How to Use ChatGPT Projects to Organise Your Work (2026 Guide)

Last updated: 13 November 2026

What are ChatGPT Projects?

Projects are ChatGPT’s way of organising ongoing work. Instead of scattering dozens of chats about one client, course or campaign across your history, you keep them together in a project. Every chat inside the project can draw on the same uploaded files and instructions, and on what was discussed in other chats in that project.

Think of a project as a smart folder with a briefing note. The files are your reference material. The instructions are the briefing that tells ChatGPT how to work. The chats are your individual working sessions.

Projects limits by plan

You can upload up to 10 files at once, and the number of projects is unlimited. Limits change from time to time, so check OpenAI’s help centre for the latest figures.

How to create a project, step by step

1. Click “New project” in the ChatGPT sidebar on web, desktop or mobile.
2. Name it clearly. Use names you will recognise in six months, such as “Client: Sharma Dental, SEO 2026” or “Book: chapter drafts”.
3. Choose the memory setting. Pick project-only memory if the project contains client or confidential context that should not influence other chats. This choice cannot be changed later.
4. Upload reference files. Add your brief, style guide, product information, past examples or data. Prioritise quality over quantity.
5. Add project instructions. Open the three-dot menu in the project and add instructions. Keep them short and specific.
6. Start a chat inside the project. Ask your first question. ChatGPT will use the files and instructions automatically.
7. Move existing chats in. Drag relevant older chats into the project to bring their context along.

Writing good project instructions

Project instructions apply only inside the project and take priority over your general custom instructions. Good instructions are short, concrete and testable. For example:

Review your instructions after the first few chats. If ChatGPT keeps making the same mistake, add a specific rule.

Choosing the right files to upload

With a limited number of files per project, every upload should earn its place. The most useful files are usually:

• A one-page brief that explains the goal, audience, constraints and deadlines.
• A style or brand guide with tone, banned words, formatting rules and examples.
• Core facts, such as product details, prices, services, team bios or policies.
• Good examples of past work you want ChatGPT to learn from.
• Data such as spreadsheets for analysis, kept tidy with clear column names.

Avoid uploading huge, mixed documents when only a small part matters. Extract the relevant sections into a cleaner file. Keep files current: when prices or details change, replace the old file instead of adding a new one alongside it, so ChatGPT is not confused by conflicting information.

Understanding project memory

Projects have their own memory behaviour, which matters for both quality and privacy.

A few rules to remember: the memory choice is made when you create the project and cannot be changed later; sharing a project switches it permanently to project-only memory; and you cannot view a list of what a project has “remembered”. If memory starts causing confusion, create a fresh project with clean files and instructions. For more, read AI memory explained.

Sharing projects with your team

Shared projects are useful for small teams, agencies and families planning together. The owner clicks Share and invites people by email or link, or by group on business plans. There are two access levels:

• Edit: can update instructions, manage files and invite others.
• Chat: can view and chat in the project but not invite people.

Members can start their own chats inside the project, branch existing chats and move their chats in. Real-time co-editing of a single chat is not supported. On shared projects, OpenAI only uses the data for model training if every contributor has the “Improve the model for everyone” setting turned on, but businesses should still check their own data policies before sharing sensitive information. See AI data privacy for business.

Project templates you can copy

Marketing campaign

Files: campaign brief, brand guide, product sheet, audience research. Instructions: “Audience: small business owners in India. Tone: practical and friendly. Every asset needs a clear call to action. Never invent statistics.”

Client account (agencies and freelancers)

Files: client brief, approved messaging, past reports, competitor notes. Instructions: “Write as the agency, not the client. Use the client’s terminology from the brief. Flag anything that needs client approval.” Use project-only memory.

Content production for SEO and AEO

Files: keyword and question list, style guide, internal link list, example articles. Instructions: “Answer-first introductions, question-based headings, tables where useful, 6 to 8 FAQs, suggest internal links from the list.” See answer-first content.

Job search

Files: CV, LinkedIn profile text, target job descriptions. Instructions: “Tailor applications to each job description. Never exaggerate experience.” See the best AI tools for job search.

Study or course

Files: syllabus, lecture notes, past papers. Instructions: “Explain step by step. Quiz me after each topic. Point out when my answer is wrong and why.”

Home or personal planning

Files: budget sheet, floor plan, quotes from contractors. Instructions: “Keep suggestions within the budget in the sheet. Show costs in rupees.”

Getting better answers inside a project

Even with good files and instructions, how you ask still matters. A few habits make a noticeable difference:

• Point to sources. “Using the pricing section of services.pdf, draft a reply to this enquiry” works better than hoping ChatGPT finds the right file.
• Set the format. Say whether you want a table, a bullet list, an email or a writing block.
• Give the audience. “For a first-time customer who is nervous about cost” changes the tone and content.
• Ask for uncertainty. “If the files do not cover this, say so” reduces made-up answers.
• Iterate in the same chat for one task, and start a new chat for a new task, so each conversation stays focused.
• Save winning prompts to a file in the project so teammates can reuse them.

