Quick answer: An AI chatbot trained on your website uses your own pages and documents as its knowledge source and answers visitor questions in natural language. Unlike scripted bots, it doesn't need a decision tree for every question, and it can hand off to a human when needed.
The short definition
A website-trained AI chatbot reads the content you already publish, such as product pages, FAQs, policies and help articles, and uses it to answer visitors' questions in real time. The answers come from your content rather than from a script someone had to write by hand.
How it works, step by step
- Crawling: the platform visits your public pages, respecting robots.txt and site limits.
- Cleaning and chunking: navigation, footers and boilerplate are stripped and the remaining text is split into passages.
- Indexing: each passage is converted into a searchable representation (an embedding) so relevant content can be found quickly.
- Retrieval and answering: when a visitor asks something, the most relevant passages are retrieved and the AI writes an answer based on them.
Scripted bots vs. website-trained AI
Scripted bots follow a decision tree. They are predictable but brittle: any question outside the script hits a dead end. A website-trained AI handles open-ended questions because it searches your content instead of following fixed paths. The trade-off is that you need good content and a sensible fallback, ideally a human handoff.
When it's the right choice
- Your website already answers most customer questions, but visitors can't find the answers.
- Your team answers the same questions repeatedly.
- Visitors arrive outside business hours.
- You want AI to handle routine questions while people handle the rest.