ChatGPT Memory and New User Controls

OpenAI has introduced a memory feature for ChatGPT, enabling the model to remember information across different chat sessions. This update reduces the need for users to repeat preferences and personal details, making future interactions more tailored and efficient.

How ChatGPT Memory Works

ChatGPT can now store information about users either through explicit requests to remember specific details or by automatically picking up details from the flow of conversation. This memory evolves over time as the user interacts more with the model.

Key examples of memory application include:

  • Formatting Preferences: Remembering that a user prefers meeting notes with headlines, bullets, and action items at the bottom.
  • Personal Context: Knowing a user owns a neighborhood coffee shop when brainstorming social media posts.
  • User Interests: Remembering a child's favorite animal to suggest relevant themes for a birthday card.
  • Professional Requirements: Recalling that a kindergarten teacher prefers 50-minute lessons with follow-up activities for lesson planning.

User Controls and Privacy

Users maintain full control over what ChatGPT remembers. Memory can be managed through the following mechanisms:

  • Direct Interaction: Users can tell ChatGPT to remember or forget specific information conversationally.
  • Settings Menu: Memory can be toggled off entirely via Settings > Personalization > Memory. Specific memories can be viewed, deleted, or cleared entirely under Settings > Personalization > Manage Memory.
  • Temporary Chat: This mode allows users to have conversations that do not use existing memories, do not create new memories, do not appear in history, and are not used for model training.

Regarding data usage, OpenAI may use provided content and memories to improve models for general users unless opted out via Data Controls. However, content from ChatGPT Team and Enterprise customers is excluded from training.

Memory for Team and Enterprise Users

For professional environments, memory allows ChatGPT to learn a user's specific tone, voice, and format preferences. This streamlines tasks such as drafting blog posts or coding in specific languages and frameworks without needing to repeat the technical stack for every new session.

Enterprise account owners have the authority to turn off the memory feature for their entire organization. As with other workspace data, memories for Team and Enterprise users are not used to train OpenAI's models.

Memory for Custom GPTs

GPTs can have their own distinct memory, which builders can choose to enable. This memory is separate from the main ChatGPT memory; for example, a specialized GPT for book recommendations might remember a user's favorite genres, but a different GPT for greeting cards would not have access to that same information. To use a memory-enabled GPT, the user must have the memory feature enabled in their general settings.

Evolution of Memory Capabilities

OpenAI has updated the memory system in stages:

  • September 2024: Memory became available to Free, Plus, Team, and Enterprise users, introducing notifications when memories are updated and a "Manage memories" shortcut.
  • April 2025: Memory became more comprehensive, referencing both "saved memories" (explicitly requested) and "chat history" (insights gathered from past chats) to tailor responses.
  • June 2025: Memory improvements rolled out to free users, providing short-term continuity across conversations, while Plus and Pro users retain longer-term understanding.

Sources