How to use ChatGPT Projects for recurring work
Short answer: A ChatGPT Project is a workspace that keeps related chats, uploaded files and project-specific instructions in one place, so every new chat inside it starts with the same context. To use one, create a project from the sidebar, add instructions and reference files, then start your chats inside it. It suits recurring work such as monthly reports.
Projects solve one specific problem: re-explaining the same background every time you open a new chat. If you want ChatGPT to remember things about you across every chat, that is a different feature, covered in our guide to custom instructions and memory. Projects are per task; custom instructions are account-wide.
Below we build one Project end to end for a recurring job, with the actual instructions, files and prompts, so you can copy the setup.
What goes inside a ChatGPT Project?
A Project holds three things: instructions that apply only inside it, files ChatGPT can refer to, and the chats you start there. Depending on your memory settings, it can also draw on earlier chats in the same project.
| Feature | Applies to | Best for |
|---|---|---|
| Custom instructions | Every chat on your account | How you like answers: language, length, your role |
| Memory | Across chats, account-wide | Facts about you that rarely change |
| Project | Only chats inside that project | One ongoing job with its own files and rules |
| Custom GPT | Anyone you share it with | A reusable tool for a team or the public |
As of November 2026, Projects are available on free and paid plans, but file limits and sharing options differ by plan. Check OpenAI's help centre for current numbers rather than relying on any blog post.
Worked example: a monthly social report for a Janakpuri cafe
Say you handle Instagram for a neighbourhood cafe in Janakpuri. Every month you export the numbers, write a one-page report for the owner and suggest next month's posts. Without a Project, you paste the cafe's background, the report format and last month's results every time. With one, you set it up once.
- In the ChatGPT sidebar, create a new project and name it “Cafe monthly report”.
- If you are offered a project-only memory option, choose it, so cafe details stay inside this project. Some memory options can only be set when the project is created, so decide up front.
- Open the project settings and add the instructions in the next section.
- Upload the reference files listed below.
- Each month, start a new chat inside the project and upload that month's data there.
The project instructions to copy and adapt
Write instructions like a brief to a new junior: who the client is, what the output looks like and what never to do.
“You are helping prepare a monthly Instagram report for a neighbourhood cafe in Janakpuri, West Delhi. The owner is not a marketer: use plain English, short sentences and no jargon. Every report has five parts: 1) headline numbers vs last month, 2) top three posts and why they worked, 3) what underperformed, 4) three ideas for next month tied to local dates and festivals, 5) one question for the owner. Use only numbers from the files or data I upload. If a number is missing, say so; never estimate. Show percentage changes with the calculation. Keep the report under 400 words.”
“Use only numbers from the files or data I upload” does the most work. Without it, a model may fill gaps with plausible-looking numbers, which is the fastest way to lose a client's trust.
Two more lines worth adding once you have run the project for a month or two: the mistakes it made last time (“do not calculate engagement rate on followers”) and the owner's preferences (“she wants the WhatsApp summary first”). Instructions are a living document; treat each correction you make in chat as a candidate rule.
Which files to add to the Project
Put stable reference material in the project's files, and upload each month's data inside that month's chat.
- A one-page cafe profile: menu highlights, price range, timings, regulars' favourite items and tone of voice.
- A blank report template with the five headings.
- A short KPI glossary: reach (unique accounts that saw a post), engagement rate (interactions divided by reach), saves and profile visits, with how you calculate each.
- Last quarter's reports, so it can compare trends.
A filled example of the cafe profile, short enough to read in a minute: “Neighbourhood cafe near Janakpuri District Centre. Open 9 AM to 11 PM. Known for cold coffee, cheese maggi and paneer tikka sandwiches; average bill roughly ₹350–500 for two. Regulars are college students on weekdays and families on weekends. Tone: friendly, light Hinglish, no slang the owner wouldn't say. Never promise discounts the owner has not approved.”
Each month, export insights from Meta Business Suite or Instagram's professional dashboard as a CSV or screenshots and upload them into that month's chat. Keeping monthly data out of the project files stops old numbers getting mixed with new ones, a common mistake when people dump everything in one place.
The prompts used each month
- “Here is November's Instagram data. Before writing anything, list the numbers you can see in a table, so I can confirm you read them correctly.”
- “Now compare with October's report in the project files. Show each percentage change with the working.”
- “Write the report using the template. Base next month's ideas on the cafe profile and December dates in Delhi such as Christmas and New Year's Eve.”
- “Rewrite the summary paragraph for WhatsApp, under 60 words, friendly tone.”
Then critique the first output like a reviewer. The weak spots are predictable: an engagement rate calculated on followers instead of reach, a top post chosen by likes when saves matter more for a cafe, and festival ideas that ignore what the cafe actually serves. Fix it in the chat once; if the same error comes back next month, add a line to the instructions.
ChatGPT Projects vs custom GPTs: which should you use?
Use a Project for your own ongoing work, where files and chats keep changing. Use a custom GPT when you want a fixed tool that other people can use, such as a caption writer for your whole team.
- Projects keep a history of chats and let you add files as you go.
- Custom GPTs are built once and shared; each user's chats stay separate.
- Claude has a similar Projects feature with its own knowledge files and instructions; see how to use Claude AI if your team prefers Claude.
If you are still getting comfortable with chats, uploads and regenerating answers, start with ChatGPT for beginners before setting up Projects.
Mistakes to avoid with ChatGPT Projects
- One giant project for everything. Make one per client or recurring job, which is also the easiest way to organise ChatGPT chats.
- Uploading confidential client data without permission. Check your plan's data settings and the client's expectations first.
- Vague instructions like “be professional”. Specific rules on word limits, sections and banned behaviour work far better.
- Never pruning files. Remove outdated templates so ChatGPT stops following old rules.
- Using a Project for a one-off question. If you will not come back to the task, a normal chat is simpler.
A Project earns its setup time when the job repeats at least monthly and follows the same format: client reports, weekly sales summaries, a newsletter, a course you are studying. Twenty minutes of setup then saves several minutes in every chat that follows, and the quality stays consistent because the rules no longer depend on what you remembered to paste.
The prompt patterns behind good project instructions are covered in our pillar guide on writing AI prompts, and ChatGPT workflows like this one are a core part of the Applied AI Professional course at Spettro School.
Frequently asked questions
Are ChatGPT Projects free?
As of November 2026, Projects are available on the free plan as well as paid plans, with lower file limits on free. Check OpenAI's help centre for current limits.
Does ChatGPT remember chats inside a Project?
It can refer to other chats in the same project when memory settings allow, and project-only memory keeps that context inside the project. Project instructions and files always apply to chats in it.
What is the difference between ChatGPT Projects and custom GPTs?
A Project is a workspace for your own ongoing task, with its own chats and files. A custom GPT is a reusable tool you can share with others.
Can I move an existing chat into a Project?
Yes. You can usually drag a chat into a project in the sidebar or move it from the chat's menu. It then sits alongside the project's other chats.
How do I organise my ChatGPT chats?
Group recurring work into Projects, one per client or task, and archive or delete one-off chats. Clear project names make search much faster.
Keep reading in AI tools
- How to Write AI Prompts That Work First Time Start here
- Claude Artifacts Tutorial: Build Two Working Tools Without Coding
- How to Use ChatGPT in Hindi and Hinglish, and Get Natural Results
- How to Summarise PDFs and YouTube Videos With AI Accurately
- AI voice generator for Hindi: scripts, fixes and consent rules
Related: Applied AI Professional · ChatGPT Custom Instructions and Memory: Set Them Up Once · All guides in AI tools