Ask ChatGPT, Perplexity, and Gemini what your company does, what you cost, and who you compete with. Record the answers verbatim, score them for accuracy, and fix the sources they cite. That twenty-minute audit is now the most consequential brand review you can run, because for a growing share of buyers, the assistant's version of you is the first version they meet.
Run the audit
Open the three major assistants and ask each one the questions a buyer would ask. What does [your company] do? What do they cost? Who are their competitors? Who should I choose for [your service] in [your city]? Is [your company] any good? What are the alternatives to [your company]? Record every answer verbatim, note which sources get cited, and mark each claim right or wrong. No tooling, no vendor, twenty minutes. We've expanded this into a full method in The Brand Audit, chapter five of Be the Answer.
Score each answer on four things: are you named at all, is the description accurate, is it current, and does the sourcing trace to pages you control or to third parties talking about you. Those four columns are your baseline, and everything you do next is measured against them.
What you'll probably find
Most businesses find at least one material error and several stale claims. A service line you retired two years ago, still leading the description. A price from two websites ago, quoted with confidence. A competitor list that includes companies that no longer exist and misses the two that are taking your deals. Occasionally something worse: the assistant confuses you with a similarly named business in another industry, and a buyer just heard about their reviews under your name.
None of this is malicious and none of it is random. Assistants repeat what the quotable web says about you, weighted toward sources that are structured, consistent, and corroborated. For most brands, the quotable web is years out of date, because nobody has been maintaining it. The audit doesn't create the problem. It just shows you a problem your buyers have been seeing for months.
The engines repeat what the quotable web says about you. Fix the sources, not the symptom.
Fixing the record
You can't email an answer engine a correction. What you can do is make the correct version of your story the easiest one to quote. Start with the about page: it should state plainly what you do, who you serve, where you operate, and since when, in sentences a machine can lift whole. Treat it as the canonical record of your claims, not a mood piece.
Then the service pages. Each one should answer, in its first paragraph, what the service is, who it's for, and roughly what it costs. Publishing price ranges feels exposed and is worth it: engines answer pricing questions constantly, and if your number isn't on the record, a competitor's is. Finally, sweep the third-party record: every directory, profile, and listing that describes you should carry the same name, the same description, and the same service list. Inconsistency reads as unreliability to a machine deciding what to repeat. The FAQ pages answer engines cite are the fastest surface to fix first.
What not to do
Skip anything sold as an AI visibility hack: prompt-injection tricks in your page copy, keyword-stuffed pages written for machines, networks of thin sites linking to yours. The engines are built by people who spend their days defeating exactly this, and the durable pattern in search has held for twenty years: tricks decay, records compound. The unglamorous version wins because it's also just a better website.
Make it a habit
Re-run the audit monthly with the same questions and watch the answers drift toward your version of the story. The drift is slow, weeks rather than days, because you're waiting for sources to be re-read and re-weighed. That lag is precisely why this rewards the early mover: a competitor who starts six months after you is six months of re-crawls behind, on every question that matters. Our AEO Playbook turns this audit into a standing monthly practice.
Your brand used to be what people said about you when you weren't in the room. Increasingly, it's what the machines say about you when you're not in the chat. The good news is that unlike the room, the chat shows its sources, and the sources can be fixed.
