Users can upload images, and the assistant understands what they show. Images and documents that are mostly text — scanned pages, photos of text, screenshots and handwriting — need optical character recognition (OCR) for the assistant to read that text accurately. OCR is a setting the user turns on before uploading. With it off, uploads are faster, but text inside a scanned document’s images is skipped.
Audience: customer IT teams
Likely cause: a text-heavy image or scanned document was uploaded with OCR turned off. Not a platform defect.
What the user sees
- answers that seem incomplete or miss key sections of a document
- the assistant ignoring charts, diagrams or image-based pages
- nothing found in a scanned PDF, even though the pages clearly contain text
Checks
- In the original file, can the user select the text with their cursor? If not, the text is an image and needs OCR.
- Was the Deep Dive OCR setting turned on in Settings before the document was uploaded?
- Whether this affects one user or several.
Fix
- Turn on the Deep Dive OCR setting in Settings.
- Upload the document again. Turning on OCR doesn’t change documents already uploaded.
- Ask the question again once the upload has finished.
Tip: uploads take longer with OCR on, especially for large or heavily scanned files. Leave it off for text-based documents, and for images where the picture matters more than the words in it.
Escalate when
- the document was uploaded with OCR on and its text is still missing from answers
- several users see the same problem
Ready to escalate? See the Escalation page for what to collect first.