Claude Cannot Read Your PDF? Check the File, Scan, and Upload

When Claude cannot read a PDF, the file is the first place to look — not Claude. The most common causes are a scanned page without a text layer, an encrypted or password-protected document, a file that is too large, or a format that uploads successfully but cannot be parsed. Each cause has a specific fix.

While you work through the problem, keeping a note of what has already been confirmed means you are not starting from zero in each attempt. fenbs keeps that kind of context note alongside the task, so the next session inherits what was already verified rather than re-reading the whole thread.

Scanned PDFs without a text layer

A scanned PDF is a set of images of pages. Unless optical character recognition (OCR) has been run on it, there is no readable text layer — the file looks blank to any tool that reads text. Claude may return an empty response, say the document appears to contain only images, or miss large sections of the content.

The fix is to run the file through an OCR tool before uploading. Adobe Acrobat's built-in OCR, several free online converters, and many document management systems can add a text layer to a scanned PDF. Once the text layer is there, the upload should work.

To test whether the issue is a missing text layer, open the PDF in a viewer and try to select and copy a line of text. If you cannot select anything, the file is image-only and needs OCR.

Encrypted and password-protected files

A password-protected PDF must have its restriction removed before you upload it. Most PDF editors let you save a copy without the password after you authenticate. If the file is encrypted at a level your editor cannot remove, you need the credentials from whoever applied the restriction. Some files carry document-level encryption without a visible password prompt — if you are unsure, check the security properties in your PDF editor.

File size and page count limits

Claude.ai limits the size and number of pages it processes per upload. Very large documents may be accepted but return incomplete output — the first portion is processed and then it stops. Check Anthropic's help pages for current file limits, as these can change.

For documents that exceed the limit, split the file into logical sections and upload the most relevant part. A focused upload also tends to produce a more specific and focused response, which is usually better than processing a 200-page document to answer a question about one section.

Upload succeeds but output is incomplete

An upload can succeed while Claude only processes part of the document. This happens with multi-column layouts, tables embedded as images, sidebars, and footnotes — formats where the reading order is not linear. Ask Claude to tell you which sections it read and which it could not access, then paste the missing content directly into the prompt as quoted text.

Common PDF types that cause problems

Print-to-PDF files from web browsers often contain text that is visually rendered but stored in a way that makes extraction unreliable. Headers, navigation elements, and decorative text can appear in unexpected positions in the reading order, confusing Claude about what belongs to which section.

PDF portfolios — files that bundle multiple PDFs into a single container — are a separate case. Claude cannot navigate the container structure to read each embedded document. Extract the individual files first and upload the one you need.

Corrupted files sometimes pass the upload step but produce nonsense output or cause the generation to stop unexpectedly. If a document consistently produces garbled or truncated output and none of the other fixes apply, try opening the file in a PDF viewer and using "Save as" to produce a fresh copy. This often resolves minor corruption.

Trying Claude Code for PDF work

Claude Code can work with PDFs through libraries in its project environment, which handles some file types that Claude.ai cannot process through the upload interface. If the file is text-based and the problem is with the web upload, see what Claude Code can read for the range of file formats it handles.

When none of the fixes work

If the file is text-based, within the size limit, and not encrypted but Claude still cannot read it, try converting the PDF to a Word document and uploading that, or copy the text content and paste it directly into the prompt. Pasting bypasses the upload entirely and gives Claude the text it needs.

How to do this with fenbs

1. Create a free account and add a task for the document review.
2. Paste the source link or file reference into the task so it is easy to find later.
3. Add a context note with what has already been checked — which sections Claude successfully read and which were empty or missing.
4. When the file is fixed and the review is complete, update the task with the verified summary and mark it done.