Migrate from claude to chatgpt.
5 hands-on-verified translation patterns — what carries over and what to watch for. Cited to the Feature Parity Map; the audit tells you whether the move is worth it.
Both products use Projects as named workspaces. Recreate each ChatGPT Project in Claude: create a new Project, copy over the project instructions, and upload reference files to the Project knowledge base. Conversations within a Claude Project inherit instructions and can reference knowledge files, mirroring ChatGPT Project behavior. The reverse migration follows the same steps in the opposite direction.
- Warning: ChatGPT Projects are available on Free plans; Claude Projects require Pro or above.
- Warning: Claude Projects have dedicated per-project memory with separate memory spaces; ChatGPT Project memory scoping may behave differently.
- Warning: Shared Projects require Team/Enterprise on Claude; ChatGPT sharing is available more broadly including on Free, Plus, Pro, and Go plans.
Both products offer persistent memory across conversations. Enable Claude's memory via Settings → Capabilities. Export or review ChatGPT memories from Settings → Personalization → Manage Memory, then manually recreate critical memories in Claude by mentioning them in early conversations or explicitly asking Claude to remember key facts. Both products allow viewing, editing, and deleting individual memories.
- Warning: No automated memory import between ChatGPT and Claude — memories must be manually recreated by sharing key facts in conversations.
- Warning: Claude memory is scoped per-project with separate memory spaces; ChatGPT memory can span projects.
- Warning: ChatGPT Memory is not available in custom GPTs; Claude memory works across all conversations including within Projects.
Recreate Custom GPTs as Claude Skills. For each GPT: create a Skill folder with a SKILL.md containing the GPT's instructions, add knowledge files as resources, and install the Skill. Claude auto-discovers and loads relevant Skills based on the task context. For simpler GPTs that are primarily instructions + knowledge, a Claude Project with custom instructions and uploaded knowledge files is an alternative approach requiring no development setup.
- Warning: GPT custom Actions (API integrations) have no direct Skill equivalent — these require building a separate MCP server or tool integration.
- Warning: GPTs are shareable via link and discoverable in the GPT Store; Claude Skills use the anthropics/skills marketplace or direct installation, a different discovery model.
- Warning: Skills auto-invoke based on task context; GPTs are explicitly selected by the user before starting a conversation.
- Warning: Claude Skills require Pro, Max, Team, or Enterprise; GPTs require Plus, Team, or Enterprise.
ChatGPT's two custom-instruction fields map to two Claude features: put 'what to know about you' content into Claude Preferences (Settings → Profile → Preferences) and put 'how to respond' content into a Claude Custom Style (Settings → Styles → Create). Claude applies Preferences globally and the active Style to all conversations, mirroring ChatGPT's global custom instructions.
- Warning: Claude splits ChatGPT's single feature into Preferences (facts) + Styles (tone/format); users must set up both to fully replicate the experience.
- Warning: Very long or complex custom instructions are better captured as Project-level instructions in Claude rather than global Preferences.
- Warning: ChatGPT custom instructions have a 1,500-character limit per field; Claude Preferences and Style descriptions have their own separate limits.
The 'How would you like ChatGPT to respond?' field in ChatGPT custom instructions maps to a Claude Custom Style. Extract tone, format, length, and structure cues from the ChatGPT field and create a matching Claude Style (Settings → Styles → Create). Claude applies the active Style to every conversation, mirroring how ChatGPT applies custom instructions globally. Moving from Claude to ChatGPT: distill the active Style's instructions into ChatGPT's 'How would you like ChatGPT to respond?' field (1,500-character limit).
- Warning: ChatGPT combines factual context ('What to know about you') and response style ('How to respond') in one feature; Claude separates these into Preferences (facts) and Styles (tone/format). This pattern covers only the style half — see pattern:chatgpt__claude/persistent-user-preferences for the facts half.
- Warning: Claude Styles are per-conversation selectable (one active at a time, changeable mid-session); ChatGPT custom instructions are global with only an on/off toggle per chat, not a multi-style picker.
- Warning: ChatGPT's 'How to respond' field has a 1,500-character limit; Claude Custom Style descriptions have their own separate limit. Long or detailed style instructions may need trimming when migrating in either direction.