Access Knowledge Base Sources in Transcripts

Automate with AI3 min readUpdated 2026-03-14

When your AI Voice Agent answers a question using knowledge base content, you can see exactly which source it referenced. This makes it easy to verify accuracy, identify gaps, and improve your knowledge base over time.

Available on: All AI Voice Agent plans with an active knowledge base.


How Source Attribution Works

Every time the agent references knowledge base content during a call, the system tags that response with the source. After the call, you can view the transcript and see:

  • Which knowledge base entry the agent used
  • The exact content it referenced
  • Whether the answer came from a URL, file, Q&A pair, or plain text source

Viewing Sources in Transcripts

  1. Go to JustCall AI and find the call in your call log.
  2. Open the transcript.
  3. Look for the View Sources button on AI responses.
  4. Click it to see the knowledge base content the agent referenced for that specific answer.

Each response may reference one or multiple sources. The system shows all sources that contributed to the answer, ranked by relevance.


What to Look For

Accurate References

The agent used the right source and gave a correct answer. No action needed.

Incorrect or Outdated References

The agent referenced content that is outdated or no longer accurate. Action: update the knowledge base source — edit the Q&A pair, refresh the URL, or replace the file.

Missing References

The agent could not find relevant content and either said "I don't have that information" or gave a generic answer without source attribution. Action: add new content to the knowledge base covering the topic.

Wrong Source Selected

The agent found a source but picked the wrong one — answering a question about return policy with shipping information, for example. Action: improve the specificity of your knowledge base entries. Break broad content into more focused Q&A pairs.


Using Sources for Knowledge Base Improvement

Review transcripts regularly to build a continuous improvement loop:

  1. Weekly review — check transcripts for the past week, focusing on calls where the agent transferred to a human (these often indicate knowledge gaps).
  2. Gap list — maintain a list of questions the agent could not answer. Add corresponding content to the knowledge base.
  3. Accuracy audit — verify that high-traffic answers are correct. Update sources when products, policies, or pricing change.
  4. Remove noise — delete knowledge base content that the agent references incorrectly or that causes confusion.

Was this helpful?