Every major AI tool now has some form of persistent memory. But they all work differently, store different things, and give you different levels of control. If you're thinking about switching tools (or just curious about where your data lives), here's the full breakdown.
Quick comparison
| Feature | ChatGPT | Claude | Gemini | Copilot |
|---|---|---|---|---|
| Memory types | Saved Memories + Chat History inferences | Synthesized memory from conversations | Saved Info + Instructions | Saved Memories + Custom Instructions |
| Where to view it | Settings > Personalization > Manage Memories | Settings > Privacy > Memory preferences | Settings and Help > Instructions for Gemini | Settings > Personalization > Copilot Memory |
| Can you edit it? | Yes, delete individual entries | Yes, edit memory preferences | Yes, add/remove entries | Yes, delete individual entries |
| Bulk export? | Not built-in (use a prompt) | Available in Settings | Not built-in (use a prompt) | Not built-in (use a prompt) |
| Import from other tools? | Paste into "More about you" | Yes, dedicated import flow | Manual via Instructions | Paste into Custom Instructions |
ChatGPT's memory
ChatGPT has the most layered memory system of the bunch. There are two separate things going on.
Saved Memories are explicit. You say "remember this," and ChatGPT stores it as a discrete entry. You can view, delete, and manage these in Settings > Personalization > Manage Memories. Simple enough.
Chat History inferences are implicit. ChatGPT watches your conversations and builds up a model of your preferences, even if you never explicitly ask it to remember anything. These inferences don't show up in the Manage Memories screen, which is why most people don't know they exist. The only way to see them is to ask ChatGPT directly. I covered this in detail in What Does ChatGPT Actually Remember About You?
ChatGPT also has a "More about you" text field in Custom Instructions where you can paste persistent context. This is a great target for importing preferences from other tools.
Claude's memory
Claude takes a different approach. Instead of storing individual memory entries, Claude synthesizes what it learns from your conversations into a holistic profile. Think of it less like a database of facts and more like a running summary of who you are and how you like to work.
You can view and manage this in Settings > Privacy > Memory preferences. Claude also has a dedicated memory import flow at claude.ai/import-memory, which is the easiest way to bring in preferences from another tool. Just paste, and Claude processes the rest. It runs daily synthesis cycles, so it may take up to 24 hours to fully incorporate new imports.
Claude's memory also works per-Project if you're using Projects. Each Project has its own separate memory space, which is handy if you use Claude for very different kinds of work.
Gemini's memory
Gemini's system is more manual than the others. The main way to save preferences is through Instructions for Gemini in Settings and Help. These are entries you add yourself, like "I prefer formal language" or "I'm a product manager at a fintech company."
Gemini also picks up some things from conversations, but it's less aggressive about it than ChatGPT. If you want Gemini to remember something, you're better off adding it explicitly through the Instructions page or saying "remember this" in a conversation.
One nice thing: Instructions apply globally across Gemini, including to any Gems you've created. So your preferences follow you everywhere within the Gemini ecosystem.
The downside is there's no bulk import. If you're bringing in a big export from another tool, you'll need to either add entries one by one through the Instructions page or paste everything into a conversation and ask Gemini to save it.
Copilot's memory
Microsoft Copilot has three layers of personalization. Custom Instructions are things you set explicitly. Saved Memories are things you asked it to remember in conversation. And there's a third layer of inferences from your chat history, similar to ChatGPT.
You can manage all of this in the ... menu > Settings > Personalization > Copilot Memory.
One thing to watch for: if you're using the work version of Copilot (Microsoft 365 Copilot), your admin controls whether memory is enabled. If your organization has Enhanced Personalization turned off, Copilot won't store memories at all. Check with your IT team if you're not sure.
Copilot also pulls in some context from your Microsoft profile (like your job title and location from Entra ID), which is a nice touch but means your export might include things you didn't explicitly tell Copilot.
Moving between them
None of these tools have built-in interoperability with each other. You can't click a button in ChatGPT that says "send my memories to Claude." That's just not how it works right now.
But you can do it manually, and it's faster than you'd think. The process is the same for every pair: use a prompt to export your memory from the source tool, review and trim it in the AI Memory Editor, then paste it into the destination tool's import or instructions field. The whole thing takes a couple of minutes.
We have step-by-step guides for the most popular migration paths:
ChatGPT to Claude | Gemini to Claude | Copilot to Claude | ChatGPT to Gemini | Claude to ChatGPT | Gemini to ChatGPT
Pick your pair, follow the steps, and you're set.