Mention Rate

Mention rate = answers naming the brand ÷ total answers. Total answers = questions × platforms × sampling runs.

Fix the denominator before the first run. Adding or dropping questions mid-flight, changing the platform mix, or sampling unevenly means two months of data are not the same measurement — this fails far more programmes than sampling noise does.

Keep the monitoring question set separate from the topic-research set. The monitoring set stays frozen so the series stays comparable; the research set can grow. Merging them means the denominator never settles.

The same prompt can return different answers by device, region and account, so a single run proves nothing. Sample repeatedly and report the ratio.

Deciding whether a brand was 'mentioned' needs both string matching and semantic review. String matching alone pulls in unrelated things with the same name; semantic review alone misses cases where a table in the answer runs the brand name into surrounding text.

Answers routinely name several brands. Mention rate is not zero-sum — one answer can count as a mention for several brands at once.

Treating it as market share. It measures whether the AI said your name, and does not convert into commercial share.

Comparing a cumulative figure with a monthly one. Stretch the window far enough and almost every brand approaches 100%, which destroys the signal. Report daily-average or monthly figures and label the column.

Reading the total without the per-platform split. Citation-source behaviour differs sharply by platform, so a rising total may be one platform moving.