (direct rivals), mid-orbit businesses (partly overlapping) and outer-orbit sites (directories, publishers and comparison pages that fill the answer).
- Open
- Few strong businesses are named and the sources are thin. A clear, well-supported page from you has a realistic chance of being used.
- Contested
- Several credible businesses are named and answers shift between runs. Winning here takes better evidence and more consistent facts than rivals have.
- Saturated
- The same few strong names dominate every run. This needs the most effort for the least return, so we usually leave it until easier questions are won.
Grading each prompt this way turns a long list of results into a short list of priorities.
The four-layer diagnosis
When an assistant ignores or misdescribes you, the cause usually sits in one of four layers. We check each.
- Entity and brand facts
- Does the assistant know who you are, what you sell and where you operate? We compare what it says with your real details and note every wrong or missing fact.
- Site and crawl access
- Can search engines and AI crawlers reach, read and index your key pages? We check robots rules, indexing, speed and structure.
- Content answerability
- Do your pages answer customer questions directly, near the top, in passages that still make sense when quoted on their own?
- Off-site corroboration
- Do other sources, including your Google Business Profile, directories, reviews and relevant publications, repeat the same facts about you?
What you receive
- The prompt set, with each prompt tagged by customer type and stage.
- A results sheet showing mentions, citations and accuracy for each platform.
- Share of voice for each topic group.
- A competitor orbit table and an Open, Contested or Saturated grade for each prompt.
- A source map showing which sites the assistants relied on.
- A list of wrong or missing facts about your business.
- Findings from the four-layer diagnosis.
- A prioritised fix list, ordered by effort against likely return.
The free audit gives a shorter version of this. The in-depth audit gives the full set.
Limits and honesty
AI visibility is measured approximately. We would rather say so than oversell.
- Answers vary
- Results differ between users, locations, times and model versions. Any audit is a sample, not a census.
- No guaranteed inclusion
- We cannot promise any assistant will mention, cite or recommend you. We improve the conditions that make it more likely.
- Change takes time
- Fixes can take weeks to show, so we re-test on a schedule and avoid changing too much at once.
- Tools are approximate
- Analytics under-count AI referrals because many assistants send no referrer. We read referral data alongside prompt results and enquiries, never alone.
Our AEO & GEO services build on this audit, and our guide to measuring AI search visibility explains the ideas in more detail.
Free audit or in-depth audit
Both use the method above. They differ in depth.
| Area | Free AI visibility audit (£0) | In-depth SEO & AI visibility audit (£495) |
|---|---|---|
| Purpose | See roughly where you stand today | Full baseline and a plan you can act on |
| Prompt set | A smaller starter set | A larger set built around your customer types and stages |
| Competitor orbit and grading | Key competitors named | Full orbit table with Open, Contested or Saturated grades |
| Four-layer diagnosis | Headline findings | Detailed findings for every layer |
| Fix list | Top priorities | Full prioritised list |
Choose your audit
Start free to see where you stand, or go straight to the full baseline.
- Free AI visibility audit: £0.
- In-depth SEO & AI visibility audit: £495.
- Prices are current; we confirm terms in writing.
See how AI describes your business today
We will show you where you appear, where you do not, and what to fix first.