Import your help center, docs site and documents into a support chatbot

Most support teams already wrote the answers. They sit in a help center, a docs site, a pricing PDF and a few spreadsheets. A support chatbot should start from that content, not ask you to rewrite it as a list of intents or a FAQ. Paperlink AI reads what you have in two ways: it imports the public pages of your website and it accepts document uploads.

How it works in Paperlink AI

Website import

Select Add website on the Documents page and paste an address, for example your help center at https://example.com/help/. Paperlink AI finds the pages, shows them grouped by section, and you pick the sections the bot may answer from, or all pages. The address sets the scope: only pages on the same host and under that folder are imported, so a help center import does not pull in your blog or careers pages.

To find pages, Paperlink AI works in this order:

1. It reads the sitemaps that your robots.txt lists, or tries the standard sitemap addresses.
2. It imports the sitemap pages inside the scope, up to 5,000 page addresses for each website.
3. Only when the sitemaps have no pages in the scope does it follow links, up to the link depth of your plan.

It respects robots.txt, identifies itself as PaperlinkAI-Crawler, and renders a limited number of pages that need JavaScript to show their text each day. Pages behind a login, files that are not HTML and HTML pages over 2 MB are not imported.

Scheduled re-crawls

Help centers change every week. Paperlink AI checks imported websites again on a schedule: weekly on Starter, daily on Growth and up. An update skips pages that did not change, so it uses no processing for them. You can add 1 website on Starter, 3 on Growth and 10 on Scale, and select Update now on a website row when you publish an urgent change. During the trial you can import 1 website, with no automatic updates.

Removed pages are handled with care. A page leaves the bot's knowledge only after it returns "not found" on two updates in a row. If more than half of a site's pages disappear at once, which usually means a broken deploy rather than a real change, Paperlink AI keeps all the pages and marks the website Needs attention.

Document upload

Drag documents onto the Documents page, or select Upload files. Paperlink AI accepts PDF, DOCX, XLSX, HTML, CSV, TXT and Markdown files of up to 25 MB each. Upload a document with the same name and new content and it becomes a new version; upload the same name and the same content and nothing changes.

What to look for in any tool

"Train on your website" is a common claim. The details decide whether answers stay correct six months later:

• Can you limit the scope? A bot that reads your whole domain will quote marketing pages and old announcements. You want to choose the help center or docs section only.
• How does it stay current, and on which plan? Ask how often pages are re-read, whether that is automatic, and what a manual refresh costs. A one-time import goes stale with your first release.
• What happens to deleted pages? A bot that keeps answering from a page you removed is a risk. So is a bot that forgets your whole site because of one failed deploy.
• Does it respect robots.txt and name its crawler? You should be able to allow or block it with a standard rule.
• Which file formats, and what about scans? Get the exact list. Spreadsheets and scanned PDFs are where tools differ most.
• Is the import limited to what you add? Importing is only half of it. The bot should answer from the imported pages and nothing else.

Check your site before you import

Most failed imports trace back to a sitemap or robots.txt problem. Run the free sitemap checker and robots.txt checker first, and see the sitemap mistakes that hide pages. To see what a page looks like as plain text, try website to Markdown; for files, try PDF to Markdown. The full walkthrough is in how to turn your help center into an AI chatbot and how to prepare documents for a chatbot.