Upload documents
PDFs, Word files, and Markdown sit alongside your recordings so one question can be answered from both. This covers formats, limits, processing, and failures.
- Updated
- 8 October 2026
- Read
- 5 min
- For
- Everyone; viewers cannot upload
From file to searchable.
Up to ten files at once; each is processed on its own.
Choose the files
On Documents choose Upload documents, or use New content, then Upload documents. PDF up to 25 MB, Word (.docx) up to 5 MB, or Markdown up to 500 KB, up to ten files at a time.
Upload documents: the accepted formats and limits, and the project to file them in. File them into a project
Optional but useful: a project keeps documents beside the recordings they relate to, and can be asked as one scope. You can move a document later with Move to project.
Let processing run
Each document is queued, read page by page, structured into sections, and made searchable. Scanned pages are read visually first, which is what uses OCR pages. You can leave the page.
Documents: the Ask bar, filters, and each document with its project, date, and status. Open it when it is ready
The document opens on Overview: a summary, suggested questions, its details, and the outline. Preview shows the original pages, and Q&A asks the document directly.
A processed document opens on Overview, with suggested questions and the outline.
What you might see.
Preparing your document
The processing screen names each stage: queued, reading pages and images, structuring sections, and making every passage searchable. A queued document may wait briefly before it starts.
"Needs attention"
Processing failed and the document page carries the reason. The two common causes have direct fixes below, then choose Reprocess.
Password-protected PDF
Remove the password by exporting or printing an unlocked copy, and upload that.
No readable text
Usually a corrupt export or an image-only file misnamed as text. Re-export from the source. Genuine scanned PDFs are supported and read visually.
Allowances for documents, pages, and OCR pages are set out in plans and usage.
Documents, answered.
PDF, Word (.docx), and Markdown. Scanned PDFs work: their pages are read visually before they can be cited.
Document pages count every page processed in any format. OCR pages count only the pages that had to be read visually because they were scanned. A scanned page consumes one of each, so the smaller allowance is your real ceiling.
Yes. Tick them in Documents and choose Ask selected, or ask a whole project so its documents and recordings answer together. Each document keeps its own citations.
Open it and read the reason on the page. Most failures are a password, an unreadable export, or an allowance that has run out. Fix the cause, then choose Reprocess on the document.
No. Viewers are read-only. Ask a member, admin, or owner to add the file.
The file and everything derived from it go together: extracted text, page anchors, and search entries. The stored-document slot frees at the same time. You confirm by typing DELETE.