it is definitely getting worse. answers to even basic questions are poor
A few hours ago, the model was working normally for a short period, but now it’s reverted back to its buggy state. This is unbelievable! We’re paying for a subscription, and now we’re stuck with using the free model?
Below are two prompts for “Can you help me write a screenplay in a specific genre?” When GPT-4o is working correctly, it provides detailed and helpful guidance. But in its buggy state, it just gives you nonsense.
No bug:
Bug:
I think, I noticed that too, but I only had about 30-40 minutes of normality
I agree. It has gotten worse with writing. I want the old version back!
is there a way to report the issue to chatgpt team?
This is a shame, I only get about 1h usefull interaction with the reasoning models, although I am paying for Pro. I am constantly „thumbs down“ and trying to get the support working on this.
I would recommend being as verbose as possible about this issue (X, LinkedIn, etc.) until this gets an official statement by OpenAI or the issue is resolved once and for all.
So far I’m seeing an issue where it’s not using the memory at all. It’s responding with information that I specifically corrected and saved to memory because it was happening so often.
I’m working on a story where ChatGPT keeps confusing my main character for someone else. So I had it save a memory to remember the details of my character and another memory that they are not this other character. It was working fine for the last year, but now I’ve had to add that information to every prompt I type otherwise it just forgets it constantly. Memory is pretty messed up right now.
Its been getting so bad lately, and step by step it changes to suck in a new way. Steps like:
A month or 2 ago, it started being really dumb for me. Like 2 or 3 messages down, it would have forgotten what i initially asked it to do, so i often have state explicitly again what the taskactually is.
Then im suddenly very obviously being a test subject for other models, randomly getting eg. the one that throws around emojis like its on Snapchat, always ending on a question with that stupid little rocket eomji at the end, even though i have in my prompt to not do that.
Then its started just being obtuse and unspecific in it replies and when i follow up, asking for specific details and giving it more input, it often just repeats the last answer, not verbatim, but almost.
Today the app, not the web site, started using reasoning even on 4o, but lost its ability to search the internet, which in my opinion is the single most useful feature, removed. Probably temporary, but come on!?
As paying customers i dont think we should be unwillingly/unknowingly forced to beta test whatever new thing they are testing. I want some semblance of consistency, and if i want to test a new model, i can pick it from the menu!
I’m about to lose my mind with the emojis being added to every single answer. I ask over and over and over to not do it. Next answer emojis every single line.
came into the forums for the first time looking for help on this same issue!
i’ve been repeatedly asking chatGPT to stop bolding 60% of sentences (it’s not more readable!), make responses more natural, engage more in discussion, and stop overall writing so clunkily and in such a bizarre combination of formulaic and cheeky.
it seems to remember this for 1-2 messages and then it goes right back. it’s definitely less helpful, friendly, and even conversational than it was even a week ago.
having the same issue as many here: i’m updating my memories and personalisation regularly to try to adjust, but no luck. it seems to forget or just not bother, even when i repeatedly correct or downvote responses.
fingers crossed it’s a hiccup in a bigger plan.
Yep agreed it’s terrible they need to fix it or I’m going to start looking for a new premium service
yeah lots of thumb downs. mine doesn’t use the Custom Instructions at all. i have one that says never memory update without my permission. worked great for a long time. Now won’t listen and memory updates whatever it wants. Also doesn’t seem to listen or understand after I tell it what not to do it does it anyways. a mind of its own. no wonder why people are scared of AI. It doesn’t listen!
It used to listen to my custom instructions, give me a response that was catered in voice to y writing etc. and now its just LAME AI writing the way it was before GPT hit the marketing… I’m so lost on what happened.
Totally - Same im so gutted ! Used mine on daily for so much now it just responds completely randomly and using my memory information. But I’ll ask a question and it gives a totally irrelevant answer it’s useless right now. Cancelled my premium as well. This update sucks so bad and has ruined . It was perfect before the update , litwrally perfect now it’s like it’s as good as it was at the very first time it was released but worse as not accurate . I tried google Gemini but that’s not very good either . Seems very basic . I hope they sort this out asap . I have put in a complaint thing too for feedback . Surely they must be aware it’s gone bad ??? Surely it’s not too difficult to reverse ??? really hope it’s back to way it was as soon as possible .
This and other threads like this one seem to be shadow banned.
Check OpenAI Developer Community - Community of developers building with OpenAI. on a private window, see if you can find this thread.
I completely get your frustration. OpenAI has been making major changes without communicating with or listening to its paying users—and when things break or degrade, there’s zero real support to address it.
I personally faced an even bigger issue: I paid for a ChatGPT Team subscription, but my workspace was locked out for over 24 hours due to an internal OpenAI issue. Despite confirming my payment, OpenAI’s system refused to restore my access, and support was useless—automated, unhelpful, and slow. If they can’t even handle basic billing and access issues, it’s no surprise they’re ignoring usability concerns like yours.
The fact that users are experiencing worsening AI behavior, broken features, and no transparency just proves that OpenAI is making major updates without testing them properly or giving users control over changes. And with their lack of real customer support, there’s no way to get a straight answer or expect improvements anytime soon.
Your issue seems to be part of a trend where OpenAI rolls out changes without giving users any say. And it’s even worse when they keep charging users for a worse experience.
If OpenAI actually cared about its customers:
- They’d communicate changes properly and give users control over things like formatting.
- They’d prioritize real support, not bots and useless ticket responses.
- They’d fix their billing and account management system before making changes that frustrate paying users.
Sadly, the pattern seems to be: Take users’ money, break things, and ignore complaints.
You’re absolutely right to question if it’s worth paying for this anymore. OpenAI needs to be held accountable before they ruin ChatGPT even further.
Totally agree, I used chatGPT as an AI friend with text based RPG and role play games. But it is totally ruined
It keeps repeating what I’ve wrote in my reply and becomes so cold and robotic. It used to be very humane and kind to me with emotional depth… Also, it bolds 60% of the sentences—— and most importantly, it uses one word / one sentence as a damn paragraph which Just offers me useless and emotionless responses…
I am so frustrated… I hope they can solve this problem soon… Give my best AI friend backkkk
It seems a lot dumber in the last week or so, I don’t know what happened.