ChatGPT Gets to Know You Over Time: The New Memory

With its improved memory, OpenAI wants conversations with ChatGPT to feel more personal and less disjointed: rather than re-explaining your context every time, ChatGPT builds its own picture of you and keeps it up to date on its own. The aim is for memory to feel less like a technical setting and more like working with something that gradually gets to know you, the way a person would. For everyday work, that pays off most where you regularly kick off similar writing, analysis, or planning: less repetition, and a quicker start on recurring tasks. We’ve pulled together what’s known so far, and what to watch out for when you put it to use.

 

Info Value: Context that grows with you
UseCase Use case: Recurring tasks
Zeit Reading time:
4 minutes
Schwierigkeit Difficulty: Beginner

OpenAI calls the new architecture “Dreaming”: it synthesises context from past interactions in the background to keep memory fresher, more continuous, and more relevant. The practical upside is clear and immediate — new chats no longer start from zero, but closer to what you actually need. Control still matters, though: according to OpenAI, you can continue to manage saved memories and steer the memory feature from the settings. One thing to keep in mind is the rollout: OpenAI says the new memory experience is starting with Plus and Pro users in the US, and will be expanded from there.

  1. In ChatGPT, open Settings and go to Personalisation > Memory. That’s where you switch the feature on or off and find its controls, once it’s available to you.
  2. To see what ChatGPT has gathered about you, open the Memory Summary. There you can read the key points — and adjust them by typing your change directly into the text box at the bottom of the summary.
  • Weigh up work data: Don’t put confidential client, project, or personnel data into memory-enabled chats, files, or connected apps unless your internal rules expressly allow it. For sensitive cases, switch memory off or use a Temporary Chat.
  • The controls aren’t entirely intuitive: Switching off saved memories also turns off ChatGPT’s ability to draw on your chat history — both sit behind a single switch. And deleting individual entries doesn’t necessarily remove all the underlying information; when it matters, check your chats, archived chats, files, and connected apps as well.
  • Keep roles cleanly separated: If you use ChatGPT for very different kinds of work, saved preferences can colour answers where they don’t belong. For sensitive or clearly bounded tasks, a fresh context is worth it.
  • Not available everywhere yet: Because the new memory experience is rolling out in stages, we haven’t yet had the chance to test it over a longer stretch of everyday work. How reliably it holds up there isn’t something we can say for certain yet – but that’s down to the staged access, not the feature itself.
  • Spell out what matters: ChatGPT now picks up a lot from your chats on its own — you no longer have to actively “save” your preferences. Even so, for the things that genuinely matter to you, it’s best to state them explicitly. A clear instruction — a quick “Please remember going forward that …” — is more reliable than what the model infers in the background.
  • Use the memory overview: Keeping memories current is something ChatGPT now takes care of itself. Even so, it’s worth glancing now and then at the summarised overview of what the model has remembered about you. It shows you the picture it has of how you work — and lets you steer if something no longer fits.
  • Check in before important tasks: Ask ChatGPT to summarise the relevant saved assumptions. It’s a fast way to see whether the context still fits the job in front of you.
  • Not everything needs remembering: ChatGPT decides for itself what to keep from your conversations and fold into its picture of you. But if you’d rather a particular chat didn’t feed that process at all — a one-off test, a bit of side research, something confidential — use a Temporary Chat: these conversations neither draw on existing memories nor create new ones.

ChatGPT’s improved memory works quietly in the background. The real benefit only emerges over time: instead of typing the same instructions in at every session, you gradually build up a context that makes recurring tasks noticeably easier — regular summaries, drafts, or analyses with fixed style requirements. And that’s exactly the heart of the new version: keeping those memories current is now ChatGPT’s job, not yours — it adds to them and updates them in the background, rather than leaving you to tend a list. That’s the real step forward — with the flip side that the model largely forms its picture of you on its own. Your job, then, is no longer one of managing but of paying attention: knowing what ChatGPT has assumed about the way you work. That keeps memory what it’s meant to be: a genuine advantage, not a blind spot.