63.1 What a Content Gap Is

A Content Gap represents a visitor question that the Agent could not answer well enough with its current Knowledge and response configuration.

The Content Gaps page describes these as questions visitors asked that the Agent could not answer well and identifies them as opportunities to add Curated Answers or new Knowledge Sources.

63.2 Opening Content Gaps

From Analytics, select Content gaps.

You can also open the full Content Gap queue from the Content Gaps summary on the main Analytics page.

63.3 What the Content Gaps List Shows

Each displayed Content Gap can contain:

• The visitor question.
• The number of times it has been asked.
• The last-seen date and time.
• The current Gap status.

63.4 Occurrence Count

The asked [count] times value helps prioritize Knowledge improvements.

A question repeated frequently may represent a more important Knowledge problem than a one-time unusual query.

63.5 Last Seen

The Last Seen information helps determine whether a Content Gap is still relevant to current visitors.

A question with repeated recent occurrences generally deserves more immediate investigation than an old isolated Gap.

63.6 Gap Statuses

The deployed Content Gap data model recognizes:

• open
• answered
• ignored

The current Content Gaps page displays each row's status.

Do not assume the presence of a status label means that the current screen exposes every possible status-changing operation. Use the controls actually presented in your deployed interface.

63.7 No Open Gaps

When no Content Gaps are present, the interface can report that visitors have not yet asked unanswerable questions.

The main Analytics page can also indicate that current Knowledge coverage is keeping up with visitor questions.

63.8 Investigating a Gap

Before adding new material, determine why the Agent struggled.

Review:

• The exact visitor wording.
• Related Conversations.
• The Agent's existing Sources.
• The extracted Documents and Chunks.
• Retrieval behavior in Playground Diagnostics.
• The Agent's confidence threshold.
• Existing Curated Answers.

63.9 Improving Knowledge with a New Source

If the information is genuinely missing, add an authoritative Knowledge Source that answers the question.

The Source may be a URL, sitemap/crawl result, pasted text, supported uploaded file, or a supported connected Knowledge provider such as Notion or Google Docs.

63.10 Improving an Existing Source

If the correct information should already be available:

1. Open the relevant Source.
2. Check its processing status.
3. Preview what the AI sees.
4. Inspect the extracted Chunks.
5. Reindex the Source if the underlying content changed or extraction needs another attempt.

63.11 Using a Curated Answer

Use a Curated Answer when a common question should always receive an exact approved response.

This is especially useful for policies, pricing language, business hours, contact instructions, or other responses that should not vary unnecessarily.

63.12 Test the Fix

After improving Knowledge:

1. Open Playground.
2. Ask the same question that produced the Gap.
3. Try several realistic variations of the question.
4. Review retrieved Sources in Diagnostics.
5. Confirm confidence and response accuracy improved.

63.13 Do Not Solve Every Gap by Lowering Confidence

A low-confidence response is often evidence that the Agent lacks sufficiently relevant Knowledge.

Lowering the confidence threshold without improving the Source material can cause the Agent to answer more aggressively without actually becoming more accurate.

63.14 Content-Gap Review Routine

For ongoing Agent quality, periodically:

• Review the highest-occurrence open questions.
• Check recently seen gaps.
• Add or improve authoritative Sources.
• Create Curated Answers where exact copy is appropriate.
• Reindex changed material.
• Re-test in Playground.

This creates a continuous feedback loop between real visitor questions and the Agent's Knowledge quality.