What file types can I upload?

Common document formats are supported. The practical requirement is that the file contains real text — a scanned image of a page has nothing to index.

Do uploads update automatically?

No. A crawled page re-indexes on schedule; an uploaded file is a point-in-time copy. Replace it when the source changes, or it will keep citing the old version.

Can I mix uploads and crawled pages?

Yes, in the same bot. Just be deliberate about it: mixing internal documents into a public-facing bot is the mistake worth avoiding.

Getting started

Add documents the crawler cannot reach

Not everything worth answering from is on a public web page. Uploads cover the rest.

Start freeNo credit card. 2 minute setup.

The most useful material is often the least public

Handbooks, specification sheets, process documents, and internal policies hold the answers people most often need. None of it is crawlable, because none of it is published.

Key information trapped in documents rather than pages
Internal processes that exist only as a shared file
Specifications that live in PDFs no search can read
Gated content that customers still ask about

Upload the document as a source

In the Knowledge tab, add the file as a source alongside your crawled site. It is indexed the same way, and answers drawn from it cite the document rather than a URL.

The Knowledge tab with the Upload file source type selected, showing a file drop area alongside the other source types. — Uploads sit beside crawled pages in the same knowledge base, and answers drawn from them cite the document rather than a URL.

Think about where the bot is deployed

This is the step that matters. Material that should not be public must only be answerable on a surface where the audience is right — an internal tool, domain-locked to it — not on your public marketing site.

Separate bots for separate audiences

Never mix internal documents into a public bot.

Domain-lock accordingly

The lock is what enforces where an internal bot can run.

Keep uploads current deliberately

Crawled pages re-index automatically; an uploaded file is a snapshot from when you uploaded it. When the document changes, replace it — otherwise the assistant will confidently cite a superseded version.

Test with questions only the document answers

Ask something that appears nowhere on your public site. If it answers and cites the document, the upload is indexed correctly.

Common questions

What file types can I upload?: Common document formats are supported. The practical requirement is that the file contains real text — a scanned image of a page has nothing to index.
Do uploads update automatically?: No. A crawled page re-indexes on schedule; an uploaded file is a point-in-time copy. Replace it when the source changes, or it will keep citing the old version.
Can I mix uploads and crawled pages?: Yes, in the same bot. Just be deliberate about it: mixing internal documents into a public-facing bot is the mistake worth avoiding.