Insights

ChatGPT Projects vs Custom GPTs: A Practical Guide for Team Collaboration

Teams get more value from ChatGPT when useful context can move with the work. Projects, Shared Projects, and Custom GPTs each support that goal, but they solve different problems.

Projects organize related chats, files, and instructions into a persistent workspace. Shared Projects extend that workspace so teammates can contribute from the same context. Custom GPTs serve a different role: they create reusable assistant behavior for defined tasks, teams, or knowledge areas.

The distinction matters because these tools are complementary, not interchangeable. Projects answer, “Where does the work live?” Custom GPTs answer, “How should the assistant behave?”

Key Observations

  • Use Projects for continuity. They keep related chats, files, instructions, and context together over time.
  • Use Shared Projects for collaboration. They help teammates work from the same project context instead of scattered individual chats.
  • Use Custom GPTs for reusable assistant behavior. They are better suited for repeatable tasks, defined knowledge areas, and consistent operating styles.
  • Use connectors carefully. Availability depends on plan, admin settings, workspace permissions, and rollout timing.
  • Design the workflow before choosing the tool. The right setup starts with how the team works, what context matters, and who should have access.

ChatGPT Projects, Shared Projects, and Custom GPTs at a glance

ToolPrimary roleBest use
ChatGPT ProjectsPersistent workspaceKeeping chats, files, instructions, and context together
Shared ProjectsCollaborative workspaceLetting teammates work from the same project context
Custom GPTsReusable assistantCreating defined behavior for repeatable tasks
ConnectorsData access layerBringing approved work sources into supported ChatGPT experiences

The simple rule: use Projects for continuity, Shared Projects for collaboration, Custom GPTs for repeatable behavior, and connectors for governed access to work data.

What ChatGPT Projects do well

A ChatGPT Project gives a team or individual a dedicated place for related work. Instead of starting each conversation from scratch, a project can keep relevant chats, files, and instructions together.

That makes Projects useful for work that changes over time, such as:

  • Planning a campaign
  • Drafting a grant proposal
  • Managing a client engagement
  • Building a knowledge base for a recurring process
  • Keeping research, decisions, and drafts in one place

Projects are best when the main need is continuity. They help preserve context so the work can continue more efficiently across sessions.

What Shared Projects add

Shared Projects extend the project model from individual organization to team collaboration. Instead of one person holding the useful context in a private workspace, teammates can contribute to a shared project and build from the same source of working context.

That matters for small teams because AI work often becomes fragmented. One person has the prompt. Another has the files. Someone else has the final draft. Shared Projects give the team a clearer place to gather the working material.

OpenAI currently describes project sharing as available across consumer and business plans, with Business, Enterprise, and Edu workspaces adding more team-oriented controls and admin considerations.

For teams, the practical value is not just access. It is continuity, visibility, and better handoff.

What Custom GPTs do well

Custom GPTs are reusable versions of ChatGPT configured for a specific purpose. They can include instructions, knowledge, capabilities, and selected tools that shape how the assistant behaves.

A Custom GPT is useful when the team needs repeatable assistant behavior, such as:

  • A proposal review assistant
  • A support response assistant
  • A brand voice assistant
  • A policy lookup assistant
  • A research intake assistant
  • A sales enablement assistant

The main value is consistency. A well-built Custom GPT gives people a clearer starting point than a blank chat because the assistant already has a defined task, tone, and operating style.

For Treyworks clients, this is often where “skills and workflows” become more useful than one-off prompting. The goal is not to make everyone an AI expert. The goal is to give the team a repeatable way to use AI on work that already matters.

Projects and Custom GPTs are complementary, not the same thing

Projects and Custom GPTs both bring more context into ChatGPT, but they serve different jobs.

Use Projects when the main need is continuity, organization, and shared context. Projects are useful for long-running work where files, conversations, and instructions need to remain connected over time.

Use Custom GPTs when the main need is specialization and repeatability. A Custom GPT is the better choice when the organization wants a reusable assistant with a defined task boundary, tone, knowledge base, or operating style.

Use both as part of the same AI operating model when a team needs shared workspaces and repeatable assistant behavior. Today, Custom GPTs cannot be used inside Projects, so the practical design question is how each tool should support the broader workflow.

A practical way to decide

Start with the work, not the feature.

Ask these questions:

  • Where does the work live? If the answer is “we need a persistent place for files, chats, and decisions,” start with a Project.
  • Who needs to contribute? If several teammates need access to the same working context, consider a Shared Project.
  • What behavior needs to repeat? If the answer is “we need the assistant to follow a defined process every time,” consider a Custom GPT.
  • What data should the assistant access? If the answer involves Google Drive, Microsoft OneDrive, SharePoint, GitHub, or other connected apps, review connector availability and admin controls before rollout.
  • What should stay private? Shared AI work needs clear boundaries around client data, internal decisions, and personal context.

This sequence keeps the decision grounded. It also avoids building a Custom GPT when the team really needs a workspace, or creating a Project when the team really needs a reusable assistant.

Connectors are the data layer

Connectors can bring approved work sources into supported ChatGPT experiences. Depending on the plan and workspace settings, those sources may include tools such as Google Drive, Microsoft OneDrive, SharePoint, GitHub, and other business systems.

