OpenAI prepares automatic memory management for ChatGPT

OpenAI appears to be moving forward with the long-anticipated upgrade to ChatGPT’s memory system, internally referred to as “golden hour.” This update was inadvertently revealed through a recent FAQ change, indicating that the new memory management functionality will soon become available to Plus and Pro subscribers.

Update: Rolled out on ChatGPT web

The core improvement is automatic memory prioritization: ChatGPT will monitor which details are most relevant, keeping the model focused on topics that matter most to the user while quietly dropping less important information. This is expected to reduce clutter and help the system remain contextually sharp, especially for those who store numerous custom facts and preferences.

From what’s known, the system will factor in how often certain topics come up in conversations. Users aren’t locked in, manual control is preserved, with options to:

1. Turn off automated memory handling.
2. Perform advanced searches with filters (e.g., by date).
3. Restore previously removed memory items.

This addresses some longstanding requests from power users, who have relied on memory for things like custom abbreviations, workflow routines, or persistent preferences.

These changes will be surfaced primarily in the memory settings area of ChatGPT’s web and app interfaces. For Plus and Pro users who push the boundaries of the tool, the update could mean less manual housekeeping and smoother, more personalized long-term chats. OpenAI’s strategy here lines up with a broader push toward making agentic AI feel more persistent and user-aware without sacrificing privacy or control. This fits their current trend of focusing on premium feature improvements as a way to differentiate subscription tiers and support advanced use cases.