Subject: Feature Suggestion — Add Conversation Context Control to ChatGPT
Dear OpenAI team,
I’d like to suggest a feature that could make ChatGPT considerably more convenient for users: giving users control over how much personal and historical context is used in a particular conversation.
This idea builds directly on functionality that ChatGPT already provides. When using ChatGPT, users can choose or enable different capabilities for a conversation, such as creating images, adding files, or using longer reasoning. I think conversation context should be treated similarly.
The problem is that Memory and personalization are extremely useful, but their usefulness depends on the conversation.
Sometimes I want ChatGPT to remember my preferences, ongoing projects, terminology, previous decisions, and other relevant information. This saves me from having to repeatedly explain the same things before getting to the actual question.
But sometimes I want to approach a subject independently — without my previous conversations, opinions, assumptions, or personal context influencing how ChatGPT interprets the discussion.
Instead of forcing users to choose between personalization and no personalization globally, ChatGPT could provide a per-conversation context option alongside the existing capability controls.
For example:
Context
* Current conversation only
* Relevant personal context
* Full personal context
* Ask before using personal context
The exact implementation could obviously be different, but the important idea is that the user should be able to decide the context scope when starting or configuring a conversation.
This would provide two major benefits.
First, convenience. When I want personalization, I shouldn’t have to repeatedly explain everything ChatGPT already knows about me just to get to the actual subject.
Second, control. When I want a fresh, independent discussion, I shouldn’t have to manually explain that I don’t want my accumulated context influencing the conversation.
I think this is particularly valuable because it doesn’t require abandoning the personalization system you’ve already built. It would simply give users another layer of control over it.
In other words, I’m not suggesting that ChatGPT should become less personalized. I’m suggesting that personalization should become more user-directed and conversation-specific.
As ChatGPT continues adding capabilities and giving users more control over which capabilities are used in a conversation, context access seems like a natural addition to the same model.
This is primarily a suggestion for user convenience, but I also think it would make ChatGPT more predictable and transparent. Users would know not only what ChatGPT can do in a conversation, but also what information it is allowed to use to understand that conversation.
I’m sharing this as a practical product suggestion based on using ChatGPT extensively. It seems like a relatively natural extension of functionality that already exists, and I think it could be genuinely useful for a broad range of users.
Please consider adding some form of per-conversation context/memory control to the product roadmap.
Thank you for continuing to improve ChatGPT and for giving users opportunities to contribute ideas and feedback.
Best,
Himanshu