ChatDocument

AI chatbot for website visitors that answers only from your documents

A website chatbot that improvises is a liability: it promises a feature you do not have or quotes a price you never set. ChatDocument puts a chat widget on your site that answers only from the documents you choose, such as product guides, terms, FAQs and manuals. Each answer shows its source, and when the documents do not cover a question, the widget says so instead of making something up.

Try it on your PDF
An office administrator feeding a stack of stapled paper into a desktop scanner beside wire trays of folders
p. 6

Add the widget with one script tag

  • Create a space and add the documents the bot may use: PDFs, Word files, Markdown, help pages by URL
  • Copy the script tag from the space settings and paste it into your site template
  • The widget is restricted to that one space, so it cannot read anything else in your account
  • Update a document in the space, or in a synced Google Drive, SharePoint and OneDrive or Dropbox folder, and the next answer uses the new version

Answers with the source attached

Visitors type a question and get a short answer with a citation to the document and page it came from. They can open the source to read the full passage, which settles most follow ups on the spot. Each cited passage is checked in code to exist and contain the quoted text before the answer is shown, so the citation is real, not decorative.

  • Answers about plans, setup steps, compatibility, policies and terms
  • Source shown with every answer
  • Honest reply when the documents do not cover the question, with the closest passages listed

Why a document grounded bot is safer on a public site

A generic chatbot answers from general knowledge whenever your content runs out, and on a public website that is where promises you cannot keep come from. ChatDocument answers only from the space behind the widget. If your refund terms do not mention a case, the bot does not invent a rule. It tells the visitor the documents do not cover it, which is your cue to add that page or to give them your support address.

Use the API when you want your own interface

On the same plans, the REST API lets you upload documents, ask questions and extract fields from your own code. Build the chat into an app, a help center or an internal tool. Read document question answering API for the calls, document chat widget for widget options, and RAG as a service for the full builder view. Support teams using the same space internally can follow AI support answers from documentation.

Plans that include the website chatbot

The widget and API come with Business and Enterprise. Questions from website visitors count toward the plan the same way as questions asked in the app, one question per cited answer.

PlanYearly billingMonthly billingCapacity for the widget
Business 249 USD per month (2,988 USD per year) 499 USD per month 7,000 questions per month, 800 per day, 30,000 new pages, 15 seats, audit log
Enterprise 745 USD per month (8,940 USD per year) 1,490 USD per month 20,000 questions per month, 2,000 per day, 100,000 new pages, 50 seats, SSO (SAML), SLA 99.9

Yearly billing is the default and saves 50%. All limits on chat with PDF pricing.

FAQ

What happens when the daily question limit is reached

The widget stops answering for the rest of that day and resumes the next day. Plan your space and plan size around your expected traffic.

Can I run different bots on different pages

Yes. Create a space per bot, for example one for pricing and terms and one for technical guides, and place each script tag where it belongs.

Are visitor questions used to train AI models

No. Questions and documents are processed by AI model providers bound by contract not to train on your content, and they are encrypted in transit and at rest.

Does the widget work on any website platform

It is a script tag, so it works on any site where you can add HTML to the page template.

Put a chatbot on your site that only says what your documents say

Start with the documents your support team already points customers to.

Try it on your PDF