How to train an AI chatbot on your website, files and FAQs

Your CAAC Go agent answers only from the content you give it, so training it means adding data sources. Open Data sources, add your website, pasted text, question-and-answer pairs or files, then use 👁 View stored content to check what it learned and the Suggestions tab to fill the gaps. The first source you add turns the AI on automatically.

Before you start

You need a CAAC Go account. The Free plan needs no credit card — sign up here.

Have your website address ready, or the documents and FAQs you want your agent to know.

Which source type should you use?

There are four kinds of content, each on its own tab under Data sources. Mix them freely: you can add as many sources as you like, because only the total size is capped.

Step 1 · Train it on your website

Data sources → Website → Train my AI

1. Open Data sources and choose the Website tab.
2. Type your address into "Your website address", for example https://yourbusiness.com.
3. Choose Entire site (recommended) to read your whole site, or Just this page to read one page only.
4. Click Train my AI.

Each page becomes its own source in your list, so you can check, turn off or remove pages one at a time.

What gets crawled first?

CAAC Go reads your sitemap.xml, and any Sitemap: lines in your robots.txt. If there isn't one, it follows links within your own site. Key pages come first — pricing and fees, locations, contact, about, hours, booking, FAQ and menu pages — and blog posts come last. If your site has the same page in several languages, it trains on it once. Sites built with JavaScript are read through a rendering fallback, which is capped at 8 pages.

How many pages one crawl reads depends on your plan, and so does your knowledge room (the total size of everything you add):

See pricing for the full plan comparison.

Import a Zendesk Help Center

Paste your public Help Center address (for example yourcompany.zendesk.com) into the Website tab. CAAC Go imports your published articles and skips drafts. If your help center is private, export the articles and upload them as a file on the Files tab instead.

Step 2 · Add text, Q&A pairs and files

Data sources → Text, Q&A or Files

1. On Text, under "Text to teach your AI", give it a title such as "Refund policy" and paste the content.
2. On Q&A, under "Question & answer pairs", type a question customers ask and your answer. Click + Add another pair to add more.
3. On Files, upload PDF, CSV, TXT or MD files (TSV works too). You can add several at once, up to 20 MB each. CAAC Go extracts the text, and you can edit it afterwards.

Word, Excel and PowerPoint files aren't supported. Save them as a PDF first, then upload the PDF.

To have the AI write Q&A pairs from a source, choose ✨ Auto-FAQ from this in its ⋯ menu. The pairs are saved right away as a new Q&A source that's switched on, so check them with 👁 View stored content and edit or turn them off if needed.

Step 3 · Check what it learned

Every source has a ⋯ menu. Use it to see exactly what your agent reads.

1. Click ⋯ on a source and choose 👁 View stored content.
2. Read through the text. Click Download .txt if you want a copy.
3. To correct a text, Q&A or file source, choose ✏️ Edit content.

Crawled website pages can't be edited this way. Update the page on your site and choose ↻ Re-crawl site, or add a Text source with the correct details.

Then go to Agent and use Chat with your agent, which uses your real knowledge search, to ask the questions customers ask you most. If the AI offers to bring in your team instead of answering, that answer isn't in your content yet. Test messages count toward your monthly AI conversation allowance.

Step 4 · Turn sources off and keep them fresh

• Switch a source's toggle off to leave it out of answers without deleting it. When it's on, it shows "On — included in AI answers".
• Website sources auto-refresh roughly once a week.
• Choose ↻ Re-crawl site any time, for example after you change prices, hours or services. If a re-crawl returns much less content than before, CAAC Go keeps the old content, so one bad crawl doesn't wipe what your agent already knew.
• Your sources are kept only in your workspace, and deleting one removes it completely.

Step 5 · Close gaps with the Suggestions tab

Data sources → Suggestions

The Suggestions tab collects questions your AI handed off to your team and replies that visitors rated 👎. Questions asked more than once and 👎 replies are listed first, under "Teach these first"; questions asked only once are grouped under "One-off questions".

1. Open Data sources and choose Suggestions.
2. Pick a question and click Add to knowledge.
3. Click ✨ Draft with AI to write a first answer, then check and edit it.
4. Click Save & teach.

It's on by default: the workspace owner gets an email about each new question to teach. To change this, use "AI alerts" under Workspace → Notifications. Analytics also shows "Questions the AI couldn't answer" and "Source coverage".

Common questions

How much content can I add?

There's no limit on the number of sources. Only the total size is capped by your plan's knowledge room: 1 MB on Free, 10 MB on Starter, 50 MB on Growth, 200 MB on Scale and unlimited on Enterprise. A single website crawl reads up to 15 pages on Free and 60 on paid plans.

Can I upload Word or Excel files?

Not directly. Save them as a PDF first. The supported file types are PDF, CSV, TSV, TXT and MD, up to 20 MB each.

My content is in one language but customers write in another. Does that work?

Yes. The agent replies in the language the visitor writes in, even if your content is in another language. There's no setting to change.

Will the AI make up answers it doesn't know?

It's built not to. It answers only from your content and is instructed never to invent prices or hours, so test the questions that matter most in Chat with your agent. When a customer asks something about your business that your content doesn't cover, it hands off to your team and asks for the visitor's name and contact details. See how AI-to-human handoff works.

Can I train it on my Shopify store?

Yes, on Starter and up. Go to Integrations → Connect Shopify to train it on up to 100 products (most recently updated first) and 50 pages. Add your shipping and returns policies with Website or Text. Refresh it with Re-sync catalog when your store changes. See CAAC Go for Shopify.

Try it on your website — no signup, no credit card. Next, set up lead and booking request capture or learn how handoff to your team works.

Last updated 2026-09-28