Short answer: Gemini reads most PDFs, scanned ones included, because it analyzes the images in a file as well as its text. When it can't, the cause is usually a limit or a setting: more than 10 files in one prompt, a file over 100 MB, a plan cap on chats with files, a document too long for the context window, or a Google Drive file Gemini isn't allowed to fetch. Each one has an error message that tells you which.
Cause 1 - Too many files in one prompt
Gemini accepts up to 10 files in the same prompt, subject to availability, and when you upload a folder every file inside it counts toward that 10. A ZIP can hold up to 10 files and 100 MB, and can't include audio or video.
The fix: attach fewer files per prompt, or merge related PDFs into one document first.
Cause 2 - The file is over 100 MB
Videos can be up to 2 GB, but every other supported file type, PDFs included, is capped at 100 MB. A scanned document at high resolution reaches that quickly.
The fix: compress the PDF, or split it into parts. Scanned pages at 300 dpi are plenty for reading.
Cause 3 - "You've reached your limit for chats with files"
This is a plan limit, not a problem with the file. Google says a Google AI plan gives you higher limits, such as uploading more files and asking Gemini to reference uploaded files in more chats.
The fix: wait for the limit to reset, continue in a chat that already has the file, or upgrade your Google AI plan.
Cause 4 - "Delete data to upload file"
Your Google account storage is full, so there's nowhere to keep the upload.
The fix: free up space across Google Drive, Gmail and Photos, then upload again.
Cause 5 - The document is too long for the context window
Gemini may warn that "your uploads may be too large for the best results", or say the prompt exceeds the context window. A long PDF can still upload but only part of it fits, so answers reflect the start of the document.
The fix: split the document into sections and ask about one at a time, or remove pages you don't need.
Cause 6 - A Google Drive file Gemini can't fetch
Adding a file from Drive needs Keep Activity turned on and Google Workspace connected to Gemini. On a work or school account, your Workspace administrator must enable access to apps in Gemini, or Drive files won't attach at all.
The fix: check those settings, or download the file and upload it from your device. If Gemini says it "failed to analyze your file correctly", Google's own advice is to re-upload it.
What about scanned PDFs?
A clean scan of printed text usually reads fine. Where Gemini still struggles is the same place every model does: faint photocopies, pages photographed at an angle, and handwriting. If the answers are vague or wrong only on certain pages, those pages didn't read cleanly. Rescan them at 300 dpi or run the file through OCR, using our guide to making a PDF searchable, and upload the improved copy.
Reading the PDF in place, with citations
Most of Gemini's limits are about getting the file into the chat. Sidenote skips that step: it's a reading assistant for Chrome, Edge and Firefox that reads the PDF you already have open, including files in Drive, SharePoint or an email, through your own signed-in session and read-only.
- No upload limits to manage. Long documents are indexed in full, and each question retrieves the relevant passages.
- Scanned pages are run through OCR on paid plans.
- Every answer is checkable. Each claim carries a citation; click it and the PDF scrolls to the exact sentence. Claims that can't be matched to a passage are dropped server-side.
Gemini vs an in-browser PDF reader
| Gemini app | In-browser reader (Sidenote) | |
|---|---|---|
| Scanned PDFs | Read as images, with quality caveats | OCR'd on paid plans, then cited |
| File limits | 10 files per prompt, 100 MB each | Reads the open file, no upload step |
| Drive files | Needs Keep Activity and Workspace connected | Read in place through your session |
| Long documents | Limited by the context window | Indexed in full |
| Verify an answer | Check the PDF by hand | Click a citation to the exact passage |
The same pattern holds across assistants: see why ChatGPT, Claude and Perplexity refuse PDFs, each for its own reasons.