Skill · OpenAI ChatGPT release notes and help center
Isolate Client Work With Project-Only Memory
Photo: (Augustin-Foto) Jonas Augustin on Unsplash
ChatGPT's project-only memory feature lets you keep professional context-client details, writing tone, project goals-completely separate by project so the AI never mixes up context between different clients or work streams. Enable it once and ChatGPT will retain preferences and decision history without relearning them each session.
ChatGPT's memory system now includes a powerful isolation feature called project-only memory that solves a common professional pain point: context bleeding across different clients or work streams. When you enable project-only memory on a project, ChatGPT uses memories and conversations only from that specific project-it won't reference personal memories or chats from outside the project, keeping sensitive client information compartmentalized.
To set this up, create a new ChatGPT project for each client or major work stream. Go to Project Settings and toggle on project-only memory (personal memory must be enabled globally first, in Settings > Personalization > Memory). Once enabled, ChatGPT will automatically extract and remember details specific to that project: the client's communication style, their brand voice, previous decisions you made together, project constraints, and even technical specifications.
This is particularly valuable for consultants managing multiple clients. Instead of telling a client "I work with strict confidentiality" and then accidentally mentioning another client's approach, you can structure your ChatGPT usage so each client's project is completely isolated. A marketing consultant can have one project for Client A with their brand guidelines and campaign preferences stored in memory, and a completely separate project for Client B where ChatGPT won't reference Client A's information at all.
Another professional use case: researchers working on multiple competing analysis streams. Each research thread can have its own project with isolated memory, so ChatGPT won't accidentally cross-contaminate hypotheses or methodologies between simultaneous studies. The feature is available on Team and Enterprise plans where admins can control whether memory is enabled, and on Plus/Pro plans for individual users.
Sources & credits
Original source: OpenAI ChatGPT release notes and help center