Why won't NotebookLM read my scanned PDF?
One sentence: NotebookLM reads text, not images, so a scanned page or handwritten note with no underlying text layer imports blank or fails outright, and the fix is running it through OCR first.
A scanned document is really a photograph of text, not text itself, so NotebookLM has nothing to read. Run it through a free OCR (optical character recognition) tool first, which converts the image into selectable, searchable text, then upload the result. The same applies to handwritten notes, though OCR accuracy there is far less reliable.
How to tell if a PDF is the problem
Open the PDF and try to select a sentence with your cursor. If you can highlight and copy actual words, it has a text layer and will import into NotebookLM fine. If clicking and dragging just selects the whole page like a picture, or nothing highlights at all, it is a scanned image with no text layer, and that is exactly what NotebookLM cannot read.
Fixing it with OCR
- Free browser tools like Adobe Acrobat's online OCR feature or Smallpdf's OCR tool convert a scanned PDF to searchable text in under a minute, no software install required.
- Google Drive does this too: upload the scanned PDF, right-click it, choose "Open with Google Docs," and Drive runs OCR automatically, producing an editable Doc you can then add to NotebookLM directly.
- Desktop tools like Adobe Acrobat Pro have more control over OCR language and accuracy settings, worth it for a large batch of scans or documents in a less common language.
Handwritten notes are a harder case
OCR was built for printed text and struggles with handwriting, especially anything less than very neat. Clean, deliberate handwriting sometimes converts reasonably well; fast, casual notes often come out garbled. If OCR mangles a handwritten source, the honest fix is transcribing it yourself into a plain text file, tedious for a long document, but the only way to guarantee NotebookLM sees what you actually wrote rather than a misread approximation of it.
Checking the OCR worked before you rely on it
After running OCR, spot-check a few sentences against the original before uploading. OCR occasionally misreads a similar-looking character, a name, a number, a unit, and those specific errors are exactly the kind of thing that then shows up confidently in a NotebookLM answer with a citation pointing at the OCR mistake rather than the original. Catching it before upload is much easier than catching it after.
Preventing this next time
When you have a choice, request or save the original digital document instead of a scan; a native PDF export from a word processor or a downloaded ebook always has a proper text layer. Save scanning for the physical documents where there genuinely is no digital original. General preparation habits for all source types are in the sources guide.
People also ask
Does NotebookLM have built-in OCR?
It has some image-to-text capability for photos and screenshots added to a notebook, but a scanned multi-page PDF with no text layer is still unreliable without running dedicated OCR on it first.
Is Google Drive's OCR free?
Yes. Uploading a scanned PDF to Drive and opening it with Google Docs runs OCR automatically at no cost, and the result can go straight into NotebookLM.
Why does my OCR'd PDF still have errors when NotebookLM answers questions?
OCR is not perfect, especially on low-quality scans, unusual fonts, or handwriting. Spot-check the OCR output against the original before uploading and relying on it for anything important.
Exporte seu NotebookLM em um clique
Extensão gratuita do Chrome. PDF, Word e Markdown. Renderizado na sua máquina — nada é enviado para a nuvem.