When the AI Agent answers questions, it cites the knowledge base articles it used. This transparency helps customers verify information and explore topics in depth.

What Are Source Citations?

Source citations are links to the knowledge base articles the AI referenced when generating its response. They appear after the AI's answer.

Example:

Customer: How do I reset my password?

AI: To reset your password, click "Forgot Password" on the login page, enter your email address, and follow the link in the password reset email. The link expires after 24 hours.

Sources: Resetting Your Password, Account Security FAQ

Why Citations Matter

Transparency

Customers can see exactly where information comes from. This builds trust because:

They know answers aren't made up
They can verify accuracy themselves
They understand the AI's knowledge source

Deeper Learning

Customers who want more detail can click through to:

Read the full article
Explore related topics
Find information the AI didn't mention

Quality Assurance

Your team can use citations to:

Verify AI is using correct sources
Identify when wrong articles are cited
Improve articles that lead to poor answers

How Citations Work

RAG Process

The AI uses RAG (Retrieval-Augmented Generation):

Customer asks a question
System searches your knowledge base using semantic similarity
Relevant articles found and passed to the AI
AI generates answer based on article content
Citations attached showing which articles were used

Citation Selection

The AI cites articles that:

Directly relate to the question
Contain the information used in the answer
Provide the most relevant context

Not every article searched is cited—only those actually used to formulate the response.

Inline Images and Videos

Beyond text citations, the AI can embed images and videos from your knowledge base articles directly in its responses. This helps customers see exactly what you're describing without clicking away.

How Media Selection Works

When answering a question, the AI:

Identifies relevant articles using semantic search
Reviews all media (images and videos) in those articles
Reads captions to understand what each piece of media shows
Selects only relevant media that directly illustrates the answer
Embeds media inline within the response

Example:

Customer: How do I connect my Slack workspace?

AI: To connect Slack, go to Settings > Integrations and click "Add Slack". You'll be redirected to Slack to authorize the connection.

[Screenshot: Settings page showing the Integrations tab with the Add Slack button highlighted]

Sources: Slack Integration Guide

When Media Is Included

The AI includes images or videos when they:

Show UI elements mentioned in the answer
Demonstrate a process step-by-step
Clarify something that's hard to explain in text
Come from articles being cited

The AI may choose to show no media if none of the available visuals are relevant to the specific question.

Supported Media Types

Type	How It Displays
Images	Displayed inline in the chat
YouTube videos	Embedded player in the chat
Loom videos	Embedded player in the chat

Improving Media Selection

To help the AI choose the right media:

Add descriptive captions to every image and video in your articles
Write captions that explain context - "Settings page showing password reset option" is better than "Screenshot"
Use relevant visuals - Only include images that genuinely help explain the content

See Creating Articles for details on adding captions to your media.

Reading Citations

Multiple Sources

An answer may cite multiple articles when:

Information spans several topics
Different articles cover different aspects
Related context enhances the answer

No Citations

If the AI responds without citations, it might mean:

The question is clarifying (not needing KB content)
The AI is indicating it can't find relevant information
The AI is offering to escalate

Single Source

Many answers come from one article when:

One article comprehensively covers the topic
The question is specific and narrow
Best information is concentrated in one place

Improving Citation Quality

Write Focused Articles

Articles with clear, specific topics get better citations:

Article Type	Citation Quality
Focused single topic	High - AI knows exactly when to cite
Broad overview	Medium - May be cited for various things
Multi-topic jumble	Low - Unclear when to cite

Use Clear Titles

The AI uses titles to understand article relevance:

"How to Reset Your Password" → cited for password questions
"Account Management" → less clear what it covers

Update Content Regularly

Outdated content leads to:

Incorrect citations
Wrong information in answers
Customer confusion

Keep articles current to maintain citation quality.

Avoid Duplicate Content

If multiple articles cover the same topic:

AI may cite the less relevant one
Customers get inconsistent answers
Consolidate into one authoritative article

Verifying Citation Accuracy

Regular Audits

Periodically test your AI:

Ask common customer questions
Check if citations are appropriate
Click through to verify article relevance
Note any issues for correction

Customer Feedback

If customers report incorrect information:

Check which article was cited
Verify if the article content is wrong
Or if the citation was inappropriate
Fix the underlying issue

Monitor AI Conversations

Review AI chat logs for:

Patterns in citation usage
Articles frequently (or never) cited
Mismatches between questions and citations

When Citations Are Absent

The AI might not provide citations when:

It Can't Find Relevant Content

If no articles match the question, the AI may:

Acknowledge it doesn't have information
Offer to escalate to human support
Ask for clarification

It's Asking Clarifying Questions

The AI might ask follow-up questions without citing sources:

"Are you asking about X or Y?"
"Could you tell me more about the issue?"

It's Offering Escalation

When escalating to human support, citations are typically not included since the AI isn't providing an answer.

Citations and Trust

Building Customer Confidence

Citations show customers:

Your AI is grounded in real documentation
They can verify any claim
Your organization values transparency

Reducing AI Skepticism

Some customers distrust AI. Citations help by:

Showing the AI doesn't fabricate information
Providing a way to "fact check" responses
Offering familiar documentation to fall back on