The Marketer's Guide to ChatGPT Work

OpenAI didn't just update ChatGPT this summer. It split it into three modes. If you've opened ChatGPT recently and noticed a new toggle at the top of the screen, this guide is for you. Here's what ChatGPT Work is, how it's different from the chat window you already know, and how to actually put it to work on marketing tasks instead of one-line prompts.

ChatGPT Work at a glance

The 5 best resources on ChatGPT Work

OpenAI's own documentation is the system of record here, since this shipped recently and the details are still settling. We've added two independent, hands-on takes for the parts OpenAI's docs don't cover: what it actually feels like to use.

1. ChatGPT is now a partner for your most ambitious work, OpenAI. The official launch announcement: what Work is, the rollout schedule by plan, and named use cases across sales, marketing, finance, ops, and engineering. Start here.
2. ChatGPT Work and Codex, OpenAI Help Center. The clearest side-by-side of what Chat, Work, and Codex are each built for, including platform availability and how usage is metered.
3. Pricing, OpenAI. The live, official pricing page for every plan. The fastest way to confirm current numbers, since OpenAI updates plans often.
4. I gave ChatGPT's new Work mode my most annoying life-admin tasks, and it handled them like a pro, TechRadar. A hands-on outsider's test of Work across five real projects, with an honest note on where it still needs a nudge.
5. How to use ChatGPT: A beginner's guide to getting started, Zapier. The friendliest starting point if you or someone on your team hasn't touched ChatGPT's newer modes yet.

What is ChatGPT Work?

ChatGPT Work is OpenAI's agentic mode for longer, multi-step projects. Where the ChatGPT you already know is built to answer a question, Work is built to finish a job: it can take action across your connected apps and files, stay on a project for hours if needed, and turn a goal into completed work.

In practice, that means research, cross-referencing information across multiple sources, and producing an actual deliverable at the end (a document, a spreadsheet, a slide deck, a report, a simple website) instead of a paragraph of suggestions you still have to go build yourself. You turn it on from a toggle at the top of the ChatGPT interface, connect the apps it needs (Gmail, your calendar, Drive, spreadsheets), and give it a brief the way you'd brief a new hire.

Chat vs. Work vs. Codex

All three now live in the same ChatGPT app, which is exactly why people mix them up. Here's the actual division of labor.

For most marketing work, that puts you between the first two: quick copy tweaks and brainstorming stay in Chat, while a real campaign brief, a competitor scan, or a monthly report is a Work job. Codex only enters the picture if the project touches your actual landing-page code.

Pricing and access

Work doesn't carry its own separate fee. It's bundled into the plan you're already on, and it draws down your plan's usage based on how much work a task actually takes. Here's the full lineup, verified against OpenAI's own pricing page and help center in August 2026.

Prices exclude applicable tax and are subject to change at OpenAI's discretion. "Usage" isn't a fixed message count: a sprawling research-and-build task uses more of it than a quick chat, which is exactly why it's worth matching the plan to how much Work-mode work you actually plan to run. Not sure what "Sol" versus "Terra" versus "Luna" means for output quality? We broke down OpenAI's model-tier naming in Sol, Terra, Luna: which one?

How to actually use it

Turning Work on is the easy part. Getting a result worth keeping depends on how you brief it. Treat it like handing a project to a new hire, not like typing a search query.

Marketing use cases worth stealing

These come straight from OpenAI's own launch examples, not a hypothetical:

• Campaign briefs from raw research. Turn customer research into a campaign brief, then produce assets adapted for different markets.
• Lead review, automated. Have Work review and summarize leads across your CRM, email, and other connected tools.
• Feedback synthesis on a schedule. Set up scheduled monitoring that turns incoming customer feedback into a prioritized list of ideas, without you re-prompting it every time.

Bonus tips for marketers

1. Brief it like you would a new hire. Audience, inputs, constraints, and what "good" looks like. Every time.
2. Start with one recurring report. Pick your weekly or monthly one and hand it to Work first; it's the easiest place to see the time you get back.
3. Wire up your CRM and shared drive. Work's usefulness scales directly with what it's actually connected to.
4. Split by mode, not habit. Chat for a quick draft, Work for the multi-step brief with a real deliverable, Codex for anything that touches actual code.
5. Watch the usage meter. Work draws down usage by effort, not message count, so a sprawling research task can burn through your plan faster than a quick chat would.
6. Keep a human review step non-negotiable. Especially on anything that reaches a customer or prospect.
7. Use scheduled tasks for what you'd otherwise forget. Competitor monitoring, feedback synthesis, a weekly digest that builds itself.
8. Loop your team in on Business. The shared workspace, connectors, and admin controls are built for exactly this kind of handoff.
9. Learn the slash commands. /status and /project save more back-and-forth than most people expect. See the full command list.

🐾 Tiny bonus: ChatGPT Has a Pet

There’s also a much less serious feature hiding in the desktop experience: a little animated pet can follow what your agent is doing. It works as a lightweight status indicator while Work or Codex runs, so you can see whether your agent is working, waiting for you, finished, or needs attention without constantly staring at the conversation. You can even customize it. Necessary? Absolutely not. Fun? Yes.

Go deeper

Frequently asked questions