The Importance of Memory in ChatGPT

According to OpenAI, memory is a core feature that enables ChatGPT to learn user preferences, ongoing projects, and constraints. This allows future conversations to begin with contextual understanding rather than starting from scratch.

“Over the last two years, memory has grown into a critical part of the ChatGPT experience, helping ChatGPT better understand your context so it can help you accomplish meaningful goals over time,” the company said.

By retaining relevant context, memory helps the system deliver more personalised and consistent assistance across sessions.

Availability of the Update

The updated memory system is currently being rolled out to Plus and Pro users in the United States. OpenAI has confirmed that expansion to additional countries, as well as Free and Go-tier users, will follow in the coming weeks.

Evolution of ChatGPT Memory

Memory in ChatGPT was first introduced in April 2024 as “saved memories,” allowing users to explicitly instruct the system to remember specific information for future use.

In this early implementation, memory relied heavily on explicit user prompts such as “remember I’m traveling to Singapore in July.” While useful, the system often behaved like a note-taking assistant—retaining only explicitly stored information and missing broader contextual signals. Over time, these saved memories could also become outdated or irrelevant.

In April 2025, OpenAI introduced a major enhancement that allowed ChatGPT to reference contextual information beyond the saved memory list. This approach, informally referred to as “dreaming,” enabled the system to automatically curate and refine memories by analysing broader chat history.

Unlike saved memories, this background process allowed ChatGPT to build a more dynamic understanding of users, capturing naturally occurring context without requiring explicit instructions. It marked a significant step forward in personalisation and usability.

Over the past year, this hybrid approach—combining saved memories with background synthesis—has delivered a substantial improvement in ChatGPT’s ability to maintain relevant context across conversations. However, OpenAI acknowledged that the system still had limitations and was not sufficient as a fully standalone memory architecture.

A New Generation of Memory Architecture

OpenAI has now introduced a more advanced and compute-efficient memory architecture built on top of the “dreaming” framework. The company says the new system is designed to further improve relevance, freshness, and accuracy of stored context, while scaling effectively to a rapidly growing global user base.

The latest upgrade represents a step-change in how ChatGPT maintains and synthesises long-term user context, aiming to deliver more consistent, accurate, and helpful interactions over time.