Can Claude Read PDFs? Yes — Here’s How It Works and What to Know

Yes — Claude can read PDFs. You can upload a PDF directly to Claude.ai and ask questions about it, summarize it, extract specific information, or have Claude analyze its contents. Here’s exactly how it works, what the limits are, and what Claude does particularly well with PDF documents.

How to upload a PDF: In Claude.ai, click the paperclip icon in the message box, select your PDF, and it uploads instantly. Then ask your question. Claude reads the full document and responds based on its contents.

What Claude Can Do With a PDF

Task Works well? Notes
Summarize the document ✅ Excellent Full document or by section
Answer questions about content ✅ Excellent Finds specific facts, quotes, data points
Compare multiple PDFs ✅ Strong Upload multiple files in one session
Extract tables and data ✅ Strong Works best on text-based tables
Analyze contracts and legal docs ✅ Strong Identifies clauses, flags issues, explains terms
Read scanned / image PDFs ⚠️ Limited Requires text layer — pure image scans may not work
Translate PDF content ✅ Strong Ask Claude to translate after uploading
Fill in or edit the PDF file ❌ No Claude reads PDFs, doesn’t modify them

PDF Size Limits

Claude supports PDFs up to 32MB per file and up to 100 pages. Documents within that range load fully — Claude reads the entire content, not just the first few pages. For longer documents, you may need to split them or work section by section.

The 200,000 token context window means very long text-heavy PDFs are handled well. A 200-page research paper, a full contract stack, or a lengthy financial report typically fits within the context window without truncation. See the Claude Context Window guide for the full breakdown.

Scanned PDFs: The Limitation to Know

Claude reads PDFs by processing the text layer — the actual characters embedded in the file. Most modern PDFs created from Word, Google Docs, or similar tools have a full text layer and work perfectly. Scanned documents — where pages are photographs of physical paper — may have no text layer, just images of text. Claude’s ability to read these depends on whether the PDF includes OCR text alongside the image.

If Claude returns a response suggesting it can’t read the content, the PDF is likely a pure image scan without a text layer. Running the PDF through OCR software first will resolve it.

Best Prompts for PDF Analysis

Summarization: “Summarize this document in 3 paragraphs. Focus on the key findings, recommendations, and any action items.”

Contract review: “Review this contract and flag: (1) any clauses that are unusually favorable to the other party, (2) missing standard protections, (3) ambiguous language that should be clarified.”

Data extraction: “Extract all financial figures from this report and organize them into a table: metric, value, and the time period it covers.”

Multi-document comparison: “I’ve uploaded two versions of this agreement. Identify every difference between them.”

PDF Reading via the API

Developers can send PDFs to Claude via the API using base64-encoded file content. Claude processes the document and responds to your prompt based on its contents — the same way it works in the web interface. This enables automated document processing pipelines: contract analysis at scale, research synthesis, financial document review, and more. See the Claude API tutorial for implementation details.

Frequently Asked Questions

Can Claude read PDFs?

Yes. Upload a PDF directly in Claude.ai by clicking the attachment icon. Claude reads the full document content and can summarize, answer questions, extract data, compare documents, and analyze contracts. The limit is 32MB and 100 pages per file.

Can Claude read scanned PDFs?

Claude reads PDFs by processing the text layer. Scanned PDFs that are pure images without a text layer may not work — Claude needs text to process, not just an image of text. If your scan was run through OCR and has a text layer embedded, it will work. Otherwise, run OCR first.

How many PDFs can I upload to Claude at once?

You can upload multiple PDFs in a single conversation — as long as their combined text content fits within Claude’s 1 million token context window (for Sonnet and Opus) or 200,000 tokens (Haiku). For most document types, that means dozens of typical-length files can be analyzed together.

Does Claude save or store uploaded PDFs?

Claude processes PDFs within the conversation context. Anthropic’s standard data handling applies — on Free and Pro plans, conversations including uploaded files may be used for model improvement unless you opt out. For sensitive documents, review Claude’s privacy policy and consider Enterprise for stronger data handling.

Related: Claude File Size Limit: PDF and Image Upload Limits
Need this set up for your team?
Talk to Will →

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *