How to Track Whether AI Engines Are Citing Your Brand
A step-by-step approach to monitoring brand visibility across ChatGPT, Perplexity, and Google AI Overviews, plus what to do once you have the data.
Why manual spot-checking isn’t enough
Typing a few prompts into ChatGPT and eyeballing the results feels like monitoring, but AI answers are non-deterministic and personalized — the same prompt can return different sources on different days. Reliable tracking needs repeated, structured prompt runs, not occasional manual checks.
Building a prompt set
Start with the questions your customers actually ask before choosing a product in your category, not just queries containing your brand name. A useful prompt set usually includes:
Category-level questions
“What’s the best tool for X” style prompts, where you want to see if you show up at all among the options a model suggests.
Comparison questions
“X vs Y” prompts against your top two or three competitors, which reveal how a model frames your strengths and weaknesses relative to them.
Direct brand questions
Prompts that ask about your product specifically, to check whether the model has accurate, up-to-date information about it.
Reading the results
Two numbers matter most: how often you’re mentioned at all (visibility), and how often you’re mentioned favorably relative to competitors (share of voice). A tool that surfaces which pages a model is actually citing is more useful than a raw mention count, since it tells you what to fix.
Turning data into action
Once you know which prompts miss you and which pages get cited when you do show up, the fix is usually content work: tightening the pages that already get cited, and creating direct-answer content for the prompts where you’re absent entirely.