Anthropic launched a dedicated page this weekend for importing saved memories from ChatGPT, Gemini, or any other AI chatbot into Claude. The whole process takes about 60 seconds and requires zero technical skill: copy a prompt, paste it into your old chatbot, copy the response, paste it into Claude's memory settings. Done.
The feature itself is almost comically simple. Anthropic gives you a pre-written prompt that asks your current AI to dump everything it knows about you into a single text block. Your name, your job, your coding language preferences, your tone instructions, your recurring projects. Then you hand that block to Claude, and it processes the memories over the next 24 hours. As developer Simon Willison pointed out, the entire "import your memories" feature is, at its core, just a really good prompt.
The timing tells the story
This didn't land in a vacuum. The #QuitGPT hashtag has been burning through social media since OpenAI announced a deal with the Pentagon on February 28th, just hours after the Trump administration blacklisted Anthropic for refusing to allow Claude to be used for mass surveillance and autonomous weapons. According to multiple reports, over 700,000 users have announced they're canceling ChatGPT subscriptions. Claude shot to the top of the App Store.
That 700,000 number is self-reported cancellations across social media, not verified churn data from OpenAI, so take it with appropriate salt. But the sentiment is real enough that Anthropic clearly decided to capitalize on it.
And here's what makes the memory import clever rather than just timely: the biggest thing keeping people on ChatGPT isn't the model quality. It is the accumulated context. Months of conversations teaching the AI your preferences, your shorthand, your project details. Starting over on a new platform means weeks of re-explaining yourself. Anthropic just cut that switching cost to roughly zero.
What actually transfers (and what doesn't)
The import captures what your previous AI stored about you: response style preferences, personal details, project context, tools you use, corrections you've made. It does not transfer full chat histories. That distinction matters. Memories are curated summaries; chat histories are everything, including the throwaway questions and debugging sessions you'd rather forget.
There's a catch, though. Anthropic's support documentation admits the feature is "experimental and still in active development" and that Claude "may not always successfully incorporate imported memories." The system also skews toward work-related context, so personal details unrelated to professional use might get dropped. If you taught ChatGPT your favorite pizza toppings, Claude might not care.
Memory is only available on paid plans (Pro at $20/month, Max, Team, Enterprise). Free users don't get to play.
The privacy angle is doing heavy lifting
Anthropic is leaning hard into a contrast with both competitors. Claude memories are encrypted, not used for model training, and exportable at any time. Google has been testing its own import feature in Gemini's beta since early February, but there's a key difference: Google's version explicitly states that imported chats get stored in Gemini Activity and may be used to train their models.
That's a meaningful gap for anyone switching providers specifically because they're worried about how their data gets used. Which, given that the current migration wave is largely driven by concerns about AI companies cozying up to military and surveillance applications, is most of the people switching right now.
OpenAI's silence
OpenAI hasn't announced a comparable import feature. Why would they? As the market leader with 300 million weekly active users, making it easy to leave isn't in their interest. Anthropic, as the challenger, benefits from making switching frictionless. This is Econ 101 playing out in real time.
Is a prompt really a "feature"?
I keep coming back to the Willison observation. The core technology here is a well-crafted prompt that extracts structured data from one AI and feeds it to another. Anthropic built a polished landing page around it, added the processing pipeline to ingest memories on their end, but the fundamental mechanism is copy-paste.
Which is either brilliantly pragmatic or slightly underwhelming, depending on your expectations. There's no API integration, no OAuth handshake with OpenAI, no automated migration tool. You are literally the middleware. But it works, and Anthropic shipped it at exactly the moment when hundreds of thousands of people needed it.
Chief Product Officer Mike Krieger has described Claude's memory as being about "creating sustained thinking partnerships that evolve over weeks and months." The import feature just means that evolution doesn't have to start from zero.
The feature is live now at claude.com/import-memory. Existing users can start from their settings page. Whether the current wave of ChatGPT defectors sticks around once the political outrage cycle fades is a different question entirely, but at least they won't have to re-explain their coding preferences if they do.
