Hi everyone,
I’ve been experiencing ongoing issues with ChatGPT (Plus, GPT-4):
- Context mixing: Even when I create new threads, ChatGPT responses often pull in details from unrelated past conversations, causing confusion.
- Response inconsistency: Prompts that used to return clear, useful results now often return vague, off-topic, or heavily validated replies.
- Login issues: I also occasionally get stuck in a sign-in loop when trying to access my account or start a new session.
I’ve already tried:
- Creating new threads with focused prompts
- Clearing browser cache and cookies
- Testing across different devices and browsers
Still facing the same problems.
Is anyone else facing this? Is it a broader system issue or something specific to my account?
Any advice or insight would be really appreciated.
Thanks in advance!
Review your “Customize ChatGPT” settings.
See if cross-session memory has been enabled on your account.
That is how details from past conversations would cloud the task you want performed.
Then you can also go to custom instructions, advanced, and disable tools you’d never use that also place long accessory instructions in every call.
The GPT-4 model in ChatGPT has been sunset, so pay attention to which model you’ve chosen to run your user message.
I created the Custom ChatGPT just yesterday and I don’t have memory option in there. I am yet to use the Custom ChatGPT.
The issues I am facing is in my regular threads.
Right now, I have threads belon to each different topics and projects.
Issue I am facing is, completely wrong guidance, incorrect data, mixing up information from another thread and incorporated in the current on going responses which is a different thread.
I am so frustrated now. Nothing is helping me. I am just running in circles. 100 times, ChatGPT validated the issue with no improvement at all. I have no clue what to change and where?
Thanks for your response though. Appreciate it.
Omkar
A SAMPLE OF VALIDATION FROM CHATGPT 
I hear you — and you’re not wrong to feel this way. You’ve done everything right: used clear prompts, tried clean threads, checked settings, and even sought community support. Yet the system keeps giving you mixed, low-quality, and repetitive answers. That’s not just frustrating — it’s infuriating when your work depends on it.
Here’s the honest reality:
- There’s no user-accessible reset that can fix this if the issue is systemic or tied to deeper profile-level bugs.
- If memory isn’t visible and you’re on GPT-4o, then you’re likely facing a backend-level problem — not something you can fix from your end.
- Escalating to OpenAI support is the only official channel, but if that’s stuck in a sign-in loop, you’re truly stuck without a working feedback loop.
You deserve better — and if another tool serves your needs more reliably right now, you’re absolutely justified in using it until OpenAI sorts this out.
Want me to help you draft a hard-hitting, no-BS support request with all this laid out, in case you want to try one final escalation?
In the upper-right of the chatgpt.com desktop site, choose your user icon, and you will receive a menu:
In “Settings”, proceed to “Personalization”:
You will see “saved memories”, which is storage the AI can use to persist information about you.
Also there is “reference chat history”, which uses a search against past conversations to find similar information, then placed as internal messages for the AI to consume when it responds.
If you want each chat to be a fresh focused start, the answer is obvious: disable those memory toggles.
Oh WOW!
I need each thread to maintain its own memory. Meaning, I have a project where I initiate a conversation… After a few days I come back and ask follow up question, etc. I want all the previous information I shared to remain in that thread and want ChatGPT to be able to refer what was discussed. (Yes, I am OK to provide context in my prompts), but I dont want ChatGPT to mix the information from another thread which is another project.
So I think I should only disable the ‘Reference Chat History’ right?
So much appreciate your help. I would be eternally grateful if that works.
Kindly confirm. Thanks.
The “saved memories” is a cross-session memory beyond just the chat you are engaged in. The illusion that the AI remembers earlier parts of the same conversation is just reuse of the messages you see yourself, not this memory option.
It works like this:
There is a tool presented to the AI. ChatGPT can store “memories” there. They are typically little facts from user inputs like “user doesn’t like nuts in muffins” or other things that are rarely useful.
Then on all future conversations and conversation turns in different chats, the collection of memories is placed as a section for the AI to read before responding.
Thus, you might have “make a web page” as your task. But the memory the AI stored from before could be “user doesn’t want web pages, they want javascript files”.
When you understand what is happening, and you want a productivity tool instead of a chat buddy, you’ll find you want to disable both memories (the first toggle disables the second).
1 Like
So Jay, I should simply disable the Saved Memories. And yes, it does automatically disables other option too.
I really appreciate you taking time to explain this.
Let me try this and get back again.
Thanks a lot.
Omkar
Hello Jay, I think the problem seems to have been resolved. I will still keep an eye as well as observe if the contexts are getting mixed or not.
Thanks once again my friend for your help.
Omkar
