Skip to main content
Connect your SiteGPT chatbot to Zendesk to provide instant AI-powered support and reduce agent workload. Your bot answers customer questions through the Zendesk messaging interface, deflecting tickets and improving response times.

​
How it works

Your chatbot integrates with Zendesk to create an intelligent support workflow:

AI ticket deflection

Answers common questions instantly, reducing ticket volume

Seamless escalation

Automatically hands off complex issues to human agents

Conversation sync

Syncs all interactions across Zendesk and SiteGPT

Performance analytics

Tracks deflection rates and conversation metrics

​
Prerequisites

Zendesk Suite account with Conversations API access
Admin privileges to create API keys and webhooks
A trained SiteGPT chatbot ready to deploy
This integration requires a Zendesk Suite plan. The Conversations API is not available on basic Zendesk Support plans.

​
Setup

1

Generate a Conversations API key in Zendesk

In Zendesk Admin Center, navigate to Apps and integrations → APIs → Conversations API.Click Create API key, enter a name (e.g., “SiteGPT Bot”), and click Next.Zendesk displays the App ID, Key ID, and Secret key. Copy both the Key ID and Secret key.
The Secret key is shown only once. Store it securely in a password manager.
2

Connect Zendesk to SiteGPT

In SiteGPT, go to Integrations → Zendesk and click Get Started.Paste the Key ID and Secret key from Zendesk, then click Connect.SiteGPT validates the credentials and provides a Callback URL. Copy this URL for the next step.
3

Create a webhook integration in Zendesk

In Zendesk Admin Center, go to Apps and integrations → Integrations → Conversations integrations.Click Create integration and configure:
  • Name: SiteGPT Bot
  • Webhook endpoint URL: Paste the Callback URL from SiteGPT
  • Request method: POST
  • Format: JSON
Enable Include full user and Include full source for complete context.
These options provide your bot with user details and device information for better responses.
4

Configure webhook subscriptions

Under Webhook subscriptions, select these events:
  • Required events
  • Conversation created - New conversations start
  • Conversation message - New messages arrive
Click Save. Zendesk displays a Webhook ID and Shared secret—copy these for your records.
5

Add a messaging channel

In Admin Center, go to Channels → Messaging and click Add channel → Web Widget.Configure the channel:
  • Name: SiteGPT Bot
  • Integration: Select the integration you created
  • Authentication: Configure as needed
  • Pre-chat form: Optional (e.g., ask for visitor name)
After saving, the channel shows status Active.Copy the widget script and embed it on your website, or test using Zendesk’s Preview function.
6

Test the integration

Open a page with the Web Widget and send a test message (e.g., “What is SiteGPT?”).Zendesk forwards the message to SiteGPT via webhook, and your bot responds through the messaging interface.
If no response appears, verify the Callback URL is correct and “Conversation message” is enabled in webhook subscriptions.

​
Webhook events

Configure which Zendesk events trigger your bot in the webhook subscription settings: Essential for bot operation:
  • conversation.created - Start new conversations
  • conversation.message - Receive new messages
Recommended events:
  • client.updated - Adapt to device changes
  • user.updated - Use latest user information
  • conversation.read - Track message engagement
  • conversation.deleted - Clean up ended conversations
  • user.merged - Handle account consolidation
More events provide better context but increase webhook volume.

​
Customization

In Zendesk:
  • Customize widget color and branding
  • Configure welcome message
  • Set up pre-chat form fields
  • Define operating hours
In SiteGPT: Changes in SiteGPT take effect immediately without reconnection.

​
Advanced features

Measure your bot’s impact on support workload:Key metrics to track:
  • Conversations handled by AI vs. escalated
  • Average resolution time
  • Customer satisfaction scores
  • Most common questions
In SiteGPT dashboard:
  • Navigate to Chat History for conversation details
  • Export data for custom analysis
  • Identify knowledge gaps
In Zendesk:
  • Use Explore to create custom reports
  • Track ticket volume trends
  • Monitor agent workload reduction
Aim for 40-60% deflection rate initially, improving to 70%+ with training.
Serve customers in their preferred language:
  1. Train your SiteGPT bot with multilingual content
  2. Enable language detection in chatbot settings
  3. Bot automatically responds in the detected language
  4. Zendesk tracks conversations by language
Supported languages depend on your training data and SiteGPT plan.
Create sophisticated handoff rules:Escalate based on:
  • Specific keywords (e.g., “refund”, “cancel”, “complaint”)
  • Low confidence scores (<70%)
  • Number of messages without resolution (>5)
  • User sentiment (negative tone detected)
  • Business hours (escalate outside hours)
Configure in SiteGPT’s human support settings.Zendesk routing:
  • Assign escalated conversations to specific agents
  • Route by topic or department
  • Set priority levels automatically
  • Trigger custom workflows
Identify and qualify leads during conversations:
  1. Enable lead collection in SiteGPT
  2. Configure trigger keywords (e.g., “pricing”, “demo”, “enterprise”)
  3. Bot collects contact information when interest is detected
  4. Leads sync to SiteGPT and can be exported to your CRM
Integration with Zendesk Sell:
  • Use Zapier to create leads in Zendesk Sell automatically
  • Sync conversation history to lead records
  • Track lead source and qualification status

​
Managing the integration

  • Monitor performance
  • Update your bot
  • Disconnect
Track your Zendesk bot’s effectiveness:In SiteGPT:
  • Navigate to Chat History for all Zendesk conversations
  • Review response accuracy and user satisfaction
  • Identify frequently asked questions
  • Export conversation data for analysis
In Zendesk:
  • Use Explore for custom analytics dashboards
  • Track ticket deflection rates
  • Monitor first response time improvements
  • Measure agent workload reduction
Review performance weekly and update training data based on insights.

