Q&A Insights
Every question your chatbot has ever answered, scored the instant it happened — and one click away from the full conversation behind it.
Q&A Analysis groups conversations into themes over a date range you choose, run monthly or quarterly. Q&A Insights is the live table underneath that report: every individual question and answer, in the order it happened, each already carrying the same 1–100 quality score computed at the moment the conversation occurred. There's no job to generate — the list is simply always current, which makes it the fastest way to spot-check a single answer, a single visitor, or a single low-scoring response without waiting for the next monthly run.
What It Does
- Lists every question asked and every answer given by your ai12z chatbot for this project, newest first (sortable by date)
- Shows the Quality Score computed for that specific answer at the moment it was given — the same per-conversation score Q&A Analysis later rolls up into its theme clusters and topic scores
- Tracks Feedback separately from Quality Score — whether a real visitor gave explicit positive/negative feedback on that answer, distinct from the automatic score
- Lets you search all data by question text, and filter by score threshold, feedback status, and Answer Engine Referral — narrow the whole table down to only conversations that started from an AI answer engine, instead of opening rows one at a time
- Provides a Conversation icon on every row that opens the complete, multi-turn transcript for that visit
- Supports Select All, Delete Selected, and Export All for bulk cleanup or exporting the raw Q&A log
- When a conversation originated from an AI answer engine referral, the conversation view surfaces the exact referral metadata — which engine, which landing page, and full device/visit details — right alongside the transcript
- Provides an Analyze button on every answer, opening a review dialog that turns a vague "this answer wasn't great" into a specific, actionable diagnosis
Understanding the Report Table
| Column | Description |
|---|---|
| Question | The visitor's question, verbatim |
| Quality Score | 1–100 score for that specific answer, computed at the moment it was given |
| Feedback | Whether a visitor left explicit feedback (shown as "Answered (No feedback)" when none was given) |
| Date | Timestamp of the exchange, sortable |
| Conversation | Opens the full transcript for that visit |
| Delete | Remove this Q&A record |
A real sample from this table shows the score doing real work, not just decorating the row: a clean, well-scoped question like "how do i find bot embed code" scored a perfect 100, while a garbled, cut-off question like "i want chatbot where based on users msg i want to open dif..." scored 75 — the same signal Q&A Analysis would later fold into a theme's aggregate score, visible here at the individual-conversation level instead.
Viewing a Full Conversation
Clicking the Conversation icon on any row opens the complete transcript — every turn between the visitor and your copilot, each answer showing its Sources count (e.g. "Sources (3)") so you can see exactly what content the bot drew from, with a footer note confirming "this is the full conversation between the user and the copilot."
AI Referral Source Metadata
This is the detail that makes Q&A Insights more than a transcript viewer: when the platform detects that a conversation started from a visitor who arrived via a generative AI answer engine, the conversation view displays a dedicated AI Referral Source panel alongside the transcript, showing:
| Field | Example |
|---|---|
| Engine | ChatGPT |
| Engine Type | AI Answer Engine |
| Referrer Domain | chatgpt.com |
Directly beneath it, a Landing Page section shows which page the visitor arrived on and its title (e.g. /pricing — "Pricing"), and a Device & Visit Details section shows the full context of the visit: device type, browser, operating system, language, IP address, country, and the exact visit timestamp.
This is the same engine-detection logic documented in AI Referral Analytics — a referrer domain like chatgpt.com normalized to the canonical engine chatgpt. Q&A Insights is where that detection becomes concrete: instead of an aggregate "43.5% of sessions came from ChatGPT" statistic, you get the actual conversation a specific ChatGPT-referred visitor had, word for word.
You don't have to open conversations one at a time to find these: use the Answer Engine Referral filter to narrow the entire table down to only conversations that started from an AI answer engine — useful for auditing every ChatGPT- or Perplexity-referred conversation in a period at once, rather than spot-checking individual rows.
Analyzing an Answer
Beyond reading the transcript, every answer carries an Analyze button that opens a detailed review dialog — the mechanism for turning "this answer wasn't great" into a specific, actionable diagnosis instead of a subjective impression. This requires logging to be enabled in Agent Settings; without it, there's no captured context for the dialog to analyze.
- Describe the concern — explain what's wrong with the answer (wrong facts, wrong tone, missed the point) and get back concrete improvement suggestions
- Context review — the system inspects the question, the answer, which tools or controls were invoked, and the prompts behind the response, and returns specific optimization recommendations rather than a generic critique
Pair this with a low Quality Score row: instead of just knowing an answer scored poorly, Analyze tells you specifically what to fix — in the prompt, the knowledge base, or the tool configuration behind it.
Perfect For
- Spot-checking a specific low-scoring answer without waiting for the next Q&A Analysis run
- Reading the real conversation behind a citation or a referral session, turn by turn
- Confirming which AI engine referred a visitor and what they asked once they arrived — the individual-conversation counterpart to AI Referral Analytics
- Filtering the whole log down to only Answer Engine-referred conversations for a focused audit
- Getting a specific, actionable diagnosis on a weak answer via the Analyze dialog, instead of just a low score
- Exporting the raw Q&A log for a support or content-quality review
Related Documentation
- Q&A Analysis — The monthly, theme-clustered report built from these same individual conversations
- AI Referral Analytics — The aggregate dashboard behind the AI Referral Source metadata shown here
- Web Content Quality — The equivalent live, per-page view, for ingested content instead of conversations
- GEO Suite Overview — How Q&A Insights fits alongside the rest of the GEO Suite