OpenAI Introduced Dreaming Memory for ChatGPT

The new background system synthesizes conversation history to improve factual recall and accuracy for U.S. users.

Updated on Sept. 21, 2026 in Artificial Intelligence

Isometric editorial illustration of a matte server rack with cooling fins, representing automated data processing architecture. — OpenAI launched a background memory-updating process for ChatGPT called dreaming, designed to synthesize conversation history to improve factual recall for users in the U.S. AI Illustration. Upload story photo >

Live Poll

Are you comfortable with AI systems automatically tracking and revising your personal history over time?

In June 2026, OpenAI launched a background memory-updating process called dreaming for ChatGPT users in the U.S. This architecture continuously reviews conversation history to synthesize a living summary of user interactions.

Why it matters

The system was designed to address issues regarding information staleness, correctness, and scalability across millions of accounts. By adjusting stored data for elapsed time, the technology ensures that recommendations remain current for individual users.

The system reduces the compute required for free users by 5x while significantly boosting performance metrics. Users retain control over the feature by accessing the Personalization section in the settings menu to edit or delete specific memories.

The players

OpenAI

This is an American artificial intelligence research organization that develops advanced large language models including the ChatGPT platform.

ChatGPT

This is a generative artificial intelligence chatbot developed by OpenAI that uses a large language model to facilitate human-like text conversations.

The details

The dreaming process operates by scanning years of chat history to extract and revise relevant personal information in the background. OpenAI has expanded the memory feature to free users and increased storage capacity for those with Plus and Pro subscriptions.

ChatGPT originally launched its core memory feature in 2024.
OpenAI began rolling out the dreaming process in June 2026.
The feature continued to expand with additional user controls by mid-September 2026.

The Tech Race

This development marks a transition from simple information retention to dynamic knowledge management within generative AI platforms. It positions OpenAI to maintain its competitive lead by solving the persistent problem of context decay in long-term AI interactions.

Users can expect more accurate and personalized responses as the AI maintains a current profile of their preferences and history. The shift to a more efficient system also enables wider access for free users while providing tools to manage or erase stored data at any time.

The takeaway

The implementation of dreaming memory demonstrates a shift toward more sustainable AI architectures that prioritize long-term relevance over simple data volume. Users should regularly review their Personalization settings to ensure that the living summary of their interaction history remains accurate.