Does Claude Remember Things About You? (And Why It Doesn't Know Your ChatGPT History)
If you used Claude before 2026, you remember it forgetting everything between conversations
If you used Claude before 2026, you remember it forgetting everything between conversations. That's no longer true — and a lot of advice online is out of date. Claude does remember things about you now. The question worth asking is what kind of memory it is, and why it still doesn't give you the continuity you actually want.
Does Claude remember things about you?
Yes — since March 2, 2026. Anthropic rolled out Chat Memory to every Claude plan, including free. Before that date, Claude was stateless by default: every conversation started cold. Guides written earlier say Claude "doesn't remember anything," and they're simply outdated now.
With Chat Memory on, Claude builds a persistent profile of you — your preferences, recurring topics, projects, communication style — and carries it across conversations. Mention you're a founder building B2B SaaS, and it'll hold onto that without you repeating it.
How does Claude's memory work?
Automatically, and on a cycle. Claude synthesizes your conversations into a memory profile roughly every 24 hours. You don't manually save entries the way you can in ChatGPT — extraction is automatic.
You stay in control of the result. In Settings → Capabilities → Memory you can view everything stored, edit it, or clear it entirely. Projects get their own separate memory space, scoped to that project rather than your global profile. One genuinely useful detail: Claude tells you when it's drawing on a memory, which ChatGPT doesn't.
So Claude's memory is real, automatic, and more transparent than most. It's still worth understanding what it isn't.
Can Claude see your ChatGPT history?
Partly — through a one-time import, not a live connection. Claude's Memory Import tool, launched alongside Chat Memory, pulls a structured export from ChatGPT (and Gemini, and Grok) into Claude's memory.
But "import" is the operative word. It's a snapshot taken once, not an ongoing sync. New things you tell ChatGPT next week don't appear in Claude. And the raw ChatGPT export doesn't carry over as usable memory on its own — the conversation history is just JSON and HTML; it doesn't change Claude's behavior unless reshaped. So Claude can be seeded from ChatGPT once, but the two don't stay in step.
There's a deeper version of this problem worth knowing about. You can connect Gmail, Notion, and other sources to Claude directly via MCP, hoping it'll pull context from them. But connectors give Claude access to a source — a way to search or retrieve from it — which isn't the same as handing it a curated, structured profile of you. Depending on how a connector retrieves, asking for "the email about our Q3 strategy" can surface what literally matches and miss the thread titled "Planning for next quarter" that never used the word. Connecting your tools gives the AI somewhere to look; it doesn't automatically produce the right context.
Why isn't that enough?
Because Chat Memory, for all its improvements, is still three things: reactive, Claude-only, and unstructured.
It's reactive — built from what you happened to discuss, accumulated over time, starting empty.
It's Claude-only — your Chat Memory profile lives in Claude. Open Cursor or ChatGPT and none of it follows. Each tool keeps its own separate picture.
It's inferred, not structured — a synthesis of conversations, not an accurate profile drawn from your real sources. You can edit it, but you're editing the model's guesses, not curating ground truth.
For "what did we discuss," that's fine. For "who am I, accurately, in every tool I use," it falls short.
What's the alternative?
A portable context layer. Instead of letting each platform build its own reactive memory, you maintain one structured profile — extracted from your real sources, owned by you — and deliver it to any tool through MCP. Accurate from the first message, identical across Claude, Cursor, ChatGPT, and anything else.
Claude's Chat Memory and a portable context layer aren't in conflict. Memory handles session-to-session recall within Claude. Portable context handles the foundational layer — who you are — that every tool should know and none should have to relearn.
→ The full distinction: AI Memory vs. AI Context: What's the Difference?
→ Moving context between tools: How to Move Your Context from ChatGPT to Claude
