Claude File Upload Stuck on Processing

PDF or image upload shows the spinner forever and never becomes attachable. File size, format, network throttling, or a backend processing queue — each has a different fix.

Published: May 23, 2026 Author: AI Productivity Guide Team 🌐 查看中文版本

You drag a PDF into Claude, or paste an image, or click the paperclip and pick a file. The thumbnail appears with a “Processing…” label and a spinner. Five minutes later it is still there. Sending the message does nothing — the file never finishes processing, so it cannot be attached to the prompt. This is usually one of four causes: file too large for the size cap, format not supported, your connection throttling the upload, or a backend processing queue backed up. Each has a clear fix path.

Common causes

Ordered by hit rate, highest first.

1. File exceeds the per-file size limit

Claude has a per-file cap (currently 30 MB for PDFs, 10 MB for images on most plans). Larger files often show the upload progress as “processing” indefinitely rather than rejecting outright.

How to judge: Right-click the file in your OS → Properties / Get Info → check the size. Anything over 30 MB on PDFs or 10 MB on images is suspect.

2. Format not supported or detected wrong

Claude supports specific MIME types (PDF, PNG, JPEG, GIF, WebP, plus several text formats). A .pdf file that is actually an HTML print-to-PDF, or an obscure image format that confuses MIME detection, can hang processing.

How to judge: Try a small known-good PDF (a one-pager you wrote in Word and exported). If that works, the original file’s format is the issue.

3. Network is throttling the upload

Slow or unstable connection (corporate VPN, throttled hotel WiFi) means the upload bytes never finish reaching Claude’s server. The UI shows “processing” but the upload is actually still in transit.

How to judge: DevTools Network → find the upload POST → check status. “Pending” with growing time means it is still uploading; “Cancelled” or “Failed” means it stopped.

4. Backend processing queue backed up

For large PDFs Claude runs OCR / parsing before the file is usable. If that queue is slow, processing takes minutes even after the bytes are uploaded.

How to judge: Wait 5 minutes for small files, 10 for big ones. Check status.anthropic.com for processing-queue incidents.

5. Page count exceeds limit even though size is fine

A small but very long PDF (1,000+ pages, low resolution scans) can stall processing because page count caps are separate from byte caps.

How to judge: Open the PDF locally and check total page count. Try splitting to 100 pages and re-uploading.

6. Browser extension is intercepting the upload

Some privacy extensions or download managers hook into POST requests and can mangle multipart form data. Upload appears to succeed but the file is empty server-side.

How to judge: Retry in incognito mode with extensions disabled. If it works there, an extension is the cause.

Before you start

Check the file size and page count locally before uploading.
Confirm the file opens correctly in your local viewer — corrupt files will not upload.
Have a known-good small file handy for control tests.
If on corporate network, note any VPN or proxy in play.

Information to collect

Exact file size (in MB), format, and page count if PDF.
Browser + OS + network type (home / office VPN / hotel WiFi).
DevTools Network status for the upload request (pending / failed / 200).
Whether other files of similar size and format upload fine in the same session.
Timestamp of attempt (to check status page).

Step-by-step fix

Step 1: Try a known-good small file

Upload a one-page PDF or a small PNG you trust. If that works in seconds, your account, browser, and network are fine — the issue is the specific file. If even the small file hangs, the issue is broader (network / browser / Anthropic side).

Step 2: Check file size and split if too big

For PDFs over 30 MB, split:

# macOS / Linux with pdftk
pdftk input.pdf cat 1-50 output part1.pdf
pdftk input.pdf cat 51-100 output part2.pdf

# Or use Preview.app on macOS:
# File → Print → PDF dropdown → save as new PDF with page range

Upload each part separately. Claude can reason across multiple attachments.

Step 3: Re-export to a clean format

For PDFs created by weird tools (browser print-to-PDF, certain scanners), re-export:

Open in Preview / Adobe Reader → File → Export As PDF

For images, open in any viewer and “Save As” PNG or JPEG. This strips metadata and forces a clean MIME type.

Step 4: Check DevTools for the actual upload status

F12 → Network tab → upload again. Watch the POST request:

Status	Meaning	Fix
200	Upload reached server	Processing queue is the bottleneck — wait
4xx	Server rejected (size, type, auth)	Read the response body for the reason
Pending forever	Bytes still in transit	Network or VPN is slow — retry on better connection
Failed / Cancelled	Connection dropped	Switch network, disable VPN

Step 5: Try a different browser or incognito

Open an incognito window or a different browser. Log in, retry the upload. If it works, the issue is extensions or cache in your normal browser.

Step 6: Reduce page count for very long PDFs

For PDFs with 500+ pages, even if under 30 MB the processing time and page caps may stall:

1. Identify the 20-50 pages you actually need
2. Use pdftk or Preview to extract just those pages
3. Upload the reduced file

For full-document analysis, split into 100-page chunks and process sequentially across multiple chats.

Step 7: Cancel and resubmit

If processing has hung for 10+ minutes:

Click the X on the thumbnail to cancel
Refresh the page
Re-upload

Sometimes the processing job dies silently and the UI does not notice — refresh and retry clears it.

Verify

Same file uploaded in incognito and a control file uploaded normally both succeed quickly.
DevTools Network shows the upload as 200 and the thumbnail switches from “Processing” to “Ready” within 30 seconds.
Send a test message referencing the file (“summarize this”) and confirm Claude reads it.
For PDFs, ask Claude to quote a specific line from the document — verifies actual parse, not just upload success.

Long-term prevention

Pre-flight file sizes before uploading — most cases of stuck processing are oversized files.
For PDFs you know are big, split proactively in 50-page chunks.
Use known-good export paths (Word → PDF, Preview → PDF) rather than obscure converters.
Keep one small known-good test file bookmarked for diagnostics.
For team accounts that upload regularly, document size caps in your internal wiki.
Re-export scanned PDFs through a clean OCR pass (Adobe Acrobat, ABBYY) before uploading — Claude’s OCR is good but a clean preprocessed version is faster to parse.

Common pitfalls

Waiting 30 minutes hoping the queue clears, instead of checking the size cap.
Uploading a 200 MB scanned PDF and expecting it to work. It will not.
Forgetting that page count is a separate cap from file size.
Blaming Claude when the actual issue is a corporate VPN throttling the POST.
Re-trying the exact same file 10 times instead of trying a known-good file as a control.

FAQ

What is the file size limit? Currently around 30 MB per PDF and 10 MB per image on most plans. Enterprise may be higher — check your plan.
How long should processing take? A few seconds for images, up to a minute for typical PDFs, longer for large multi-hundred-page documents.
Can I upload multiple files at once? Yes, but each counts against limits independently. Five 25 MB PDFs is fine; one 100 MB PDF is not.
Does Claude OCR scanned PDFs? Yes, but quality depends on the scan. Clean text PDFs are always faster.
Where do uploaded files live? They are attached to the conversation. Deleting the chat removes the file from accessibility.

Tags: #Claude #Troubleshooting #uploads