When attaching files to Deep Research it tries to open them (via python etc) but can‘t achive to do that for like 5-10mins. Then it switches to search for similiar content (based on title of the file etc) and just makes up everything on his own.
All other models (o3, o4-mini etc) don‘t have the problem with the same files ..
I tried switching up browsers, devices and even accounts but it does happen all the time.
You can‘t even get it to access a 1 page pdf and read through it.
Is there any way to fix this? I‘m in urgent need of deep research right now.
Does anyone other than me also face the same problem? I can‘t find people talking about it online ..
I’m having the same problem here. No matter the format of the files, Deep Research won’t read them. Files in docx or pdf format are not working, for example. The his nightmare started a pair of days ago! Please, if anyone solves the problem or finds a workaround, let us know!
no I did not and somehow it does not bother anyone? how can a feature that the pro mode (230€) is for, not work for 1 week and we do not even get a response from the dev team that there is such a bug and it will get fixed?
I am experiencing the exact same issue as a Pro account user, and based on the GPT explanation, it seems to be a recent platform-wide bug that needs to be fixed. It used to work like a dream…
Yes, this is a major issue! Please fix it immediately. This easily wasted the equivalent of an entire workday trying to make it work. I hope pro users are going to get some credits returned on the deep research prompts wasted since this bug became an issue.
Having the same issue, been going on for about a week now. Doesn’t matter the file type I try, it spends 15-30 minutes trying different attempts to open it but never can. Never had an issue before this week
I don’t think it will get fixed until this thread gets more traction and is elevated up. I recommend everyone share this thread with as many other users you know who are having the same problem and encourage them to comment here.
I did the exact same thing: spend at least 10h last week to try to fix/workaround this issue. Did at least test it 50 times because I need it right now and only within this upcoming week! I even contacted the support multiple times but they didn’t know of anything and asked me so many irrelevant questions ..
As far as I know it can’t even use python to make visualizations for web searches? So how come nobody noticed this bug before? Are there so few pro users out there that use deep research on a daily basis? Or is it that only a proportion of total users are affected? But then again: tested it on the accounts of friends and family and got the same results .. vpn usage, switching devices and accounts didn’t help.
I have both a Pro account and a Teams account and can confirm it does not work on either account. I have tried across multiple devices and browsers. However, Gemini 2.5 is working well in deep research mode and was able to handle my prompts without any editing to create content of similar quality as what I was getting through ChatGPT. This seems like a major misstep in a market this competitive.
In my case only chatgpt is capable of doing my task at the moment .. I could provide multiple data sets in a zip file and deep research could do every analyses I wanted .. unfortunately gemini doesn’t accept zip files or more than 10 files ..
I feel your pain. It is a major issue. Like I said, until this thread becomes the most popular bug reported, I doubt it will get any attention from them. Time to ask friends for a favor and get them to help us escalate this.
You also can report this issue via the help bot. It won’t help you, but if you ask it to have a human reach out about the issue, it will escalate it to a support specialist.
I did it already a few days ago, and nothing happened, sadly. I had to try out Claude.ai, and it works way better. GPT is experiencing numerous issues, and OpenAI appears to be unresponsive. I mean, it’s even impossible to open a ticket or something with a real person! I suggest we address this issue by bombarding social media with it.
It is finally marked as an issue .. https://status.openai.com/
Interestingly enough, just after fixing the API issue .. I think they ignored it on purpose for the last week to fix what pays more money.