Projects, instructions, and memory

Home Tools Platforms ChatGPT Projects, instructions, and memory

Keep related sources together and distinguish deliberate project instructions from recalled context.

Conversation & research · Documented · Documentation checked 2026-10-04

ChatGPT Chat and Work; local Codex and the OpenAI API are separate surfaces.

Tools, usage, and controls vary by plan, client, region, rollout, and workspace policy.

The worked example is illustrative and has not been tested in a signed-in product account.

What it does and how it connects

Keep related sources together and distinguish deliberate project instructions from recalled context.

A project intentionally groups conversations, sources, and instructions. Memory carries useful context across work where enabled. These are different mechanisms: the project is a named working container, while remembered context may be incomplete or outdated. A web project does not automatically mount a folder on your computer.

Walkthrough

Use the product named in the access panel. The steps below combine documented controls with the editorial example.

  1. Create a project for one topic and add only the approved source files.
  2. Write short project instructions defining audience, terminology, and evidence standards.
  3. Check personalization settings for the memory behavior available to your account.
  4. Start a fresh conversation and ask which sources and instructions it is using.

Worked example

Create an AI literacy project containing your approved glossary and workshop outline. Ask for a new lesson in the same voice. If an old preference conflicts with the current outline, make the current project instructions explicit and ask the assistant to identify the conflict.

Using this project’s approved outline and glossary, draft a beginner lesson on AI agents. Before drafting, list the project sources you can actually access. Follow the current outline if recalled preferences conflict with it. Flag any missing reference instead of pretending to have read it.

Inspect and verify the result

Check references against files actually present. Ask a follow-up in a new chat to test whether essential instructions are still supplied. Put rules that must always apply in deliberate project instructions or checked-in documentation, not only in memory.

Limits and troubleshooting

Memory is a recall aid rather than a complete archive. Local Codex memory is separate from ChatGPT web memory. If expected context is missing, attach it explicitly; repeatedly asking the model to remember does not supply an unread file.

Data, permissions, and actions

Projects retain their attached context according to the product and workspace controls. Shared project access and connected-source permissions need separate review. Exclude passwords, personal records, and preferences you would not want reused.

Images and multimodal interaction

Use images as inputs or generated drafts and understand the different voice and video boundaries.

Explore Images and multimodal interaction

From evidence to a beginner lesson

Create a sourced workshop lesson and an editable handout from approved material.

Explore From evidence to a beginner lesson

A weekly evidence digest with review

Repeat a public-source investigation and deliver a private draft digest.

Explore A weekly evidence digest with review

Plugins, connectors, and skills

Separate packaged capabilities, reusable instructions, and the connections that let AI reach other software.

Explore Plugins, connectors, and skills

Sources and verification

Documentation checked 2026-10-04. Examples are illustrative unless described otherwise.

Projects and chats · OpenAI

Memories across product surfaces · OpenAI

Share this page