You can't improve what you don't sample. A simple, repeatable method for tracking whether AI answers mention you.
Start with a prompt panel: 20 to 40 prompts that mirror how buyers actually ask. Mix category prompts ('best X for Y'), comparison prompts ('A vs B'), and problem prompts ('how do I solve Z').
Run the same panel across each engine on a fixed cadence — monthly is enough for most teams — and record three things per run: whether you were mentioned, in what position, and whether a link or citation was attached.
From that you can compute citation share (your mentions divided by total brand mentions), presence rate (share of prompts where you appear at all), and competitive displacement (who appears when you don't).
Keep the panel stable. Changing prompts between runs makes movement unreadable. Add new prompts as a separate cohort instead.
Finally, tie it to pipeline qualitatively: ask new inbound leads whether an AI assistant was part of their research. The answer rate is climbing fast.
Free audit, 48-hour turnaround. No commitment.