Custom GPTs and ChatGPT Projects: A Simple Guide for Small Business Owners
Picture this. It’s 7:40 a.m. You are drafting a client update from your phone before school drop-off. Yesterday’s notes are in Drive. Your tone guide is in another chat. You keep retyping the same instructions. Fifteen minutes later, you are still stitching pieces together. Sound familiar? Projects and custom GPTs fix this by remembering your instructions and files so you can jump straight to the work.
TLDR: Use a Project when you want a tidy workspace that remembers instructions and files across chats and across your team. Use a custom GPT when you want a reusable “mini app” you can open from your sidebar or the GPT Store. We prefer Projects for most small-business workflows because they are faster to set up and easier for teammates to use. You can also see how this compares to Google’s approach in our Gemini Gems guide.
Where Projects and custom GPTs fit in your workflow
Both options sit in the middle ground between one-off prompting and full automation. With either approach, you define the task, tone, and constraints once, then reuse them. Projects group multiple chats, shared files, and your instructions into one space. Custom GPTs package instructions and optional tools into a reusable “GPT” you can open from the GPTs list or share.
OpenAI’s help docs describe custom GPT creation via the GPT Builder, which has two tabs, Create and Configure, for building and polishing your GPT. Source: OpenAI Help, 2025. OpenAI’s Projects overview explains that Projects are smart workspaces that keep related chats, files, and custom instructions together so responses stay on-topic. Source: OpenAI Help, 2025.
Quick comparison: which to choose first
| Use this when… | Pick Projects | Pick a custom GPT |
|---|---|---|
| Multiple chats share the same context and files | ✔ Keeps instructions and files across chats | Possible, but each chat lives inside the GPT you open |
| Team needs shared access | ✔ Shared projects with shared files and instructions | Shareable, but the builder UI can feel heavier |
| Open from a sidebar as a “tool” | Good for ongoing work | ✔ Excellent. A custom GPT opens like a reusable app |
| Fastest to get started | ✔ Usually a few minutes for a solid setup | More steps and options during creation; clunky and confusing user interface that can intimidate casual users (and honestly, even pros like us!) |
How to set this up
Here is the short version. Projects: create a project, add instructions and files, start chatting. Custom GPTs: open the GPT Builder, write your instructions, and configure capabilities, knowledge, and actions.
Set up a ChatGPT Project (step by step)
- Create a Project. In ChatGPT, create a new Project and name it after the job, like “Monthly Client Updates” or “Proposal Drafting.” Projects are workspaces for related chats, reference files, and your custom instructions. OpenAI Help.
- Add project instructions. Paste a short “house style” and task definition. Keep it concrete: audience, tone, must-do, must-avoid.
- Upload reference files. Add samples, price sheets, or FAQs. As of September 3, 2025, OpenAI’s release notes list file limits per project: Free allows up to 5 files, Plus/Go/Edu allow up to 25, and Pro/Business/Enterprise allow up to 40. OpenAI Release Notes, 2025.
- Start your first chat inside the Project. Ask for a draft, a summary, or a checklist. Every new chat inherits your instructions and files.
- Invite teammates if needed. OpenAI’s announcement notes you can add members to collaborate in a shared project with shared instructions and files. OpenAI Blog, 2025.
Build a custom GPT (step by step)
- Open the GPT Builder. Go to chatgpt.com/gpts and click + Create, or open chatgpt.com/gpts/editor. OpenAI Help, 2025.
- Use the Create and Configure tabs. In Create, describe what you want. In Configure, refine instructions, upload files as knowledge, turn tools on or off, and set sharing. OpenAI Help, 2025.
- Decide on discovery. Keep it private, share with a link, or list it in the GPT Store. OpenAI introduced the GPT Store on January 10, 2024. OpenAI, 2024. For the basics of GPTs themselves, see OpenAI’s original announcement. OpenAI, 2023.
- Test with real inputs. Paste a real email or document. Edit your instructions until the response looks right.
Day-to-day view
With Projects, you open the same project every time you work on that client or task. Your instructions and files are already there, so each new chat starts on the rails. Teammates see the same context. It feels like a shared folder that also remembers how you want the work done.
With a custom GPT, you open the GPT by name from the GPTs menu and then start a chat. It is great for a reusable “assistant,” like a lead-reply helper or proofreading buddy. The builder UI is powerful but takes a few extra clicks to maintain compared with simply tweaking project instructions.
Results to expect
Plan 15 minutes to set up your first Project and 20 minutes for your first custom GPT. Simple math: if you save 12 minutes on three drafts per week, that is 36 minutes a week or about 31 hours a year back to your calendar. If a teammate also uses the same project, double it. Your numbers will vary, but it adds up fast. If you want to save even more time, you should explore AI workflow automation.
Seven practical instruction templates you can paste into a ChatGPT Project or Custom GPT today
Copy these into Project instructions or a new custom GPT and adjust for your tone and tools.
1) Lead reply assistant
What it does: Sends warm, on-brand first responses and asks for missing details. See a real-world angle in our real estate AI workflows guide for how responsiveness impacts results. :contentReference[oaicite:0]{index=0}
Starter instructions: “You write quick, friendly first replies. Confirm the need, ask up to three clarifying questions, and propose a 20 or 30 minute call. Never invent pricing. Offer two time slots from the availability text I paste.”
2) Proposal polish
What it does: Turns rough outlines into clear sections with next steps.
Starter instructions: “Rewrite in plain language. Keep headings. Flag any gaps with a short ‘Missing info’ list at the end.”
