An AI search audit answers one question with evidence: when your buyers ask ChatGPT, Gemini, Claude, and Perplexity for advice, do you exist? Here is how to run one properly.
Per prompt, per assistant: were you named, in what position, with what framing, stating which facts, citing which sources. Aggregated across repeated runs, that yields your mention rate, your framing quality, and a map of who owns the answers you are losing.
Thirty to fifty prompts spanning three layers: category questions, qualified variants with budget, size, and industry, and the life moments that trigger demand. The moment layer is where audits find the surprises, and where our research finds the most open territory.
Fresh sessions, no account history, repeated runs per prompt, and verbatim logging. Single runs mislead because models are probabilistic; three to five runs per prompt is the floor for patterns you can act on.
Three gap types emerge: absence, where you are never named; framing, where you are named but hedged; and factual, where the model states wrong things. Each has a different fix, corroboration, positioning consistency, and source correction respectively.
Aethon runs this audit for you, free, and walks the results on a 30-minute call across all four assistants live. Book it and bring your hardest prompts.
A structured measurement of whether and how AI assistants name your brand across your buyers' prompts, including position, framing, and cited sources.
A disciplined manual audit of thirty prompts across four assistants takes a few days. Aethon's founder-run audit turns around in one business day.
You can run one yourself for free with the method above, and Aethon's baseline audit is also free with a demo call.