AI Visibility Audits: Find Out Who the Engines Recommend Instead of You
We put the questions your buyers ask to the engines they ask, and record what comes back. Usually it is a competitor.
Delivered in working days, not weeks. You keep the question set and use it as your benchmark afterwards.
An AI visibility audit measures whether AI engines name your brand when buyers ask questions in your category, and explains why they do not on the Generative Engines and is a part of GEO.
We build a question set from real buying behaviour rather than brand-name lookups, run it across ChatGPT, Gemini, Perplexity and Google AI Overviews, and record three things: who gets mentioned, who gets cited as a source, and which domains the engines lean on for your subject. The output is a diagnosis rather than a score, so every absence maps to a cause: missing content, weak entity signals, no third-party corroboration, or a crawler you are blocking. Delivered in working days, not weeks.
An audit is not a dashboard
A measurement. Useful, and available cheaply.
If you want the measurement running continuously afterwards, that is citation tracking, and it is a separate and cheaper thing.
METHOD
How we actually do it
Measured, not guessed.
Build the question set from real buying behaviour.
Not brand-name lookups, which flatter everyone. The questions a buyer asks before they know who you are: category comparisons, "best X for Y", requirements, alternatives, objections.
Run it across the engines that matter.
ChatGPT, Gemini, Perplexity and Google AI Overviews behave differently and cite differently. We test each rather than assuming one stands for all of them.
Record mentions, citations and sentiment.
Being mentioned and being cited are different outcomes with different fixes. We log which sources the engines pull from, because those sources are the route in.
Turn the gaps into a plan.
Every absence maps to a cause: missing content, weak entity signals, no third-party corroboration, or a crawler you are blocking. The plan feeds the GEO programme and ongoing citation tracking.
HOW WE SCORE IT
AI share of voice
Share of voice is the proportion of AI answers, across a fixed question set and several engines, in which your brand appears, measured against your competitors. It is the number every audit reports. The diagnosis explains it.
Share of questions where your page is listed as a source.
Share of answers that name your brand, linked or not.
Share of answers that actively suggest you. The one closest to a sale.
Whether your content shapes the answer or just sits in the footnotes.
How the engines describe you, including anything outdated or wrong.
What you get
Deliverables, not adjectives.
The question set is yours to keep and re-run, with us or without us. It is the benchmark everything afterwards gets measured against, so handing it over rather than holding it is the point.
Request a sample auditWho this works for
Most people arrive having asked ChatGPT about their own category once and not liked the answer.
An audit is evidence. "We are named in two of thirty questions and our competitor is named in twenty-one" moves a budget conversation in a way an opinion does not.
Outdated positioning, a merged company, a product you discontinued. Worth knowing before a buyer reads it.
Fair enough, and this is the cheapest way to do it. You keep the question set whichever way you go.
Questions, answered straight
No fluff, as promised.
Can you actually influence what an AI says?
You cannot edit the model, but you can change what it reads. Engines synthesise from sources they can crawl and trust, so coverage, corroboration and clean entity signals reliably shift outcomes over time.
How often should we re-run the audit?
Quarterly. Models update, competitors publish, and the answer set moves. A single audit is a snapshot; the value is in the trend, which is what citation tracking is for.
Is this different from SEO?
It overlaps heavily and is not the same. Strong SEO is the foundation, but engines weigh corroboration across sources and clarity of entity far more than a blue-link ranking does.
What is AI share of voice?
The proportion of AI answers, across a fixed question set and several engines, in which your brand is cited, mentioned or recommended, compared with your competitors. The audit reports it alongside the reasons behind it.
How many questions do you test?
A defined set as standard, built from your buying behaviour rather than from a keyword list. The count matters less than the construction: thirty questions a buyer would actually ask is more useful than three hundred that flatter you.
Which engines do you test?
ChatGPT, Gemini, Perplexity and Google AI Overviews as standard, because they behave and cite differently and testing one tells you very little about the others. Additional engines can be added if your buyers use them.
Do you test in Vietnamese as well as English?
Yes, and it is a genuinely separate exercise rather than a translated question set. Buyers phrase things differently, the engines draw on different sources, and the competitive picture is usually completely different.
What if the audit says we are doing fine?
Then you have a documented baseline, evidence for the people who were worried, and the question set to re-run in six months. We have delivered that result and we would rather report it than invent a problem.
Do we have to continue with you afterwards?
No. You keep the question set, the baseline and the prioritised plan, and all of it is written so another team could act on it. An audit that only works if you hire the auditor is not an audit.
Our working process
How we work
Five steps, in the same order, on every engagement.
Audit
Site, search visibility, AI citations and competitors. The findings are yours either way.
Strategy
A prioritised plan tied to outcomes: what happens in month one, and what it achieves by month six.
Build or optimise
The team that wrote the recommendation implements it. No handover gap.
Measure
One dashboard. Weekly written updates, monthly reports.
Iterate
Quarterly reviews. What worked gets more budget. What didn't gets cut.
How engagements are shaped
Projects are quoted after the audit. If your budget can't move the needle, we'll say so.
Proof from real work
Related reading
Want to know what the engines say about you?
Tell us what you're working on. You'll get an honest answer about whether we can help, and exactly how.