An AI visibility audit has one job: tell you whether AI assistants name and cite your brand when buyers ask, how you compare to competitors, and exactly what to change to move that number. Everything below — the prompt runs, the content read, the entity and off-site checks — exists to support those answers. An audit that hands you a vague "improve your authority" has failed at its actual job.
We measure six dimensions. For each, the inputs are what we examine, and the finding is the decision-relevant conclusion we deliver. We publish the framework openly; the proprietary prompt sets, scoring thresholds, and tooling stay with the engagement.
The six dimensions
- 01
Answer share
Inputs: A defined set of buyer prompts — the questions your customers actually ask — run across ChatGPT, Perplexity, Gemini, and Google AI Overviews.
Finding: How often each engine names you, how high in the answer you land, and how that compares to competitors. This is the baseline number everything else is measured against.
- 02
Citation footprint
Inputs: The web mentions of your brand and the specific pages engines cite as the source of a claim.
Finding: Whether you're merely mentioned or actually cited as a source — and which of your pages earn citations versus which get talked over. Mentions get you considered; citations get you named.
- 03
Content extractability
Inputs: The structure of your highest-value pages — how answers, evidence, and claims are arranged on the page.
Finding: Whether your pages are answer-shaped: self-contained, answer-first passages backed by statistics, quotable lines, and citations that an engine can lift cleanly — or buried prose it skips.
- 04
Entity & authority
Inputs: Your sameAs profiles, directory and review presence, third-party mentions, and author/E-E-A-T signals.
Finding: Whether your brand is a clearly identified, corroborated entity an engine trusts enough to repeat — or an ambiguous one it can't confidently attribute an answer to.
- 05
Structured data & machine legibility
Inputs: Schema coverage (Organization, Person, Service, FAQ, Article), llms.txt, semantic HTML, and crawler access in robots.txt.
Finding: Whether engines can parse your identity, your facts, and your offerings unambiguously — the difference between content a model can confidently quote and content it has to guess at.
- 06
Off-site & community presence
Inputs: Comparative listicles and roundups, category directories (Clutch, G2), community sources (Reddit), and your coverage in the Bing and Google indexes the engines draw from.
Finding: Whether you appear in the third-party sources engines actually cite for your category. This is the largest blind spot for most brands — comparative listicles alone are roughly a third of all AI citations, and most of that lives off your own domain.
The output: an AI Visibility Score
The six dimensions roll up into a single AI Visibility Score and, more usefully, a ranked list of gaps. Each gap is written to be acted on, and carries three things:
The gap
What specifically is keeping you out of the answer, in plain terms.
Impact on answer share
How much closing it is expected to move your share of AI answers — so the work is prioritized by payoff, not effort.
The fix & effort
The concrete change and a realistic estimate of the work, so it can be scheduled rather than admired.
The depth scales with the goal: a focused baseline to see where you stand, or an ongoing program that re-scores on a schedule and tracks your share of answer as it moves. The constant is measurement — visibility in AI answers is a number, and a framework that doesn't produce one is just an opinion with a logo on it.
For the step-by-step version of the audit, see the AI visibility audit playbook. For the bigger picture on why this matters, start with the AI visibility & GEO guide.