Hit Logs show what members ask your community's AI assistant (AI Copilot) and how well it answered. Use Admin → AI Copilot → Hit Logs to check answers and their sources, find questions the assistant struggles with, and add the missing answers to your knowledge base.
Who can do this: Owners and admins. · Where: Web (admin console). · Plan: Requires a platform plan that includes AI Copilot.
What the AI assistant's Hit Logs record
Go to Admin → AI Copilot → Hit Logs. The page is called Ask Hit Logs and has two tabs: Recent Questions and Content Gaps.
Recent Questions lists each question members ask the AI assistant:
| Column | What it shows |
|---|---|
| Query | The member's question. An error code appears under it if the assistant couldn't answer |
| Confidence | High, Medium, or Low |
| Source | Where the answer came from: KB (knowledge base), Community, or Hybrid |
| Mode | Keyword or Semantic search |
| Post-check | The result of the final check on the answer: passed, failed, or rejected_low_confidence |
| Coverage | How much of the answer is backed by its citations, from 0 to 1 |
| L1 / L2 | How many knowledge base results (L1) and community post results (L2) were found |
| Latency and Time | How long the answer took, and when it was asked |
Hit Logs don't show which member asked a question. AI writing tools and AI course outlines aren't logged here.
Filter the AI assistant's recent questions
Use the filters above the table to narrow Recent Questions:
- Source Scope: KB, Community, or Hybrid.
- Retrieval Mode: Keyword or Semantic.
- Post-check: for example, rejected_low_confidence to find answers held back because the evidence was weak.
Click Clear Filters to reset them, and Refresh to load new questions.
Check the details of an AI answer
Click a question to open Question Detail. It shows the Confidence and Coverage scores, then:
- Answer: the full answer the member received.
- Citations: the sources listed under the answer, with a snippet of each.
- Retrieved fragments: everything the AI assistant found while searching, including results it didn't cite. Each fragment has a score, which only ranks fragments against each other; it isn't a percentage.
If a good source was retrieved but not cited, or the wrong post ranks first, adjust your community content settings. See Choose which spaces your AI assistant can search.
Fill content gaps with Q&A entries
The Content Gaps tab groups questions the AI assistant answered with low confidence, so you can see which topics your sources don't cover. Each row shows the Question, how often it was asked (Count), the Avg confidence, the Top source, and when it was Last seen.
To add an answer:
- Click Add to KB next to the question. Mateflow creates a Q&A knowledge source for you if you don't have one yet.
- In Add Q&A to Knowledge Base, review the Question, and write the Answer. Markdown works. To start from a draft, click Draft with AI, then check it carefully.
- Optionally, add Alternative phrasings, one per line.
- Click Add to Knowledge Base.
The entry is indexed right away, and the gap is marked as resolved and leaves the list. To remove a gap you don't want to answer, click Dismiss.
Other ways to improve AI answers
Q&A entries give exact answers to single questions. For wider gaps:
- Upload files or import a website into the knowledge base. See Add knowledge sources for your AI assistant.
- Turn on community content, or add the spaces where answers live. See Choose which spaces your AI assistant can search.
- Publish the answer as a post in a space the AI assistant searches.
Members can't rate AI answers, so Confidence, Post-check, and Content Gaps are the main signals of answer quality. After you add content, check Recent Questions again to see whether confidence improves.
Related articles
Was this guide helpful?