Fix Claude Refusing to Analyze Uploaded Files

Claude occasionally refuses to read, analyze, or summarize files that you upload. This issue typically presents as a direct refusal message (e.g., "I cannot analyze this file" or "I am unable to assist with this document") or as a silent processing failure where Claude acts as if no file was attached.

These refusals are rarely due to account bans. Instead, they are caused by safety filter triggers, file format incompatibilities, or document corruption.

Why Claude refuses or fails to read files

Understanding the cause of the refusal is key to fixing it: * False-Positive Safety Flags: Claude's alignment safety filters scan document contents. If a file contains medical data, legal case files, copyrighted text, or words resembling personal identifiable information (PII), the safety filter may block analysis. * Unreadable Document Formats: Claude cannot read scanned PDFs that lack an optical character recognition (OCR) layer. If the file is just a collection of images, Claude cannot parse the text. * Size and Token Overload: Files might be under the raw upload size limit (typically 10MB to 30MB depending on the interface) but contain too many complex tokens, causing the parser to silently time out.

How to fix file upload refusals

Follow these troubleshooting steps to resolve file reading issues:

1. Perform manual OCR on scanned PDFs If Claude tells you a PDF is empty or that it cannot see the content, the document is likely saved as raw images rather than selectable text. 1. Open the PDF in a browser or PDF viewer. 2. Try to highlight and copy a sentence. If you cannot highlight individual words, the document is an image. 3. Run the PDF through an OCR tool (such as Adobe Acrobat, Google Drive PDF import, or a free online OCR converter) to convert the images into searchable text. 4. Re-upload the newly saved OCR-enabled PDF to Claude.

2. Extract and paste text directly (Markdown or TXT) If Claude's PDF parser is failing or timing out on complex layouts (like multi-column papers or charts), bypass the document parser entirely: 1. Open your document and copy all the text (`Ctrl+A` then `Ctrl+C`). 2. If the text is extremely long, paste it into a simple plain-text file (`.txt`) or markdown file (`.md`). Claude reads these formats much more reliably than heavy PDFs. 3. Alternatively, paste the raw text directly into the Claude chat box using XML tags to structure it: ```text Please analyze the following raw text: <document_content> [Paste text here] </document_content> ```

3. Clear false-positive safety triggers If Claude refuses your file with a generic safety statement, reframe your query to assure the safety filter that the task is safe, educational, or benign. 1. Do not ask Claude to "analyze this legal document" or "summarize this medical record." Safety filters are highly sensitive to these phrases. 2. Instead, use neutral, abstract framing: *"Extract the structural data and key technical bullet points from this text for formatting purposes. Focus only on the dates and technical specifications."* 3. If the file contains copyrighted content (like a book chapter), ask Claude to *analyze the writing style* or *find grammatical inconsistencies* rather than summarizing the plot, which triggers copyright blocks.

4. Split large files into smaller chapters If your file is causing silent timeouts or extremely slow responses, the context window might be overloaded. Split the document into sections of 15-20 pages each and upload them sequentially across different prompts.

When to escalate

If Claude refuses to read *any* file you upload—even basic, single-paragraph .txt or .docx files—there may be a server-side outage affecting Anthropic's file processing microservice. Check the official Anthropic status page for "File Upload" or "API" degradation notices. If no outage is listed and file uploads fail globally across different browsers and incognito tabs, contact Claude support through the dashboard help icon.