Each day the agent runs your agreed list of questions across the assistants and stores every answer. For each one it notes whether you were named, where you sat in the list and how the answer described you. Because the full text is kept, you can open any result and read exactly what a customer would have seen.
Overnight changes are flagged, so nobody has to scan a dashboard each morning. A drop on one assistant, a new competitor appearing or a shift in tone gets a short note with the answers behind it. Your team decides what deserves action. The agent only reports what it saw.
The question list is the most important input. We draft it with you from what customers really ask, then refine it once the first results are in. Answers can vary between runs, so we read trends over weeks rather than any single result. A short list of well-chosen questions beats a long list nobody reads.
Being named is not the whole story. A mention in the first position with a clear recommendation is different from a passing name at the end of a long list. The agent records position and sentiment separately, so your team can tell strong visibility from weak. It also notes when you are mentioned without a source, which is harder to act on.
Mention tracking is not a fit if no assistant has anything to say about your field yet. In that case the early reports will mostly show absence. That is still useful as a baseline, but we would talk honestly about whether other agents should come first.
ChatGPT, Gemini, Perplexity, Claude and Copilot, with more added as they matter.