Home›AI Receptionist & Lead Capture›How to Train a Conversation AI Bot in GoHighLevel
AI Receptionist & Lead Capture

How to Train a Conversation AI Bot in GoHighLevel

By William Welch ·October 21, 2026 ·12 min read

Follow along — get 30 days free →

In This Guide
  1. What bot training does
  2. Pick the right source type
  3. Step-by-step: train a Conversation AI bot
  4. Keep sources current
  5. Launch checklist
  6. FAQ
  7. Related guides

This tutorial also has a podcast episode

Listen on Spotify — "Go High Level" podcast

To train a Conversation AI bot in GoHighLevel, go to AI Agents → Knowledge Base, create or open a knowledge base, and add sources such as website pages and FAQs. Then open the bot under AI Agents → Conversation AI → Agents List and check that the knowledge base is listed under Knowledge Base Triggers. Test the knowledge base with Test retrieval, then test the bot in the Test your agent panel.

TL;DR
  • Training means giving the bot approved business information it searches before it answers, instead of relying only on general AI knowledge.
  • A knowledge base can combine source types. In our sandbox it offers Web crawler, FAQ, Rich text, Tables and File upload.
  • For website pages, use the Web crawler: pick a mode (Exact URL, All URLs with the path or All URLs in this domain), enter the URL and click Train this page.
  • For direct answers, open the knowledge base's FAQ tab and click Add FAQ.
  • Test twice: Test retrieval checks that the content can be found; Test your agent checks the full conversation.

What bot training does

Conversation AI bot training gives the bot access to approved business information it can retrieve when it answers customer questions. Instead of relying only on general AI knowledge, the bot searches its connected Knowledge Base for relevant content and uses it to generate a response.

Training content can include website and help-center pages, frequently asked questions, policies and written instructions, product, service, or pricing tables, uploaded documents, Google Drive files, and approved web-based sources.

Interface note

The help article says some accounts show Web Crawler and Custom Bot Responses directly in a Bot Training tab, while the refreshed experience manages sources under AI Agents → Knowledge Base. In our sandbox (captured 2026-10-21) the bot editor has no Bot Training tab. It has Build and Deploy views, and knowledge bases are connected in the Knowledge Base Triggers section.

Pick the right source type

Match the source to how the information is structured. A Knowledge Base can combine multiple source types. HighLevel recommends using the smallest set of clear, relevant sources that covers the questions the bot should answer.

Source (help article)Best used forIn our sandbox
Web CrawlerPublic service pages, help articles, pricing pages, policies, and other website contentWeb crawler
FAQDirect, approved answers to common customer questionsFAQ
Rich TextPolicies, instructions, business details, and notes written directly in HighLevelRich text
File UploadSupported documents containing searchable textFile upload (Files tab)
TablesStructured information such as pricing grids, product catalogs, or service listsTables
Web SearchSupported external web information that is not stored in your uploaded business contentNot shown
Google DriveImporting supported Google Docs, PDF, DOC, and DOCX filesNot shown

"In our sandbox" lists the source buttons and tabs our knowledge base showed on 2026-10-21. Web Search and Google Drive are in the help article, but our sandbox showed no button or tab for them.

Create or open a Knowledge Base→ Add source content→ Connect it under Knowledge Base Triggers→ Test retrieval, then the bot

Step-by-step: train a Conversation AI bot

1
You

Open the bot

  1. Go to AI Agents → Conversation AI.
  2. Open Agents List (the help article calls it Agent List).
  3. Click an existing bot's name, or click Create Agent to make a new one.
  4. The bot opens in the editor's Build view. The help article's Bot Training tab is not there in our sandbox; the knowledge settings are in the Knowledge Base Triggers section on the right.
AI Agents, Conversation AI tab, Agents List sub-tab and the agents table
Conversation AI: Agents List. 1AI Agents · 2Conversation AI · 3Agents List · 4your bots (name, status, type, assigned channels).
Conversation AI bot editor in Build view with Knowledge Base Triggers expanded and the Test your agent panel
Bot editor: Knowledge Base Triggers and Test your agent. 1Build · 2Knowledge Base Triggers · 3a connected knowledge base ("Active for all queries") · 4Add Knowledge Base Trigger · 5Test your agent · 6Send a message.
2
You