Using projects on mobile and desktop

Projects sync across ChatGPT on the web, the desktop app and the iOS and Android apps. Mobile is handy for quick questions, voice conversations and capturing ideas with project context, such as dictating meeting notes into a client project on the way back to the office. For uploading files, writing longer instructions and reviewing detailed outputs, the web or desktop app is more comfortable.

Projects vs custom GPTs vs custom instructions vs memory

ChatGPT has several ways to give it context. They overlap, which can be confusing. Here is when to use each:

A useful rule of thumb: if the context is about you, put it in custom instructions or memory. If it is about a piece of work, put it in a project. If it is a tool you want to reuse or share with many people, build a custom GPT.

Working with data inside projects

Projects are useful for recurring analysis. Upload a clean spreadsheet, such as monthly sales or website traffic exports, and ask ChatGPT to analyse it, build charts or spot trends. Because the file stays in the project, you can come back next month, upload the new data and ask for a comparison. Tips:

• Use clear column names and one header row.
• Remove personal data you do not need.
• Ask ChatGPT to explain its method and show its calculations so you can check them.
• Verify important numbers in your own spreadsheet before acting on them.

For website analytics specifically, our guide to tracking AI traffic in GA4 shows what to export.

Common mistakes with ChatGPT Projects

• One giant project for everything. Context gets muddled. Create separate projects per client, campaign or subject.
• Too many overlapping files. Old and new versions of the same document confuse ChatGPT. Replace, do not stack.
• Long, vague instructions. Short, specific rules work better than paragraphs of general advice.
• Using default memory for client work. Choose project-only memory at creation for anything confidential.
• Forgetting that deletion is permanent. Download important outputs before deleting a project.
• Trusting outputs without checking. Projects improve context, not accuracy guarantees. Always fact-check.
• Sharing without thinking. Shared members can see files and chats. Remove sensitive material first.

Projects and the wider ChatGPT toolkit

Projects work well alongside other ChatGPT features:

• Writing blocks for drafting and editing inside project chats, now that Canvas has been retired. See ChatGPT Canvas is gone.
• Deep research for in-depth reports you then store and build on in the project. See how to use deep research.
• Voice mode for talking through ideas on the go, with project context. See ChatGPT voice mode.
• ChatGPT Work in the desktop app for longer agent tasks, where available on your plan.

An example: running a content calendar in one project

Here is an illustrative example of how a small marketing team might run a monthly content calendar inside one project.

1. Set up: A project called “Content: Q1 2027” with project-only memory. Files: brand guide, list of target questions, product sheet, internal link list. Instructions: tone, length, formatting and SEO rules.
2. Planning chat: “Using the question list, propose 12 blog topics for January to March, grouped by theme, with a suggested title and target question for each.”
3. Brief chats: One chat per article to create an outline and brief for the writer.
4. Draft chats: Writers draft in separate chats, section by section, using writing blocks.
5. Social chat: “For each published article, write a LinkedIn post and two Instagram captions.”
6. Review chat: At month end, upload performance data and ask for a summary of what worked.

Because everything sits in one project, each new chat already knows the brand, audience and rules, which saves repeating the same context dozens of times a month.

Project ideas by role

Troubleshooting

Keeping projects tidy over time

• Archive finished projects. Export important outputs, then delete or leave them clearly labelled as complete.
• Review instructions quarterly. Businesses change; update tone, offers and rules.
• Rename chats. Give each chat a descriptive title so you can find past work quickly.
• Keep a change log file. A short document listing key decisions helps both people and ChatGPT stay consistent.
• Start fresh when needed. If a project becomes cluttered or confused, create a new one with clean files.

Privacy and security tips

Projects make it tempting to upload everything about a client or your business. Be selective:

• Do not upload passwords, ID numbers, bank details or health records.
• Remove customer personal data unless you have a clear reason and permission.
• Use business plans with appropriate data controls for company work.
• Follow your organisation’s AI policy; our AI usage policy template can help you create one.

Who benefits most from Projects

Projects help anyone who returns to the same work repeatedly: freelancers juggling clients, marketers managing campaigns, students preparing for exams, founders writing investor updates and families planning events or renovations. They help least for one-off questions, where a normal chat is quicker. If you notice yourself pasting the same background information into new chats again and again, that is the signal to create a project.

Our take

Projects are one of the most underrated ChatGPT features. Ten minutes setting up a focused project with a clear brief, good reference files and specific instructions can save hours of repeated explanations and produce noticeably more consistent results. Start with one project for your most frequent type of work, refine the instructions after a week, and expand from there.

Frequently asked questions