Does ChatGPT actually look at PDFs?

When I present a PDF file to ChatGPT, I notice that the interaction can sometimes be a bit limited.

I understand that GPT-3.5 isn’t doing any visual analysis and that the text from the document is just extracted.

However, is GPT-4 or GPT-4o actually able to visually look at the document itself? For example, if the document has graphs/charts, would it be able to identify those?

Additionally, let’s say that I have a document which has no text layer. Would the models be able to natively engage in OCR (without having to run python code) of the document instead of me having to manually screenshot each page and send images?

1 Like