ChatGPT’s memory system has improved, especially for free users. OpenAI redesigned the chatbot’s method of storing user information, aiming to make interactions more personalized and efficient. The update includes a background process called “dreaming,” which helps retain context without needing explicit prompts. The new architecture offers users better control over how their data is handled and stored.
The initial memory feature, introduced in April 2024, required users to give direct instructions for ChatGPT to remember details. Over time, this led to outdated or irrelevant information. OpenAI’s “dreaming” system helped by synthesizing data from past conversations. However, the company admits the early version wasn’t sufficient. The new system combines both methods, creating a significant improvement in personalization and context handling.
Related: Thief uses Waymo to steal yoga clothes
Users can now see a “memory summary” generated by ChatGPT. This summary captures what the chatbot knows about them, including preferences and past interactions. Users can edit or add information, ensuring accurate data. For example, if someone mentions using a specific camera model, the chatbot might suggest compatible lenses during future photography-related chats.
The update also enhances how ChatGPT follows user preferences. If a user discusses travel plans, the chatbot tailors recommendations accordingly. A user planning a trip to Singapore might receive itinerary ideas with street photography spots based on past conversations. The system automatically updates memories to avoid suggesting outdated information, like mentioning a past trip as if it were upcoming.
Related: Matt Davies Stockton Provides Tips to Choose Reliable Online Stock Trading App for Beginners
OpenAI is rolling out the new architecture to Plus and Pro users in the US immediately. Free users will soon see similar improvements due to backend efficiency gains. For Plus and Pro accounts, the update allows more detailed memory retention. The feature will expand to other regions in the coming weeks, though global timelines remain unclear.
The new memory system works with another tool called “sources,” which shows what information ChatGPT used to generate responses. This feature, introduced with GPT-5.5 Instant, lets users edit or delete context if needed. Together, these tools improve transparency and control over data, though OpenAI hasn’t addressed privacy concerns directly.
Related: The best CRM software in Singapore
The update represents a major shift in how ChatGPT processes information. The “dreaming” system now runs more efficiently, reducing computational load while improving accuracy. Free users, who previously had limited access to advanced memory features, benefit from this change. The new architecture is designed to scale better, though its impact on mobile devices or slower internet connections is unclear.
The rollout highlights OpenAI’s focus on balancing user experience with technical feasibility. While the new system offers more personalized interactions, it also raises questions about data storage and user consent. OpenAI hasn’t provided detailed documentation on memory retention or security. For now, the update prioritizes usability, leaving privacy considerations for future updates.
