How to track brand mentions in AI answers

Typing your name into ChatGPT once a week is not tracking. Answers vary between runs, accounts, regions and model versions.

Why manual checks mislead you

Two runs of the same prompt can return different brands. Chat history and personalisation shift the result further. A single check tells you what one session said, not what the platform generally says.

Build the prompt set first

Write down the prompts you want to win, keep them fixed, and change them only on purpose. Without that, every chart you build afterwards compares against a moving target.

Read mentions in context

The count is only the start. What matters next is how you are described, which brands are named beside you, and which links the model attached to the claim.

Turn tracking into a weekly habit

Fifteen minutes a week: read the new answers, note where a competitor replaced you, pick one page to fix. Slow, compounding and boring — which is why it works.