I would like to request the ability to create and customize multiple profiles within a single ChatGPT account and Pro subscription. Users should be able to name and organize these profiles according to their own needs.
Many people use ChatGPT across several parts of their lives. A single account is convenient, but keeping those contexts distinct requires more than organizing conversations into projects. Separate paid accounts also mean maintaining multiple subscriptions to keep comparable access to Pro features.
Each profile should have its own:
- Memory and contextual knowledge
- Chat history and search results
- Custom instructions
- Files and projects
- Connected apps and integrations, with permissions scoped to that profile
- Personalization settings, such as tone and language preferences
Profiles should be isolated by default. Information from one profile should not appear in another profile’s conversations, memory, search, or connected app access unless the user explicitly chooses to share it.
A visible profile switcher, similar to browser profiles, would make the active context clear.
For example, a user could create Personal and Work profiles. Switching to Work would show only that profile’s chats, files, projects, and integrations, reducing accidental exposure of personal conversations during screen sharing. Personal and Work would be just one example of how users could organize their profiles.
ChatGPT Projects with project-only memory help by keeping conversational context within a project. Project instructions and files are useful too. But Projects do not replace the complete profile experience described above. The requested separation would apply across all chats, projects, integrations, and personalization within the active profile, with profile memory available across that profile’s projects.
All profiles would remain under one Pro subscription for the same user, with plan usage limits shared across profiles. Users should not need additional Pro subscriptions simply to keep different contexts separate.
Could OpenAI consider adding this capability?