Deep Research delays and "phantom responses" — normal or not?

ChatGPT Bugs
, ,
1

Hey everyone,

Just wanted to check if anyone else has experienced this kind of behavior from Deep Research and what you’d recommend in my case.

I launched a pretty big research request — a long, structured prompt with around 9 sections — and ChatGPT said it had kicked off a full-scale deep research job. It mentioned a proper PDF would be sent later, after some asynchronous processing.

But here’s what’s happening:

  • It’s been almost three days, and I still haven’t received any final result or PDF.
  • When I ask ChatGPT to regenerate or show me earlier sections, the answer changes every time — no consistency, no sources, no signs that it’s pulling from an existing research result.
  • The model confidently claims the research is “still being prepared”, and that I should wait — but I haven’t seen any documentation or forum threads saying this is a thing (like 1–2 day background research?).

This makes me wonder:
Is this a real feature (long-running async research), or is it just the model hallucinating about how it works?

Also, I noticed that in large requests, the model tends to only return the final section, and skips the earlier ones completely.
Probably this is due to output length or context limits — which means I may be running into a cap without realizing it.
So I’m wondering — is Deep Research really meant for smaller scopes? Should I split up my request into multiple, smaller parts?

Would love to know if anyone has:

  • waited over a day and eventually got results?
  • confirmed whether ChatGPT can actually re-access a pending Deep Research job after a regeneration?
  • tips on how to structure large prompts so they don’t get clipped or ignored?

Thanks a lot in advance!

P.S. I did message support a couple of days ago, but still no response. Just curious — how long do they usually take to get back?

1 Like
2

YES! I am going on 9 days for a deep research request. It keeps saying it will send and apologizes. What is the issue?!!!

3

Actually, support got back to me and confirmed that this is, unfortunately, a hallucination.

At the moment, Deep Research doesn’t have the ability to generate PDFs, run asynchronously in the background, or restore past research attempts. If you don’t receive a result within ~30 minutes, you should assume that the research has either finished in its current state — or stalled.

Still kind of amazed how confidently ChatGPT can explain its imaginary inner workings :slight_smile: