Short answer: yes. NotebookLM, which Google renamed Gemini Notebook on 16 July 2026, now reads scanned PDFs and image files. Google's help centre lists images as a supported source type, and when we re-tested in October 2026 a scanned PDF imported with text we could question and cite. Scans aren't the brick wall they were in 2025, but poor scans, old imports and a few edge cases still trip it up.
What Gemini Notebook reads in 2026
Google's "Add or discover new sources" help page lists the source types: PDFs, Word, text, Markdown, CSV and PowerPoint files, Google Docs, Slides and Sheets, web URLs, YouTube videos, audio files, ePub files, and images in formats including JPEG, PNG, TIFF, HEIC and WebP. Each source can hold up to 500,000 words or 200 MB, and a free notebook holds up to 50 sources.
That image support is what changed the scanned-PDF story. A scan is a picture of a page; Gemini Notebook now reads the text in that picture instead of looking only for a text layer.
What still goes wrong with scans
- Poor image quality. Google's own page says "certain types of images may not work as well". Faint photocopies, phone photos at an angle and low-resolution faxes give any model less to read, and the answer quietly inherits the gaps.
- Handwriting. Neat block capitals usually come through; cursive and hurried notes are unreliable. See can AI read handwriting in PDFs?.
- Old imports. A scanned PDF you added before image support improved may still be stored as it was processed then. Remove it and add it again.
- Lost Drive access. A source imported from Google Drive becomes inactive if you lose access to the file or it's deleted. It still counts toward your source limit but isn't referenced in answers.
- Web pages. A source added by URL imports only the page's text: images, embedded videos and nested pages are skipped, and paywalled pages aren't supported.
- Locked files. Password-protected PDFs still fail. Unlock a copy first.
How to check a scanned source in five minutes
- Ask a pinpoint question. Pick a fact that appears only on one scanned page, such as a date, figure or name, and ask for it.
- Compare with the page. Open the PDF and check the answer against the image. If it's wrong or missing, the page didn't read cleanly.
- Fix the weak pages. Rescan at 300 dpi, straighten skewed pages, or run the file through OCR to add a real text layer. Our guide to making a PDF searchable covers the free and paid routes.
- Re-add the source. Upload the improved copy and delete the old one so it doesn't use a slot.
Reading a scanned PDF without uploading it
If your scans live in Google Drive, SharePoint, a wiki or an email, an in-browser reader skips the upload step. Sidenote is a reading assistant for Chrome, Edge and Firefox that reads the PDF you have open, runs OCR on scanned pages on paid plans, and answers with a scroll-to-source citation: click it and the document scrolls to the exact sentence and highlights it. Claims it can't match to a real passage are dropped server-side, so a bad read shows up as "not found" rather than a confident guess.
Gemini Notebook vs an in-browser reader for scanned files
| Gemini Notebook | In-browser reader (Sidenote) | |
|---|---|---|
| Scanned and image-only PDFs | Read, with quality caveats | OCR'd on paid plans, then cited |
| Workflow | Upload sources into a notebook | Read the file where it already is |
| Verify a claim | Citation to the source | Click to the exact highlighted sentence |
| Audio and Video Overviews | Yes | No |
| Best for | Synthesising a set of uploaded sources | Reading and citing whatever you have open |
Gemini Notebook is genuinely good at what it's for: pulling a set of sources into one notebook and working across them, with Audio Overviews, mind maps and reports on top. For scans you just need an answer from, with a citation you can check against the exact line, reading in place is quicker. See the full Sidenote vs Gemini Notebook comparison and our roundup of Gemini Notebook alternatives.