Shared Projects: Collaborating as a Team

Collaborative AI for Teams

So far, everything we’ve built in this course—projects, instructions, and sources—has been personal. Shared projects change the game: multiple people work within the same project, accessing the same files, instructions, and conversation history. This marks the shift from individual AI to team AI, solving a problem you may have already encountered: when every colleague “tinkers with ChatGPT on their own,” results vary wildly in quality and tone because there is no shared context.

A shared project works like a standard project, except ChatGPT leverages everything inside it—conversations, files, and instructions—to respond to every member. Everyone picks up where others left off. Sharing is available on Free, Go, Plus, Pro, Business, Enterprise, and Edu plans, both on the web and in mobile apps.

Creating and Sharing a Project

Start with a complete project, as in Lesson 8, including a name, instructions, and files. Do not share an empty shell; the framework must exist before you send invitations. Next, select Share to open the sharing panel. On personal plans, you invite people one by one (Only those invited). On Business, Enterprise, and Edu plans, you can also invite a group or share a link restricted to your workspace. Finally, an owner can open the project to Anyone with the link: anyone logged into ChatGPT with the link can join. Each invitation grants one of two access levels:

The logic is simple: the owner controls the framework and membership, Edit access keeps the content alive, and Chat access lets users work within it. Assigning the right level at the invitation stage prevents you from later discovering that three different people have rewritten the instructions.

Limits by Plan

File and collaborator limits depend on the owner’s plan: 5 files and 5 collaborators on Free; 25 files and 10 collaborators on Go and Plus; 40 files and 100 collaborators on Pro. In Business, Enterprise, or Edu workspaces, a project supports up to 100 collaborators and 40 files. Workspace administrators can also restrict who creates or shares projects.

What Changes in a Shared Project

A shared project automatically and permanently uses project-limited memory: ChatGPT has no access to members’ personal memories, custom instructions, or external conversations. What each person knows about you stays with you. Members can start new conversations and create a branch of an existing one to explore a different angle without losing the original thread; the profile icon next to a conversation indicates who started it. Anyone can archive or delete conversations they created. If a member leaves the project, they can take a copy of their conversations to a new project.

What This Looks Like in Business

Consider three typical teams. A marketing team creates a “Content Strategy 2026” project: instructions cover the editorial charter, brand voice, and target keywords; files include the editorial calendar, personas, and approved article examples. The content lead is the owner, writers have Edit access, and the intern has Chat access. Each writer opens a conversation for their article, and everyone produces work with the same context and rules.

A product team sets up a “Product Roadmap” project using the same model: vision, technical constraints, and methodology go in the instructions; the backlog, user stories, and mockups go in the files. Finally, a legal team centralizes its “Standard Contracts”: mandatory clauses and internal terminology in the instructions, templates and terms of service in the files. The pattern is always the same: a repository managed by a lead and leveraged by the whole team.

Best Practices for Sustaining a Shared Project

First, governance: every shared project needs an identified owner who writes and maintains the instructions, manages reference files, and controls access. Without this lead, instructions become inconsistent and files pile up without organization.

Second, instructions: in a collective setting, they must be more explicit than for personal use because they serve as the rules of the game for people who do not confer before every conversation.

## Team Context
We are the DataFlow communications team (5 people).
This project centralizes marketing content creation.

## Common Rules
- Tone: Professional, informal (use "you"), no excessive technical jargon
- Language: English (English technical terms accepted)
- Review: All content is reviewed before publication
- Confidentiality: No client data in conversations

## Process
1. One conversation per content piece
2. Iterate until the final draft
3. Copy to the team tool for approval

Third, conversation discipline: one conversation per content piece, and use a branch rather than hijacking a thread when you want to test a different angle on a colleague’s work. Two final warnings. First, size: a project with twenty people and hundreds of conversations becomes unmanageable; create multiple projects by team or initiative instead. Second, confidentiality: all members see the project’s files and conversations. Never share information with people who should not have access, and reserve Anyone with the link for non-sensitive content.

Your Turn

A project for two. Create a project with team instructions, select Share, and invite a colleague—or your own secondary email address if you are working alone—with Chat access. Ask them to open a conversation in the project, then verify from both sides that the instructions and files are accessible. Next, upgrade them to Edit access and have them add a file. Finally, create a branch of one of their conversations and observe that the original thread remains intact.

Official Sources: Projects in ChatGPT (OpenAI Help Center: sharing, access levels, links, plan limits, project-limited memory, branching, member departure)

Key Takeaways

• Project sharing is available on all plans; ChatGPT leverages all project content for every member
• Two access levels: Edit (instructions, files, invitations) and Chat (view and chat); the owner manages members
• Files and collaborators vary by plan: from 5 and 5 on Free to 40 and 100 on Pro and in workspaces
• A shared project permanently switches to project-limited memory: no personal data leaks in
• A responsible owner, explicit instructions, one conversation per content piece, and caution with open links

Test Your Knowledge

Writing blocks, memory, projects: three tools, three uses. Can you tell them apart?

1. What replaced Canvas, and what action remains essential?

Answer: As of May 28, 2026, writing blocks and code blocks appear directly in responses. The essential action remains unchanged: edit what you know yourself, and select a passage before requesting a targeted revision.

2. What can a code block do that plain code text cannot?

Answer: Copy it, modify it, open it in full screen, preview HTML, React, SVG, Mermaid, or Vega, and execute supported Python with a console in an isolated environment.

3. Memory or custom instructions: when to use which?

Answer: Use instructions for what must always apply (language, tone, business context); use memory for what evolves over exchanges. You can review and correct memory in settings, and deleting a conversation does not necessarily erase the memory derived from it.

4. What does a project offer compared to isolated conversations?

Answer: A space that groups files, instructions, and conversations: every new conversation inherits them. With project-limited memory, it becomes airtight, ideal for a client file.

5. What do shared projects change?

Answer: Multiple people work with the same files, same instructions, and shared conversations, with Edit or Chat access; memory is limited to the project, with no access to members’ personal memories.

When combined effectively—instructions for the permanent, projects for the file, and writing blocks for the deliverable—these tools make ChatGPT a complete workstation.