Custom GPT for a Real Estate Team: Who Owns It Now?

By Emily Terrell — Top Coach and Speaker at Tom Ferry International. Licensed since 2016. Closing 70+ deals/year while coaching agents nationwide.

A custom GPT for a real estate team needs one named owner, use-only access for agents, and master files kept outside the platform. The timing matters: OpenAI has scheduled custom GPTs to retire December 11, 2026, and migration doesn’t carry over sharing. This guide covers the ownership model, the platform decision, and the migration checklist.

Key Takeaways

• Every shared AI assistant needs one named owner, and that owner should be the team leader or ops lead, not whichever agent built it first.
• Agents get permission to use the assistant, and only the owner and your compliance lead get permission to edit it.
• OpenAI has scheduled custom GPTs to retire on December 11, 2026, and migrating a GPT doesn’t carry over who it was shared with.
• Keep the master instructions and knowledge files in a folder your team controls, so a platform change costs an afternoon instead of a rebuild.
• Review the assistant every quarter alongside your AI policy, because one unreviewed edit changes the output for every agent at once.

What is a shared custom GPT for a real estate team?

A shared custom GPT is one AI assistant that a team leader builds once and gives every agent access to. The whole team then works from the same instructions, voice, and compliance rules. In ChatGPT, that’s been a custom GPT, which is now being replaced by plugins. In Claude, it’s a Project shared on a Team or Enterprise plan. For one agent, an assistant like this saves time. For a team, it’s infrastructure, and infrastructure needs an owner.

Why does this matter for real estate teams right now?

Two things changed this year.

The first is that agents use AI but don’t trust what it produces. According to RPR’s February 2026 survey of 225 NAR members, 82% of agents use AI in their business. In the same survey, 63% named accuracy of outputs as their top concern, and 49% named compliance or legal issues (RPR). According to NAR’s 2025 Technology Survey (September 2025), only 17% of agents reported a significantly positive impact from AI, while 46% said it had no noticeable impact (NAR).

“The real opportunity now is confidence.”
— Reggie Nicolay, SVP of Marketing and Training, Realtors Property Resource

Here’s the thing nobody wants to tell you: a team assistant can close that confidence gap, and it can also scale a mistake. With shared instructions, one compliance rule gets applied by forty agents. One bad edit gets applied by forty agents too.

The second change is the platform itself. OpenAI no longer lets personal accounts, including Free, Go, Plus, and Pro, create new GPTs (OpenAI). It has also scheduled custom GPTs to retire on December 11, 2026, with workflows moving to plugins. Migration doesn’t carry over a GPT’s sharing settings, and permission to use someone else’s GPT doesn’t give you permission to migrate it (OpenAI). So if an agent built the GPT your whole team uses, that agent controls the migration unless you run an Enterprise workspace.

“Whoever built your team’s GPT owns it until you decide otherwise. Write the files down where you control them, and a platform change becomes an afternoon of migration instead of a rebuild.”
— Emily Terrell, Tom Ferry Coach

The ownership model: four decisions every team leader makes

Building the assistant takes twenty minutes. The decisions below determine whether it still works in a year.

Who owns the team’s AI assistant?

The team leader or operations lead owns it, never the agent who happened to build it first. Ownership decides who can migrate the assistant, who can delete it, and where it goes when someone leaves.

The platforms handle this differently. In a managed ChatGPT Enterprise workspace, if a GPT’s owner is deactivated, ownership moves to a workspace owner and the GPT is marked unassigned so it can be reviewed (OpenAI). A GPT built in an agent’s personal account has no safety net like that. It stays with the agent. On Claude Team and Enterprise plans, project creators manage sharing and can change permissions or remove access (Claude Help Center).

If the assistant your team depends on lives in someone else’s account, rebuild it under the owner’s account in a workspace the team controls.

Who can edit the assistant, and who can only use it?

Agents use it. One or two people edit it.

ChatGPT has three permission levels: Can chat, Can view settings, and Can edit (OpenAI). Claude has two. Can view lets members see the project’s instructions and knowledge and chat inside it without changing anything. Can edit lets members change the instructions and knowledge and add or remove members (Claude Help Center).

Here’s why this matters. Say an agent edits the instructions to fix the output on their own listing. That edit now changes the output for every agent on the team, and nobody reviewed it. Route every edit through the owner, and keep a one-line changelog for each change: the date, what changed, and why.

Where do the master files live?

In a folder your team controls, outside the AI platform. The platform only holds a copy.

The files are the same four that make any assistant work: voice, process, constraints, and source of truth. I break each one down in why most custom AI assistants fail. What changes for a team is where those files live.

This is your insurance against retirement. OpenAI’s migration only uses GPT’s latest published version. Drafts and unpublished edits don’t transfer, and neither do custom actions (OpenAI). If the only copy of your instructions lives inside the GPT, you’re depending on a migration tool working correctly. If the master copy lives in a document, you can paste it into whatever comes next in one sitting.

When does the assistant get reviewed?

Every quarter, on the same schedule as your AI policy. Check four things:

• The instructions still match your written AI policy.
• The knowledge files are current, including market stats, scripts, and disclosure language.
• The access list matches your roster.
• Three familiar prompts still produce output you’d publish.

Offboarding happens outside that quarterly schedule. Remove an agent’s access the same day they leave.

Should a team build a custom GPT or a Claude Project in 2026?

Don’t start a new custom GPT. Personal plans can’t create one anymore, and anything built in a Business or Enterprise workspace is on a retirement schedule.

If your team already runs on ChatGPT Business or Enterprise, build your instructions as a skill inside a plugin. Under the migration plan, a GPT’s instructions become a skill and its knowledge files become the plugin’s reference files. In Enterprise workspaces, sharing a plugin with people or groups requires its own permission (OpenAI). Confirm with your admin who has that permission before you build.

