A custom AI chatbot trained on your website, documents and Q&A

CenterChat turns the information your business already has into an agent that answers customers. Import your website, add documents and notes, write exact Q&A for the facts that must be precise, and choose OpenAI or Google Gemini to write the replies. Every answer shows the source it came from.

How do you train an AI chatbot on your own website?

You don't retrain the AI model. You give the chatbot a searchable copy of your content: it imports your website pages and documents, splits them into passages, and when a question arrives it retrieves the passages that match and writes the reply from them. This is retrieval-augmented generation (RAG). Clean, current content matters more than the amount of content.

Updated by the CenterChat team

How CenterChat learns your business

Website import
Enter a home page to import up to 25 pages of the site, or a deeper URL to import that section. It follows your sitemap and respects robots.txt.
Documents and notes
Upload text, Markdown, CSV or JSON files up to 1 MB, or paste policies and service details as text.
Exact Q&A
Write approved answers for facts that must be word-perfect, such as refund windows or prices.
Retrieval that understands meaning
Semantic search finds the right passage even when the customer uses different words from your page.
Sources on every answer
Each reply shows which source it used, and the agent says so when nothing supports an answer.
Model choice with fallback
Pick OpenAI or Google Gemini per agent; if a provider fails, CenterChat falls back automatically.

How to set it up

Most owners finish setup in one sitting. Agencies repeat the same three steps for each client.

  1. Step 1

    Import and add

    Import your website, then add files, text and Q&A for anything the site doesn't cover.

  2. Step 2

    Test in the studio

    Ask real customer questions, check the cited sources, and fix gaps with Q&A.

  3. Step 3

    Publish and improve

    Install with one script tag, then use improvement suggestions to add missing knowledge.

Prepare content the AI can trust

Retrieval can only return what your content says. Remove outdated pages and conflicting policies before you import, give each policy one current source, and use Q&A for numbers and dates that must be exact.

Keep internal-only material out of public agents. Anything in an agent's knowledge can appear in an answer.

  • One current source of truth per policy
  • Clear headings and plain language
  • Dates on time-sensitive information
  • No secrets or internal notes in public knowledge

What "grounded" means in practice

A grounded answer is one you can trace to a passage in your knowledge or to a live tool result. CenterChat instructs the model to answer business questions only from those, shows the source, and offers a next step when it has nothing to go on. Test this deliberately: ask questions your content cannot answer and confirm the agent declines rather than guesses.

Knowledge sources

SourceUse it forLimits
Website importHelp centre, product, pricing and policy pagesUp to 25 pages per import; public pages only
File uploadExported FAQs, product data, notesText, Markdown, CSV or JSON, up to 1 MB each
TextPolicies and service descriptions you paste inPlain text
Q&AFacts that must be exactOne approved answer per question

What it does not do

So you can decide quickly whether CenterChat fits, here is what this part of the product does not cover today.

Questions, answered

More in the full FAQ.

Does CenterChat train the AI model on my data?

No. Your content is indexed so the agent can retrieve relevant passages at answer time; the underlying OpenAI or Gemini model is not retrained on it.

Which files can I upload?

Text, Markdown, CSV and JSON files up to 1 MB each. For PDFs or Word documents, paste the text or save it as Markdown first.

How many pages can it learn from?

Each website import reads up to 25 pages; import further sections by entering deeper URLs. You can add files, text and Q&A on top.

Will it answer in other languages?

Yes. In text chat it replies in the visitor's language, even when your content is in English.

Which AI models does it use?

You choose OpenAI or Google Gemini for each agent. Model usage is included in your plan; you don't need your own API key.

Put an agent on your site this week.

Every feature on every plan. Start yourself, or tell us what you need and we'll set it up with you.