You’ve logged into ChatGPT and clicked the paperclip-upload icon, but to no avail. What should you do?
We’ll cover the troubleshooting steps (to take) when ChatGPT experiences an error reading documents. Generally, such errors are a result of a simple upload misstep. Fortunately, there’s a quick (and easy) solution. Let’s dive in.
Currently, ChatGPT can only accept basic file types: Excel, JSON, PDF, CSV, PNG, JPG, EPUB and so on. If you’ve uploaded a complex or proprietary file type, consider converting it into an image or PDF.
Also, check that you haven’t uploaded too many documents. Currently, ChatGPT’s limit is 20. If you’re only uploading one file, try uploading a different file. Do you get the same error message?
Is your document password-protected? If so, ChatGPT won’t be able to open it. Assuming you have the correct permissions, you can copy and paste the document’s contents into a non-password-protected file.
A corrupted file is another reason the document might be unavailable. Corrupted documents have essential data missing, therefore they’re impossible to open. If you think your file might be corrupted, try opening it. If you see an error message, then follow these instructions to resolve it.
You can also link a document accessible online for ChatGPT to read (e.g. ‘summarise this article by Evolution AI (URL) in 100 words or less’). However, if the link is invalid, ChatGPT won’t be able to read the document. Check the link’s validity by opening it in your web browser before pasting it into ChatGPT.
(Note: ChatGPT and other AI virtual assistants can’t access information from Google Drive, so you won’t be able to upload collaborative documents like Google Docs).
Is your document too large? ChatGPT’s document size limit is 512MB, meaning it might reject files with hundreds of pages. If the document is a poor-quality scan, you might find that ChatGPT either produces an error message or an incomplete or inaccurate output.
Try clearing your browser’s cookies. Cookies can cause odd browser behaviour, including changes in document permissions or viewing errors. By clearing your cookies, you can allow ChatGPT to set up new, compatible cookies, which might allow better performance.
Also, if you’re using a VPN, try temporarily disabling it. VPNs can conflict with security protocols, restricting access to a document. Plus, some VPNs disable access to third-party cookies, preventing complete functionality for cloud-based services like ChatGPT.
You can check to see OpenAI’s service status. It’s unlikely that a service status will cause OpenAI to generate errors when reading documents, but it’s helpful to cover all bases.
Let’s say you need your document read ASAP, and you don't have time to check all the document settings. You can use another free AI virtual assistant, such as Claude or Gemini. If they fail to read your document, there’s likely something wrong with the document itself (such as it being corrupted or locked). Playing around with the document’s type, format, length and accessibility will hopefully resolve the issue quickly.
Take a moment to check the following:
Have you tried to read, analyse and ask ChatGPT to monitor for errors simultaneously? The results will likely be garbled and unsatisfactory. Though breaking down the prompts into more digestible actions is more time-consuming, you will likely experience a more coherent output.
Rather than just referring to a ‘balance sheet’, ask ChatGPT to read the ‘the balance sheet on page 9’ (for example). It can also be helpful to refer to the file type (e.g. PDF).
Prompting ChatGPT to adopt specific personas, such as an analyst or editor, can yield more detailed and informative responses.
Overall, specificity is best. Prompts based on vague verbs like ‘analyse’, ‘fix’, ‘adjust’, ‘check’ or ‘rearrange’ without qualification will likely produce disappointing results. For example, the AI might misunderstand the requested outcome or produce a (technically) correct result yet refuse to address the prompt’s core focus.
Take ten seconds to check that your prompt is as informative as possible. The results will speak for themselves.
If you have individual documents (not containing sensitive data), ChatGPT can be a quick and helpful tool. However, (like most commercial AI tools), ChatGPT can glitch and reject document reading. By following these six steps, you’ll unlock ChatGPT’s document reading ability without continually experiencing irritating errors.
If you’re looking for fast, accurate data extraction (without a side of error messages), try Evolution AI. And, for more tips about getting the best of Large Language Models like ChatGPT, follow us on LinkedIn and X (formerly Twitter).