Create or open a knowledge base

  1. Go to AI Agents → Knowledge Base.
  2. Click Create knowledge base, or click an existing knowledge base's name to open it.
AI Agents Knowledge Base tab with the Create knowledge base button and the knowledge base list
AI Agents: Knowledge Base. 1Knowledge Base · 2Create knowledge base · 3your knowledge bases.
3
You

Add a source

  1. Open the knowledge base. Its source tabs are All, Web crawler, FAQ, Tables, Rich text and Files.
  2. Pick the source type. The help article says to click Add Source; our sandbox had no Add Source menu. An empty knowledge base shows "No knowledge sources added!!" with buttons for Web crawler, FAQ, Rich text, Tables and File upload, and the FAQ tab has its own Add FAQ button.
  3. Add the content and wait for processing to complete.

The help article recommends File Upload for supported documents containing searchable text, Tables for structured pricing, product, inventory, or service data, and Rich Text for information you maintain directly in HighLevel. It also describes importing supported files through Google Drive and manually re-syncing them after changes; Google Drive was not shown in our sandbox.

Inside a knowledge base: name, source tabs All, Web crawler, FAQ, Tables, Rich text, Files, and the Knowledge sources toggle
Inside a knowledge base: source tabs. 1knowledge base name · 2source tabs (All, Web crawler, FAQ, Tables, Rich text, Files) · 3Knowledge sources / Knowledge gaps.
4
You

Add website pages with the Web Crawler

If the bot already has website sources, review them first. The help article says to expand Uploaded Links and confirm the correct pages were trained, each source has a successful status, the refresh timestamp is current, and the content is still relevant. This helps prevent duplicates, outdated information, and conflicting answers. Our sandbox knowledge base had no crawled pages, so we could not show that list.

  1. Choose Web crawler. The Add website pop-up opens ("Crawl and extract content from a website to train your bot").
  2. In Select mode, pick Exact URL (one specific page, the default), All URLs with the path (pages that share a URL path), or All URLs in this domain (pages across the domain).
  3. In Enter URL, enter the full URL, including https://.
  4. Click Train this page. The help article calls this step Get Data, then has you review the pages HighLevel finds, select only the ones relevant to customer questions, and click Train Bot.
  5. Wait for each selected URL to show Trained.

We did not submit a URL in the sandbox, so the screens after Train this page are described from the help article.

Add website pop-up with Select mode, Enter URL, Cancel and Train this page
Web crawler: the Add website pop-up. 1Select mode · 2Enter URL · 3Train this page.
Select mode list: Exact URL, All URLs with the path, All URLs in this domain
Web crawler: Select mode options. 1Exact URL · All URLs with the path · All URLs in this domain.
Don't train the whole site by default

HighLevel's best practice is not to train every page automatically. Exclude outdated, duplicated, navigation-only, legal boilerplate, and unrelated pages, because they may weaken retrieval quality.

5
You

Add FAQs for direct answers

FAQs give concise, approved answers to common questions that other sources may not cover clearly. To add one:

  1. In the knowledge base, open the FAQ tab.
  2. Click Add FAQ (top right, or in the middle of an empty list). The help article calls this + Add Q & A.
  3. In Your question goes here, enter a clear, customer-style question.
  4. In Your answer goes here, enter a concise, complete answer. Each field shows a 1,000-character counter.
  5. Click Save.
  6. Test several natural variations of the question.

FAQs work well for business hours, pricing rules, refund or cancellation policies, service areas, contact details, and common setup instructions.

FAQ tab in the knowledge base with the Add FAQ buttons
Knowledge base: FAQ tab and Add FAQ. 1FAQ tab · 2Add FAQ · 3Add FAQ (empty list).
FAQs pop-up with question and answer fields and the Save button
The FAQs pop-up. 1Your question goes here · 2Your answer goes here · 3Save.
6
You

Connect the knowledge base to the bot

  1. Return to the Conversation AI bot and expand Knowledge Base Triggers. The help article calls this selecting the knowledge base in Bot Training.
  2. Check that your knowledge base is listed. In our sandbox, Existing knowledge base shows Active for all queries.
  3. The section says, "Configure when each knowledge base should be used," and explains that the agent routes conversations based on these conditions. It also shows an Add Knowledge Base Trigger button; we did not open it.
  4. No knowledge base yet? Use the Create New link under the button.

