Which files work, and size limits

When you're done, you know which files to give your Chirp and how big they can be.

Who this is for: anyone adding knowledge files to a Chirp.

Files that work

File	How it's read
PDF (.pdf)	The text, page by page, so answers can mention the page. Scanned PDFs (pictures of pages) have no text and are turned away, and so are PDFs with a password.
Word (.docx)	The text. Pictures are left out.
Excel (.xlsx)	Each sheet in turn. When every cell of the first row holds a heading (text, not a number), each row is read with them, like "Price: 12; Size: large". Otherwise cells are read one after another.
Text (.txt), Markdown (.md)	As written.
CSV (.csv), TSV (.tsv)	As written, row by row. A passage from the middle of the file doesn't carry the column headings, so an Excel file with headings often answers better.
JSON, XML, YAML (.json, .xml, .yaml, .yml)	As written.
Web pages (.html, .htm)	The text, without scripts and styles.

Not accepted: older Word and Excel files (.doc, .xls), PowerPoint, .rtf, and pictures. Save them as PDF, .docx or .xlsx first. A PDF with some scanned pages is accepted, but those pages add no text.

How big

• Per file: the drop zone shows your plan's limit, such as "25 MB each", measured on the file you upload. To start with: Essentials 25 MB, Creator 50 MB, Academy 50 MB, Business 100 MB.
• Per Chirp: up to 20 files.
• In total: your account's knowledge storage: see Knowledge storage and your plan's limit.
• Very long text: a file with more text than a Chirp can use is turned away with "This file has more text than a Chirp can use. Split it into smaller files."

Your plan's current limits are also on the Pricing page.

Tips for files that answer well

• One topic per file, with clear headings, works better than one huge file.
• Use the words your visitors use. A Chirp with lots of text searches its files by words: see How a Chirp uses its files.
• Text beats pictures of text. A screenshot of a price list inside a Word file isn't read.