ChatGPT's Memory Feature Now Learns About You Automatically
OpenAI's chatbot builds and updates its profile of users without explicit prompts, raising new questions about control and accuracy.
ChatGPT's memory capabilities have evolved significantly since their introduction, shifting from manual user commands to an automated system that continuously builds and refines what the AI knows about you.
The change represents a fundamental shift in how the chatbot personalizes responses—and introduces new considerations about accuracy and user control.
From manual to automatic memory
When OpenAI first launched ChatGPT's memory feature in April 2024, users had to explicitly tell the bot to "remember this" for information to be stored. These memories were static, creating awkward situations where the AI would reference future events long after they had passed.
An update in June 2024 introduced what OpenAI calls "dreaming." The system now automatically determines what information is worth remembering and updates those memories over time. According to details first reported by WIRED, ChatGPT now handles three main functions: carrying forward useful context, following user preferences and constraints, and staying current as circumstances change.
The automation makes the feature more convenient, but it also means users have less direct control over what the AI considers important. If you spend a week researching DIY projects, ChatGPT may incorrectly assume you're a home improvement enthusiast. Research a vacation destination, and the bot might think you live there.
How to review and edit what ChatGPT remembers
Users can access their memory profile through the Personalization settings. In the web app, click your profile icon in the bottom left, select Personalization, and scroll to the Memory section. Mobile users can find it by tapping the menu button, then their profile icon, then Memory.
The interface displays a summary of what ChatGPT has learned, covering topics like interests and communication preferences. Users can highlight any item and choose "Make a correction" or "Don't mention this again." A prompt box at the bottom accepts direct commands like "forget that I'm a journalist."
Disabling the memory toggle stops ChatGPT from using stored information in responses, but doesn't delete existing memories. To completely wipe the slate clean, users must select the three dots at the top, choose Delete, and turn off memory—though re-enabling the feature will prompt ChatGPT to re-index past conversations.
Why it matters
As AI chatbots become embedded in daily workflows, the data they accumulate about users directly shapes the assistance they provide. Automatic memory systems like ChatGPT's dreaming feature offer genuine convenience, but they also make assumptions that may not align with user intent. For business users relying on AI for professional tasks, understanding how these memory systems work—and knowing how to correct them—is essential for getting accurate, relevant responses. The shift from explicit to implicit memory collection also raises broader questions about transparency and control in AI tools that are increasingly positioned as productivity essentials.
Temporary chats offer an alternative
For conversations users don't want remembered at all, ChatGPT offers temporary chat mode. On the web, clicking the dotted speech bubble in the top right starts a temporary session. Nothing from these chats is stored in memory, and they're deleted when closed.
The details of ChatGPT's updated memory system were first reported by WIRED.
This is an original analysis by the Omega editorial team. Source reporting: WIRED.
Want systems like this working for your business?
Book a Call
