AI analytics
Questions & answers
Charts summarize; this is where you read the receipts. The Questions page lists every stored answer, and each question drills into its full per-engine history.
The collected answers table
Questions (under AI analytics in the sidebar) is a paginated table of every answer collected for the current brand, 20 per page, filtered by the top-bar date range and platform:
| Column | Meaning |
|---|---|
| Question | The prompt that was asked — click it to open the full answer history. |
| Platform / Model | Which provider and which exact model produced the answer. |
| Source | Manual (you pressed run) or Scheduled (the scheduler fired). |
| Mentioned | Whether the analysis found your brand (or an alias) in the answer. |
| Rank | Your 1-based position when the answer was a ranked list. |
| Sentiment | Tone of the mention, −1.00 to +1.00. |
| Collected | When the answer was gathered. |
The question detail page
Clicking a question opens its detail page: the prompt text with its topic and status badges, and the answer history — every stored answer as a card with the platform and model, the timestamp, the full answer rendered as markdown, and badges for mention, rank, sentiment, and citations.
- Arriving from a platform-scoped row shows answers from that engine only — an “only” badge appears with a Show all engines action to widen back out.
- The same detail page is reachable from Question Mining (tracked questions) and from the questions list on Brands.
How to actually use this page
- Audit surprises. A weird sentiment score or unexpected rank? Open the answer and read exactly what the model said — the raw text settles every argument.
- Compare engines on one question. The same prompt often gets a glowing answer from one engine and a shrug from another; the history view puts them side by side.
- Watch an answer evolve. With a schedule running, the history shows how an engine’s answer to the same question drifts over weeks — including the moment your brand starts appearing.
Raw count vs. analyzed count
This table counts raw answers; analytics pages count analyzed ones. Empty or failed model replies are stored here but never analyzed, so the analytics totals can be slightly smaller. That’s expected.