Files and documents
What you can upload, how files are checked, and how large document sets work.
Supported files
Documents, spreadsheets, slides, images and text.
- PDF, Word (.docx), Excel (.xlsx), PowerPoint (.pptx)
- Images: PNG, JPEG, WebP, TIFF (photos, scans, screenshots; text is read with OCR)
- Text, Markdown, CSV and web pages (.html)
How files are checked
Every upload is scanned before anyone reads it.
- 1The file is scanned for malware.
- 2Its text is extracted (with OCR for scans) and split into passages.
- 3When it shows Ready it can be used in a question.
Note: Files that fail the scan are rejected and never reach a model.
Large documents
When the council reads passages instead of everything.
Short documents are read in full. Very long ones are searched: the council plans its searches, finds the most relevant passages by keywords and by meaning, and reads those together with a map of the document's headings.
Collections
Ask across hundreds of documents (Pro Max).
A collection groups many documents in a workspace so questions can search across all of them. Collections are part of Pro Max.