3) Policy checker
What it does: Skims contracts or policies for risky words and missing pieces.
Starter instructions: “Scan this document for deadlines, signatures, address mismatches, and words like wire, unpermitted, leak. Output a table: Item, Why it matters, Due by.”
4) Marketing copy coach
What it does: Produces caption options under 60 words in your voice.
Starter instructions: “Suggest captions for this image. Give three options. Tone: warm, direct, trustworthy. Avoid buzzwords. End with a short call to action.”
5) Meeting prep brief
What it does: Creates a one-pager with facts, questions, and an email draft.
Starter instructions: “Summarize key info from this meeting in five bullets, list three open questions, then write a follow-up email I can send.”
6) Client follow-up nudge
What it does: Drafts a polite check-in when someone has not replied in 3 to 5 days.
Starter instructions: “Draft a polite follow-up email in under 100 words. Restate the ask in one line. Offer one next step. No filler.”
7) Customer feedback scan
What it does: Analyzes comments and lists themes by sentiment.
Starter instructions: “Write a one-paragraph summary, bullet top themes tagged Positive, Neutral, or Negative, then give three action ideas.”
Common pitfalls and simple fixes
- Objection: “I tried a custom GPT and it feels heavy.” Workaround: Put your instructions and files into a Project first. You can still keep a custom GPT later for quick access, but the Project becomes your source of truth. (Or, use Gemini Gems as an alternative to custom ChatGPT).
- Instructions drift. If outputs wander, tighten the first 3 lines of instructions. Add 2 to 3 banned words that often creep in.
- Version confusion. Keep one Project per client or process. If you need variants, use a heading like “Alt tone B” inside the same instructions so the context remains shared.
Sharing and permissions
You can invite teammates to a shared Project so everyone works from the same instructions and files. For custom GPTs, you control sharing in the GPT Builder. Decide whether to keep it private, share by link, or list it in the GPT Store.
Practical limit to keep in mind: as of September 2025, OpenAI notes Project file uploads per tier are 5 on Free, 25 on Plus/Go/Edu, and 40 on Pro/Business/Enterprise. If you hit a limit, compress or split large files and keep only the most-used references in the Project. For more details, see the OpenAI Release Notes, 2025.
Alternatives to Custom GPTs: Gemini Gems and Automated Workflows
If you find ChatGPT isn’t your style, check out Google’s approach to solving the same problem in our Gemini Gems guide. Gemini Gems are a very easy to set up alternative to Custom GPTs; one of our friends swears by them, and we’ve even set up a few for our own use.
However, if you’re already exploring custom GPTs, you’re well on your way to automation. If you want drafts to appear without you lifting a finger, step up to AI workflow automation. Our automation primer explains how this works and why it saves real time. You can also browse real examples on our Solutions page; or for more of a practical view, see our guide for real estate teams. Contact us to talk through your use case. If you want practical AI tips in your inbox, join our mailing list or follow us on X and LinkedIn. If this guide helped, please share it with a friend!
Related Posts:
Troubleshooting and limitations
- Can’t see Projects? Check OpenAI’s release notes to confirm availability by plan and recent changes. As of September 3, 2025, Projects are available on the Free tier with file limits noted above. OpenAI Release Notes, 2025.
- Which feature should I use to “teach” ChatGPT globally? OpenAI’s Custom instructions apply to all new chats and are available across platforms. They are different from Project instructions, which apply within a specific Project. OpenAI, 2023, OpenAI Help, 2025.
- Builder confusion. In the GPT Builder, use the Create tab to describe what you want and the Configure tab to fine-tune instructions, files, tools, and sharing. OpenAI Help, 2025.
FAQ
What is the difference between Project instructions and Custom instructions?
Answer: Custom instructions apply to all new chats unless you change them. Project instructions apply only inside that Project and are shared with teammates if you invite them. OpenAI Help, OpenAI Help.
When should I pick a custom GPT instead of a Project?
Answer: Pick a custom GPT when you want a reusable “assistant” you can open from the GPTs menu or list in the GPT Store. Pick a Project when you want a space that holds instructions, files, and multiple chats together for an ongoing effort. OpenAI Help, OpenAI, 2024, OpenAI Help.
Answer: Yes. OpenAI says members can add files and instructions to a shared project together so every new chat starts with the latest information. OpenAI Blog, 2025.
Are there file limits in Projects?
Answer: Yes. As of September 3, 2025, OpenAI’s release notes list 5 files on Free, 25 files on Plus/Go/Edu, and 40 files on Pro/Business/Enterprise. OpenAI Release Notes, 2025.
Do I have to pick one?
Answer: No. Many teams use a Project for shared context and a custom GPT as a quick-launch assistant that follows the same instructions. Start with the simpler one, usually a Project, then add a custom GPT later if it helps.
Sources
- OpenAI Help Center (2025). Creating a GPT.
- OpenAI Help Center (2025). Using Projects in ChatGPT.
- OpenAI Help Center (2025). ChatGPT — Release Notes.
- OpenAI Blog (2025). More ways to work with your team and tools in ChatGPT.
- OpenAI Blog (2024). Introducing the GPT Store.
- OpenAI Blog (2023). Introducing GPTs.
- OpenAI (2023). Custom Instructions for ChatGPT and OpenAI Help Center (2025). ChatGPT Custom Instructions FAQ.
- OpenAI Academy (2025). Shared Projects and Custom GPTs.
Note: Product names and limits change. If a label looks different in your account, check OpenAI’s help center and release notes for the latest UI and rollout details.