​
Troubleshooting

If your bot isn’t responding in Zendesk:Verify webhook configuration:
  • Ensure the Callback URL in Zendesk matches the URL from SiteGPT
  • Check that “Conversation message” is enabled in webhook subscriptions
  • Confirm the webhook integration is active
Check API credentials:
  • Verify the Key ID and Secret key are correct in SiteGPT
  • Ensure the API key hasn’t been revoked in Zendesk
  • Confirm your Zendesk plan includes Conversations API
Test the webhook:
  • Send a test message through the Web Widget
  • Check Zendesk webhook logs for errors
  • Verify your SiteGPT bot has training data
Use Zendesk’s webhook testing tool to verify the endpoint is reachable.
If responses are slow:
  • Normal delay: 2-5 seconds for AI processing
  • Check training data: Large knowledge bases may take longer
  • Verify webhook delivery: Check Zendesk logs for delays
  • Test bot performance: Try a simple question to isolate the issue
Zendesk’s preview window may be slower than production. Test on your live site for accurate timing.
If you see authentication failures:
  1. Verify the Shared secret matches between Zendesk and SiteGPT
  2. Check that the webhook endpoint URL is correct
  3. Ensure the API key has necessary permissions
  4. Try regenerating the API key and reconnecting
After regenerating API credentials, update them in SiteGPT immediately.
If the Web Widget doesn’t load:
  • Verify the widget script is correctly embedded in your HTML
  • Check browser console for JavaScript errors
  • Ensure the messaging channel is set to “Active” in Zendesk
  • Test in an incognito window to rule out browser extensions
  • Confirm your website allows third-party scripts
The widget script should be placed before the closing </body> tag.
If conversations don’t appear in SiteGPT dashboard:
  • Verify the integration is still connected
  • Check that webhook events are being delivered
  • Ensure “Conversation created” is enabled in subscriptions
  • Review SiteGPT webhook logs for errors
Contact SiteGPT support if syncing issues persist after verification.

​
Best practices

Build trust gradually:
  • Set confidence threshold to 80%+ initially
  • Let AI handle only clear, well-documented questions
  • Monitor escalation rates and adjust threshold
  • Gradually lower threshold as accuracy improves
Recommended thresholds:
  • Week 1-2: 85% (conservative)
  • Week 3-4: 75% (moderate)
  • Month 2+: 65% (aggressive)
This approach minimizes incorrect responses while building confidence.
Maximize AI’s impact on support workload:Train on common tickets:
  • Export top 50 ticket topics from Zendesk
  • Add these topics to your SiteGPT training data
  • Create Q&A pairs for frequent issues
  • Update weekly based on new ticket trends
Strategic escalation:
  • Let AI handle tier 1 support (account questions, how-tos)
  • Escalate tier 2+ issues (technical problems, bugs)
  • Route sensitive topics (billing, refunds) to humans
  • Use keywords to identify escalation needs
Aim to deflect 60-70% of tier 1 tickets within 3 months.
Ensure excellent customer experience:Monitor regularly:
  • Review 10-20 conversations weekly
  • Check for incorrect or unhelpful responses
  • Identify confusing questions
  • Update training data based on findings
Collect feedback:
  • Enable feedback buttons in responses
  • Track satisfaction scores
  • Follow up on negative feedback
  • Use insights to improve accuracy
Set clear expectations:
  • Identify the bot as AI in welcome messages
  • Explain what the bot can help with
  • Make escalation options visible
  • Provide estimated human response times
Quality conversations build customer trust and satisfaction.
Create a seamless AI-human handoff:For agents:
  • Provide access to full conversation history
  • Include AI confidence scores in escalations
  • Show which knowledge base articles were used
  • Enable agents to provide feedback on AI responses
For customers:
  • Maintain context during handoffs
  • Avoid asking customers to repeat information
  • Acknowledge AI limitations transparently
  • Thank customers for patience during escalation
For managers:
  • Track deflection rates by topic
  • Monitor agent workload reduction
  • Identify training opportunities
  • Measure ROI on AI investment
A well-integrated workflow improves efficiency and satisfaction.
Protect customer data and maintain compliance:
  • Store API credentials securely (use environment variables)
  • Enable webhook signature verification
  • Regularly audit API key permissions
  • Rotate credentials every 90 days
  • Monitor webhook logs for suspicious activity
  • Comply with GDPR and data privacy regulations
  • Document data handling procedures
Security builds customer trust and protects your business.

​
Next steps

Human support

Configure escalation workflows

Lead collection

Capture and qualify leads

Chat history

Review Zendesk conversations

Webhooks

Set up custom integrations