Has chatgpt gotten worse at reading text from PDFs?

Assessing the Current Capabilities of ChatGPT in Extracting Text from PDFs

In recent months, users have observed noticeable changes in ChatGPT’s ability to interpret and extract text from PDF documents. Historically, the model demonstrated strong performance, even with older or lower-quality scanned documents. For example, a few months ago, one user successfully uploaded a series of antiquated incorporation records from the 1940s, and ChatGPT was able to accurately process and read the content without issue.

However, recent reports suggest a decline in this functionality. Several users now encounter messages such as “No text could be extracted from this file,” even when working with PDF documents that originate from digital sources like Microsoft Word—print-to-PDF files that typically maintain higher quality and accessibility.

This shift raises questions about the consistency of ChatGPT’s document processing features, particularly its ability to handle complex, low-quality, or older scans. It’s worth considering whether recent updates or adjustments to the underlying models have impacted this functionality, or if the changes are related to the way PDF files are being formatted, compressed, or encrypted.

For professionals and researchers relying on ChatGPT for document analysis, this observation underscores the importance of testing the tool’s current capabilities with your specific document types. If you’ve noticed similar issues, sharing your experiences can help the wider community understand the scope and potential causes of these changes.

As AI tools continue to evolve, ongoing user feedback is crucial. It remains to be seen whether improvements or new features will restore this functionality or if alternative methods will be necessary for extracting text from complex PDF files.