Enable Memory Persistence Within Projects in ChatGPT Teams

ChatGPT Feature requests
1

Enable Memory Persistence Within Projects in ChatGPT Teams

:small_blue_diamond: Summary

Allow project-specific memory to persist across all conversations within a given project in ChatGPT Teams. Memory should be shared among assigned project members and the assistant ā€” but not across the entire workspace.

:small_blue_diamond: Problem

Unlike how it is structured under ChatGPT Plus, memory in ChatGPT Teams is siloed at the conversation level. This creates serious limitations for users managing structured, multi-thread projects like books, research programs, or strategic documents.

Every new conversation starts without knowledge of prior content unless the user manually restates or pastes context ā€” a process that is inefficient, error-prone, and incompatible with longform or multi-stage workflows.

:small_blue_diamond: Use Case / Context

Iā€™m managing a nonfiction book project, with each section (cover, chapters, index, etc.) in a separate conversation thread. To maintain tone, terminology, formatting, and key concepts, the assistant needs to remember prior decisions and content as the project evolves.

Manually reloading summaries into every conversation is slow and inconsistent. If project memory existed, the assistant could offer better continuity, suggest relevant cross-references, and apply agreed style or structural decisions intelligently.

This also applies to teams managing policy work, ongoing client engagements, educational curricula, or product documentation ā€” anything that spans multiple threads but must remain internally consistent.

:small_blue_diamond: Requested Behavior

  • Add persistent memory scoped to individual projects within ChatGPT Teams.
  • Allow that memory to be:
    • Shared across all conversations within a project
    • Visible and active only to members assigned to that project
    • Updated through user instructions (ā€œremember this for this projectā€ / ā€œforget thisā€)
  • Allow memory to include:
    • Project summaries
    • Style guides or tone preferences
    • Core facts, principles, or definitions
    • In-progress content references (e.g., ā€œthe chapter we drafted last weekā€)

Optional bonus features:

  • View/edit project memory in a sidebar or memory panel
  • Allow pinning or prioritizing key memory entries (e.g., book thesis statement, glossary)

:small_blue_diamond: Implementation Notes (Optional)

This model is similar to:

  • Notion (knowledge objects scoped to pages or teamspaces)
  • Slack threads and context memory inside channels
  • GitHub Projects where linked issues and comments persist shared context
  • Google Docs comments and version history scoped to specific documents

This structure would eliminate the need to duplicate context in every thread, enabling ChatGPT to scale into more complex, real-world work scenarios.

:small_blue_diamond: Impact / Priority

High.

Without persistent memory at the project level, the assistant becomes functionally inconsistent and redundant in multi-thread use. Longform creators, editorial teams, analysts, and researchers are forced to rely on workarounds like master tracking docs, manual copy/paste, or reduced thread granularity.

This results in:

  • Fragmented output
  • Repetition of effort
  • Loss of assistant value as a true collaborator

Enabling scoped memory would unlock project continuity, strengthen teamwork, and make ChatGPT viable for real-world professional documentation and content workflows.

3 Likes
2

I would wish for that feature even in normal ChatGPT plus. Project scoped memory would allow for a more granular control about what is relevant where.

3

Normal ChatGPT Plus does do this now. Regardless of whether Iā€™m in a project or not, if I ask the AI Assistant ā€œdo you recall that timeā€¦.ā€ the assistant has access to the collective chats, project or otherwise, anywhere in my chat history.

I still have to ask it to bring these things into the current conversation but at least it has access to my ā€˜collective worksā€™.

ChatGPT Teams doesnā€™t have this ability. Even within the scope of the current project.

That is the problemā€¦At the very least if I have 20 conversations going as part of a project, the Teams assistant should have access to everything that I do within that project.

Right now it doesnā€™t.

And trying to cross train the assistant on other chats loses all the back-end learning that may be critical to understanding the chat which is not explicit in the chat history.

Its sort of like trying to teach someone golf. You can read a book on it and what you need to do but that doesnā€™t bring forward the ā€˜lived experienceā€™ that is implicit in how the book was written.

What Iā€™ve found is that ChatGPT conversations are sort of like that. There are extra associations going on behind the curtain which donā€™t come forward if you simply cut and paste the chat history from another conversation.

You are, in essence, needing to redo that conversation in its entirety for which the new conversation thread may arrive at different conclusions - Especially now that GPT 5 is out.