The bot editor screenshot under step 1 shows the Knowledge Base Triggers section.

7
You

Test retrieval first

Test the Knowledge Base first:

  1. Go to AI Agents → Knowledge Base and open the knowledge base connected to the bot.
  2. Wait for new or updated sources to finish processing.
  3. At the bottom right of the knowledge base page, click Open on the Test retrieval widget ("Ask anything your users might ask"). The help article calls this the Retrieval Tester.
  4. In the Test retrieval panel, enter a realistic customer question.
  5. Review the answer and the retrieved sources. Update unclear, missing, or outdated content, then retest the same question.
Test retrieval widget at the bottom right of the knowledge base page with its Open button
Knowledge base: the Test retrieval widget. 1Test retrieval · 2Open.
8
You

Test the bot in Test your agent

  1. Return to the Conversation AI bot. The Test your agent panel is on the right of the Build view (the help article calls it Test your Bot, in Bot Training).
  2. Type a question in Send a message. The panel displays examples such as "What services do you offer?" and says test messages are never sent to your contacts.
  3. Test a common question, a paraphrased version, an FAQ, and an out-of-scope question.
  4. Use the reset icon before you start a separate test scenario (per the help article).
  5. Confirm the response matches your approved source content.

The bot editor screenshot under step 1 shows the Test your agent panel.

After launch

Once the bot is live, use Response Info beside supported AI messages in Conversations to see which Knowledge Base sources contributed to a response.

This is built into GoHighLevel. Try it free for 30 days →

Keep sources current

The help article recommends keeping sources current to help prevent outdated prices, policies, services, and instructions from appearing in customer conversations. It says to update sources whenever business information changes and test retrieval after major updates before relying on a source in live conversations. Its maintenance guidance lists the following source controls:

According to the help article, Auto Refresh can re-crawl web URL sources on a daily, weekly, or monthly schedule. It recommends a cadence that matches how often the page changes.

Launch checklist

FAQ

Which source type should I use?

Web Crawler for public website content, FAQ for direct answers, Rich Text for information you maintain by hand, File Upload for documents, and Tables for structured data.

Should I train the bot on my entire website?

Only when most pages are relevant to customer conversations. Focused pages usually produce cleaner results than unrelated sections of a large site.

Can I use Google Docs?

Yes. The help article describes two methods: import supported files through the Google Drive integration, or use a publicly viewable Google Doc URL in the direct Web Crawler interface. For the URL method, it says to set sharing to Anyone with the link, select Exact URL, paste the document URL, and click Get Data. Our sandbox showed Train this page instead of Get Data; we did not submit a Google Doc URL.

Why isn't the bot using a source I just added?

Confirm that processing is complete, the source has a successful status, and the correct Knowledge Base is attached to the bot. Then use Test retrieval (the help article's Retrieval Tester) to confirm the information can be found.

Why is the bot giving conflicting or outdated answers?

Look for duplicated, overlapping, or outdated sources. Remove or update the incorrect version, then retest the Knowledge Base and the bot.

Is Knowledge Base testing the same as testing the bot?

No. Test retrieval (the Retrieval Tester in the help article) checks whether source content can be found. Bot testing checks the whole conversation, including instructions, goals, actions, and response behavior.

Source: HighLevel Support, Training Your Conversation AI Bot (article 155000008470; modified Wed, 19 Aug, 2026 at 1:02 PM; fetched 2026-10-21).

Screenshots: our HighLevel Cloud Sandbox (English UI, captured 2026-10-21). Where the live screen uses a different label from the help article, this guide uses the live label.

Ready to try this?

$0 for 30 days — just a ~$1 card-verification hold (no subscription charge). Set up everything in this guide inside your trial.

Start Free 30-Day Trial
Cancel anytime — $0 for the first 30 days
Share
William Welch
GoHighLevel Consultant & Agency Automation Specialist
I help agencies replace 5-10 disconnected tools with one platform. I've built and managed GoHighLevel automations across CRM, email, SMS, WhatsApp, and AI — and I publish everything I learn here. More about me →