Skip to content
eWiki The practical AI reference

Tutorials

How to Use ChatGPT Projects for Files, Instructions, and Recurring Work

White desk with stacked gray folders, a paper-filled tray, a blue pen, a ruler, and several dark strips.

A practical guide to using ChatGPT Projects for shared files, project instructions, focused chats, and recurring work.

When to Use ChatGPT Projects

ChatGPT Projects keep related chats, files, instructions, and sources together. They are designed for work that continues over time, produces multiple outputs, or relies on the same source material across conversations. A chat outside a project is better suited to self-contained work that does not require shared project context.

For example, a project could contain separate chats for research, drafting, review, and follow-up while making the same uploaded files and project instructions available to those chats. OpenAI also identifies recurring activities such as weekly research and content drafting as potential uses for projects.

As an editorial suggestion, create a project when you can identify material or guidance that several future chats should share. For a single question or isolated deliverable, consider starting with an ordinary chat and moving to a project if the work expands.

Dark desk with folders in an organizer, a stack of papers bearing faint line drawings, a black pen, ruler, and lamp.

Create a Project and Add Shared Context

To create a project, select New project in the ChatGPT sidebar, enter a name, and choose an icon and color. Once it exists, you can add reference material by uploading files or pasting text. Supported uploads described by OpenAI include PDFs, spreadsheets, documents, and images. Upload allowances depend on the user’s plan.

A useful setup approach is:

  1. Choose a short name that identifies the continuing body of work.
  2. Add files that should be available across multiple project chats.
  3. Put durable behavioral guidance in project instructions.
  4. Begin a new chat inside the project for the first specific outcome.

This is a suggested organization pattern, not a guaranteed way to improve results. Review uploaded material before relying on a response, especially when the project contains changing or time-sensitive information.

Person sorting white sheets into three gray desk trays beside a mug, lamp, notepad, and pen.

Write Project Instructions for Repeated Guidance

Project instructions apply across chats in the project. OpenAI’s current interface instructions direct users to the project’s more-options menu and Project settings to add them. These instructions apply within that project and take precedence over global custom instructions.

Consider using project instructions for guidance that should remain relevant to several outputs, such as the intended audience, preferred response structure, or an established terminology list. For example, you might suggest concise answers, request bullet points, or ask ChatGPT to identify questions that need clarification.

Keep request-specific directions in the individual chat. That separation is an editorial recommendation: project instructions can hold durable guidance, while each chat can state the requirements for its particular deliverable. Revisit the instructions when the project’s scope or audience changes.

Separate Recurring Work by Outcome

OpenAI recommends starting a separate chat for each distinct outcome. Each chat retains its own transcript, while chats in a ChatGPT project can use the same uploaded files, connected sources, and project instructions.

As a practical suggestion, name chats after their intended output—for example, “Q3 launch brief,” “Interview themes,” or “Final copy review.” This can make the project’s chat list easier to scan. If an existing eligible chat belongs with the work, ChatGPT supports dragging it onto the project or using Move to project from its menu. After it is moved, it receives the project’s instructions and file context.

When a response will be useful in later project chats, its message menu includes Add to project sources. OpenAI gives summaries, decision notes, drafts, and analyses as examples of material users may choose to save.

Keep Active and Finished Work Organized

The Projects interface provides several organizational controls. You can pin a project near the top of the sidebar, pin a frequently revisited chat, rename chats, search for past chats, and archive finished chats. Archived chats can later be restored from ChatGPT settings.

Pinning only changes an item’s sidebar position; OpenAI states that it does not add context or change ChatGPT’s access. A suggested routine is to pin current work, give each chat a short outcome-based name, and archive chats after their work is complete. Use search when you remember wording from a past conversation but not its title.

Conclusion

ChatGPT Projects are most appropriate when multiple chats need the same files, sources, or continuing instructions. The core structure is simple: keep reusable context at the project level, place one-off inputs in the relevant chat, and start a separate conversation for each distinct result.

My practical recommendation is to begin with one active, recurring task. Create a clearly named project, add only the references needed across its chats, write a short set of durable project instructions, and open one outcome-focused chat. As the work develops, save genuinely reusable responses as project sources and use pinning, descriptive names, search, and archiving to manage the conversation list.

Sources

  • [

Projects and chats | ChatGPT Learn ](https://learn.chatgpt.com/docs/projects)

Disclosure

This article was prepared with AI assistance.

Improve this guide

Report a problem or suggest the next question

Submissions enter a private editorial queue. They are reviewed and deduplicated before any public update or new page is created.