Connector availability changes quickly, so teams should treat it as plan-specific, admin-dependent, and subject to rollout timing.

That means connector planning should include:

  • Access control: Who can connect which sources?
  • Data boundaries: What information should be available to ChatGPT?
  • Workspace policy: What do admins allow across the organization?
  • Use-case fit: Does the connector support the actual workflow, or is it just technically available?
  • Maintenance: Who reviews access when people change roles or projects end?

Connectors can make AI more useful because they reduce the need to manually upload and re-explain context. They also increase the importance of good governance.

Build the workflow before you share the workspace

Shared AI tools work best when the team agrees on how they should be used.

Before rolling out Shared Projects or Custom GPTs, define a few operating rules:

  • Name the use case. Be clear about the job the workspace or assistant is meant to support.
  • Set ownership. Decide who maintains instructions, files, and access.
  • Document the boundaries. Identify what belongs in the workspace and what should stay out.
  • Review outputs. Keep human review in place for client-facing, legal, financial, hiring, or policy-related work.
  • Revisit the setup. AI workspaces should improve as the team learns what helps and what creates noise.

The point is not to slow the team down. Good governance makes the system easier to trust, easier to maintain, and easier to hand off.

Example: A nonprofit development team

A nonprofit development team might use a Shared Project for a grant cycle. The project can hold the grant guidelines, prior proposals, program notes, budget context, and draft language.

That same organization might use a Custom GPT as a reusable proposal review assistant. The assistant could check draft language against the nonprofit’s voice, required sections, funder priorities, and internal review standards.

The Shared Project gives the team a place to work. The Custom GPT gives the team a consistent review process.

Those are different jobs. Together, they create a more durable workflow.

Example: A professional services firm

A professional services firm might use a Project for each client engagement. The project can hold meeting notes, strategy documents, draft deliverables, research, and decisions.

The firm might also use a Custom GPT for internal quality review. That assistant could apply the firm’s style guide, check for missing assumptions, flag unclear recommendations, and suggest next-step questions.

The Project keeps the client work organized. The Custom GPT helps the team apply a repeatable standard across engagements.

Why this matters for small teams

Small teams rarely have extra capacity for complicated systems. They need AI workflows that reduce coordination cost, improve consistency, and fit into how work already happens.

Projects, Shared Projects, Custom GPTs, and connectors can all help. The value comes from choosing the right role for each one.

When the setup is clear, AI becomes easier to use across the team. People know where context belongs, which assistant to use, what sources are approved, and who owns the workflow.

That is where ChatGPT starts to move from personal productivity into team operations.

FAQ

Are Shared Projects the same as shared chat links?

No. A shared chat link gives someone access to a specific chat view. A Shared Project gives approved members access to the broader project workspace, depending on plan, permissions, and sharing settings.

Can a Custom GPT be used inside a ChatGPT Project?

OpenAI currently says Custom GPTs cannot be used inside Projects, with or without sharing. Teams can still use Projects and Custom GPTs as part of the same operating model, but they should treat them as separate surfaces.

Should our team start with Projects or Custom GPTs?

Start with Projects when the team needs a shared place for ongoing work. Start with a Custom GPT when the team needs a repeatable assistant for a defined task. If both needs exist, map the workflow first. Then decide what belongs in the project, what belongs in the assistant, and who owns each part.

Do connectors replace uploaded files?

Not always. Connectors can reduce manual uploads when supported sources are available, but availability depends on plan, workspace settings, app permissions, and rollout status. Some teams may still use uploaded files for specific projects or controlled workflows.

Are Custom GPTs still useful if Projects exist?

Yes. Projects organize work. Custom GPTs shape assistant behavior. A team may need both, especially when it wants shared context for ongoing work and consistent guidance for repeatable tasks.

Final thought

ChatGPT’s team features are becoming more useful because they are moving closer to how real work happens. Teams need shared context, repeatable workflows, governed access to data, and clear ownership.

Projects, Shared Projects, Custom GPTs, and connectors each support part of that operating model. The strongest setup starts with the work itself: what needs to be organized, what needs to repeat, what data should be available, and who is responsible for keeping the system useful.

Treyworks helps teams make those decisions before the tools get rolled out. As an AI and web development partner, we help map use cases, governance needs, and practical AI workflows so your team can build faster, operate smarter, and measure what is working.

Need help deciding how ChatGPT should fit into your team’s workflow?

Schedule a consultation to design a setup your team can actually maintain.

Karl B GPT Case Study: An AI Advocacy Assistant for Music and Arts Education

August 17, 2026

Built for the teachers, parents, and advocates making the ask for music and arts programs Why we took this…

Read Article
What Does Human Review Mean When a Nonprofit Uses AI?

June 10, 2026

Turn human review into a clear nonprofit workflow with a named reviewer, five practical checks, and a simple record for improving AI-assisted work.

Read Article
What Information Should Never Go Into an AI Assistant?

June 6, 2026

As a default, do not put sensitive information, secrets, or unapproved public claims into a general AI assistant. For…

Read Article