If your team is choosing a platform now, a shared Claude Project on a Team plan is the simpler starting point. Free accounts can build up to five projects for testing, but sharing requires a Team or Enterprise plan (Claude Help Center). Give agents Can view access and keep Can edit for the owner. For what goes inside each project, see what a Claude Project is and how agents use one.

Either way, the files are what you’re actually building. The platform holds them.

What should a team do before December 11, 2026?

If your team uses custom GPTs today, work through this list in order:

1. List every GPT the team relies on and who created it. The creator controls migration unless you’re on Enterprise.
2. Copy the instructions and knowledge files into your team folder now. Do this before you migrate anything.
3. Finish edits and publish. Drafts don’t migrate. Publishing doesn’t require public sharing.
4. Migrate once the option appears for your account. Enterprise workspaces should note that new GPT creation is planned to end October 26, 2026.
5. Test the replacement with three familiar prompts and one hard case before anyone switches.
6. Re-share it. A migrated plugin starts private, so no one on your team can see it until you share it again.
7. Rebuild any custom actions separately. They don’t transfer.
8. Give agents a switch date and tell them where the replacement lives.

OpenAI’s dates are subject to change, so check the notice in your own workspace before you plan around them.

How I use this in my own business

I run my business on separate Claude Projects, one for each recurring job: CMA reports, listing presentations, buyer showing packets, and the blog you’re reading now. None of the master instructions live only inside Claude. Each one is a document in my operations folder, sitting next to my transaction checklist, and the project holds a copy.

When an instruction changes, the document changes first, with a dated line saying what changed and why. My transaction coordinator uses the projects. She doesn’t rewrite the rules inside them. I run my closing checklist the same way: one owner, and everyone else follows the process. That split is part of how 70+ transactions a year run on roughly five hours of active management a week.

Common mistakes

1. Letting the builder be the owner. The agent who was curious enough to build it first shouldn’t control the team’s infrastructure after they move on.
2. Giving every agent edit access. Forty editors means forty versions of your compliance rules.
3. Keeping the only copy inside the platform. If the platform is the only place your instructions exist, a retirement notice turns into a rebuild.
4. Starting a new custom GPT in late 2026. You’d be building on a product with a published end date.
5. Forgetting offboarding. A departed agent who still has edit access to the assistant your team uses is a compliance gap with a login.

This is general information, not legal advice. Consult your broker or attorney on your state’s advertising and privacy rules.

Frequently Asked Questions

Can I still create a custom GPT for my real estate team?

Not on a personal plan. OpenAI ended new GPT creation for personal accounts, including Free, Go, Plus, and Pro. In ChatGPT Business and Enterprise workspaces, creation depends on workspace settings, and custom GPTs are scheduled to retire on December 11, 2026. A team starting today should build its instructions as a plugin skill in ChatGPT or as a shared Claude Project instead.

When are custom GPTs being retired?

OpenAI’s standard retirement date for custom GPTs is December 11, 2026. Enterprise workspaces with an approved deferral get until February 11, 2027. Existing GPTs keep working until then, and past conversations stay accessible afterward. After retirement, the GPTs and their pages become inaccessible, so migrate and test the replacement plugin before that date rather than after.

What happens to my team’s GPT when an agent leaves?

That’s decided by where the GPT was built. In a managed ChatGPT Enterprise workspace, a deactivated owner’s GPT transfers to a workspace owner and is marked unassigned for review. A GPT built in an agent’s personal account stays in that account. Build team assistants in a workspace the team controls, and keep the master files outside the platform.

Should agents be able to edit the team’s AI assistant?

No. Agents should use the assistant, not change it. ChatGPT offers Can chat, Can view settings, and Can edit permissions; Claude Team and Enterprise offer Can view and Can edit. Give agents chat or view access, and limit edit access to the named owner and your compliance lead. One unreviewed edit changes the output for every agent on the team.

Is a Claude Project better than a custom GPT for a real estate team?

For a team starting from scratch in late 2026, a Claude Project is the safer starting point, because custom GPTs are scheduled to retire and a new GPT can’t be created on personal plans. Sharing a Claude Project requires a Team or Enterprise plan. If your team already runs on ChatGPT Business, build in plugins instead. The files you write work in either.

Can I share a Claude Project with my agents on a free plan?

No. Free Claude accounts can create up to five Projects, but sharing a Project with other people requires a Team or Enterprise plan. On those plans, the creator can share by email, in bulk, or with the whole organization, and owners can turn sharing off. A free account works for building and testing before you buy seats.

What should a team AI assistant never contain?

Client names, contact records, financial details, loan documents, and anything tied to an identifiable transaction. A shared assistant multiplies exposure, because agents with view access can see its knowledge files. Upload templates, scripts, redacted examples, and compliance rules instead, and check your brokerage’s AI policy before adding anything you’re unsure about. This is general information, not legal advice.

Bring this to your team or event

Emily Terrell speaks at brokerage events, real estate conferences, and team trainings on AI, systems, and social media — the exact playbook in this post, delivered live to your audience. As a Top Coach and Speaker at Tom Ferry International and an active agent closing 70+ transactions a year, Emily speaks from the stage about what’s working right now, not theory. Recent stages include NAHREP and eXp Con.

Book Emily to speak at your next event:
Email: eterrell@yourcoach.com
Phone: (210) 400-9191
Web: coachemilyterrell.com/keynote

For real estate agents who want to implement this: Get the weekly real estate prompt library at weeklyrealestateprompts.com or follow @coachemilyterrell on Instagram for daily systems and AI breakdowns.