ChatGPT Memory: The Feature That Made ChatGPT Personal

ChatGPT Memory is OpenAI's feature that lets ChatGPT remember details across chats — your job, your writing style, your dog's name, your dietary restrictions. It was launched broadly in 2024 and redesigned in 2025 to include 'reference chat history' which pulls from recent conversations in addition to explicit memories. By 2026 it defines the personalized ChatGPT experience.

• Saving obvious personal context — you tell it once, it knows forever.
• Adapting tone and style — learns you prefer concise bullet points over prose.
• Professional context — remembers your company, role, projects for work chats.
• Managed memory view — you can see, edit, and delete every stored memory.
• Incognito mode — turn off memory for individual chats with 'Temporary chat'.

• Over-saves — stores random facts like 'user likes tacos' from a passing mention, pollutes future answers.
• Confuses contexts — work facts leak into personal chats and vice versa (Projects solve this, but default memory doesn't).
• Privacy concerns — this is personal data on OpenAI's servers, subpoenable in legal cases.
• Deletion isn't strongly expected after review — removed memories may persist in training snapshots.
• Reference-chat-history is opaque — you can't easily see what historical context is shaping a response.

• Included on all paid plans (Plus, Pro, Team, Enterprise) and free tier.
• Storage capacity scales with plan — Free has small capacity, Pro has the most.
• Enterprise can disable memory org-wide for compliance reasons.
• No separate cost — it's a feature, not an add-on.

Who should bother: daily ChatGPT users who want personalization, professionals whose work context is consistent. Who shouldn't: privacy-conscious users, anyone in regulated fields (healthcare, legal), shared account users. Memory is a great feature for a single-user personal account and a compliance headache everywhere else.