If you've used Claude in the past year, you probably remember it forgetting you the moment a new chat started. That changed, then changed again twice more since, which is exactly why the question keeps coming up. People who left Claude for that reason, or picked ChatGPT specifically because it remembered them, want to know if the gap actually closed or if this is a smaller feature wearing a bigger name.

It closed, and the two systems now look more alike than they used to. Claude used to keep memory as a single running summary. It doesn't anymore. As of Anthropic's August 2026 update, Claude stores memory the way ChatGPT does, as a list of individual entries you can open, edit, or delete one at a time.

Quick answer: Yes, Claude has memory, and since March 2, 2026 it's available on every plan, including free. As of August 2026, Claude stores memory as individual entries grouped under Topics in Settings, added to as you chat rather than summarized when a conversation ends, per Anthropic's own help center. ChatGPT works the same way structurally: saved memories you can view and delete one at a time, plus a separate reference to your chat history. If you want either one's memory to follow you into your actual texts, folk's import tool reads a copy-pasted export from both and adds it to its own long-term memory.

Does Claude have memory in 2026?

Yes, and the rollout has happened in stages. Anthropic first shipped memory in August 2025, limited to paid subscribers. By October 2025 it covered every paid tier: Pro, Max, Team, and Enterprise. On March 2, 2026, Anthropic dropped the paywall completely, so a free Claude account gets the same memory a paying one does, 9to5Mac reported at the time.

That same March announcement introduced an import tool: a prompt you can paste into ChatGPT or another chatbot asking it to list everything it has stored about you, so you can carry that answer into Claude instead of starting over.

Then, on August 25, 2026, Anthropic rebuilt memory again. It's now unified across Claude's chat product and Claude Cowork, its cloud coding tool, so context carries between the two instead of living in separate silos, 9to5Mac reported at the time, and it's on by default across web, desktop, and mobile for Free, Pro, and Max plans.

How does ChatGPT's memory feature work?

ChatGPT splits memory into two separate mechanisms. The first is saved memories, specific facts you asked it to remember, or that it flagged as worth keeping, stored as a running list you can review in Settings. The second is reference chat history, where ChatGPT can pull relevant details out of old conversations even if you never explicitly asked it to remember them.

You can turn either one off independently, and you can delete individual saved memories one at a time. One detail catches people off guard: deleting a chat does not delete the memories that came from it. Those live in the separate memory list and have to be removed there too.

What's the actual difference between Claude's memory and ChatGPT's?

Less than there used to be. ChatGPT's saved memories accumulate as a flat, growing list of individual facts. Claude now does the same thing structurally, entries grouped under Topics rather than one flat list, but the same idea: pick one, read it, edit it, or delete it without touching anything else.

The deletion behavior used to be the clearest difference between the two, and it isn't anymore. Removing the conversation that a memory came from does not remove that memory in either app now. ChatGPT never deleted memories when you deleted their source chat, and per Anthropic's help center, Claude doesn't either: "when a conversation expires or is deleted, related memory entries generated from it won't be removed, but you can delete individual memories at any time." Cleaning up memory and cleaning up chat history are two separate chores in both now.

One real difference remains: what Claude will store in the first place. By default, Claude excludes sensitive subjects, health, race, ethnicity, religion, politics, and gender identity among them, from memory entirely, and you have to opt in under Settings > Memory if you want those tracked too. ChatGPT doesn't offer an equivalent named toggle for excluding topic categories.

Neither approach is strictly better overall. Both are now easier to audit entry by entry than the old single-summary version of Claude was. ChatGPT's separate chat-history layer still picks up incidental detail without you asking it to remember anything, at the cost of being a second place to check.

Does it matter which one you pick just for the memory?

Less than it might seem, because both versions of memory share the same limitation: they only work inside that one assistant's own app. Whatever Claude or ChatGPT remembers about you stays there. Neither one reaches out to check in, reminds you about anything on a schedule, or shows up anywhere you didn't open its interface, memory in both cases is something you have to go ask about, not something that acts on your behalf.

If you've already built up memory in one and are weighing a move, the mechanics above matter less than whether you can bring what you already have with you. See how to move your ChatGPT memory to a texting AI assistant for the copy-paste steps that work for either Claude or ChatGPT.

How folk's memory compares

folk is the personal AI assistant that lives in your texts (iMessage, Telegram, and WhatsApp), and its memory is built to close the gap both Claude and ChatGPT leave open:

• It follows you into a messaging app, not just a browser tab. Long-term memory persists across conversations and across all three apps folk supports, same facts whether you text it on iMessage, Telegram, or WhatsApp. See memory for how it works day to day.
• You can see the whole thing, not guess at it. A memory viewer in the dashboard shows what folk has stored, grouped by people, topics, and ongoing things, the same idea as Claude's Topics list. Delete a single entry there or ask folk in chat to forget something specific, and a full wipe is one command away (CLEARMEMORY in chat, or the dashboard settings).
• It imports from either one. If you already have memory built up in ChatGPT or Claude, folk's import page takes a pasted export from either and folds it into folk's own memory, so switching doesn't mean starting from zero. The fuller walkthrough is in moving your ChatGPT memory to a texting AI assistant.
• Memory acts, it doesn't just recall. Because folk remembers you inside the same thread it can also message you first, reminders, follow-ups, a morning check-in, using what it knows rather than waiting for you to ask it to recall anything.
• Group-scoped memory too. In a shared iMessage or Telegram thread, folk keeps memory scoped to that group as well as to you individually, useful for anything a group chat needs to remember collectively.

The bottom line

Claude caught up to ChatGPT on memory, first by March 2026 for every user, not just paying ones, then again in August 2026 by rebuilding it as individual entries instead of one running summary. The two now solve the same problem in similar ways: both a list of entries you manage one at a time, plus ChatGPT's separate chat-history layer that Claude doesn't have an exact match for. Whichever you've been using, that memory isn't stuck there. Bring it with you if you're trying a texting AI assistant, then get started with folk from a place that already knows your basics instead of a blank one.

For the deeper walkthrough on the actual copy-paste move, read how to move your ChatGPT memory to a texting AI assistant. If context windows disappearing mid-conversation is the problem you're actually hitting, that's a different failure with a different fix, covered in why AI assistants forget your conversation. And if you just want to check what either one has actually saved about you before picking a side, see how to see everything an AI assistant remembers about you.