What does an AI visibility audit include?

What are the components of a real AI visibility audit?

A real audit has five parts. First, the question set: the questions patients actually ask, geo-scoped to the practice's city and specialty, not generic "who is the best" prompts. Second, multi-engine coverage, because ChatGPT, Claude, Gemini, and Perplexity answer differently and a practice can win on one and be absent on another.

Third, citation and mention tracking per answer: for every answer, whether the practice's name appears, whether its website is cited, and which other sources the engine quoted. Fourth, competitor mapping — who wins the answers instead, named practice by named practice. Fifth, the gap diagnosis: why the engines recommend those competitors and not this practice.

What does a fake or lightweight audit look like?

A lightweight audit runs one query on one engine, screenshots the result, and calls it a finding. That is screenshot theater. A single ChatGPT answer is a snapshot from one run of one engine, and AI answers vary between runs, so a lone screenshot proves nothing about whether a practice is reliably recommended.

The tell is missing methodology. If an audit does not state which questions were asked, how many engines were covered, and how citations were counted, its conclusions cannot be reproduced or trusted. A real audit can be re-run by anyone and reach the same counts.

What should the audit deliverable contain?

The deliverable should contain counts, the actual answers, the sources cited, and a repeatable methodology so movement is attributable. Counts give the score: how many of the answers named or cited the practice. The actual answers let the practice read what engines tell patients, in full, not in summary.

The sources cited explain why competitors appear and the practice does not, which is where the work begins. Done well, a real audit names the competitors winning each answer instead of the practice, so the gap is explained rather than merely flagged. The repeatable methodology is what makes a second audit comparable to the first — so when a number moves, the practice knows the work caused it rather than random variation between runs.

The audit also sizes the opening it finds. In Tenva's gap check, 12 of 16 buyer queries about AI visibility for practices have no authoritative answer source, so the audit shows not only where a practice is absent but how many answer slots remain unclaimed.

Does Tenva run this audit on itself?

Yes. Tenva runs exactly this method on itself. The audit method on this page is the one Tenva publishes its own baseline with. Before optimizing its own site, Tenva was cited in 0 of 40 AI answers to the questions its buyers ask — measured with the same question set, the same four engines, and the same citation tracking described above.

Publishing its own number is the proof that the method is honest. A vendor that will not measure itself the way it measures clients is selling screenshots, not measurement.