Follow ZDNET: Add us as a preferred source on Google.
ZDNET's key takeaways
- Claude AI now lets you copy memories from another AI service.
- The goal is to help you easily switch to Claude.
- The process uses instructions that you can copy and paste.
Building a digital second brain usually involves weeks of tedious back-and-forth, as you slowly teach an AI your professional quirks, favorite hobbies, and specific writing style. While most platforms rely on this slow-burn "memory" to personalize your experience, Anthropic's Claude is cutting out the middle man.
A new migration tool for Claude now lets you instantly import your existing preferences and background data from another AI. Instead of starting from scratch and repeating your life story, you can port over your curated profile in a single move.
With the new memory import option, you can transfer memories to Claude from another AI, such as ChatGPT, Google Gemini, or Microsoft Copilot. The objective is to help you switch to Claude without having to start over with fresh new memories and other preferences.
Also: New GPT-5.4 clobbers humans on pro-level work in OpenAI's tests - by 83%
Claude has been picking up a lot of steam lately, as evidenced by the iOS app's top spot among free apps in Apple's App Store. At the same time, ChatGPT is being hit by a QuitGPT campaign. Clearly, Anthropic smelled an opportunity here to encourage people looking to jump ship to its own AI. Even if you're not quitting another AI but simply want to give Claude the same memories, you can still use the memory import tool.
(Disclosure: Ziff Davis, ZDNET's parent company, filed an April 2025 lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.)
How to use the memory import tool
Here's how to try this.
First, set up your account with Claude if you don't have one yet. A free or paid plan will suffice, as you can perform the memory import either way. Head to Claude's memory import page and click the button to start importing to Claude.
If you're already using Claude, you can also access the memory import via settings. For that, click your account name at the bottom of the left sidebar, then choose Settings and select the Privacy setting. On the Privacy page, click the Manage button next to Memory preferences, then select the Start import button.
Also: Why I stopped using ChatGPT for everything
A small window pops up, prompting you to import memory to Claude. The necessary instructions are conveniently generated for you. Just click the Copy button to copy them.
Next, sign in to the other AI service from which you want to import your memories. Paste the copied instructions at the prompt and then submit them. The instructions read as follows:
I'm moving to another service and need to export my data. List every memory you have stored about me, as well as any context you've learned about me from past conversations. Output everything in a single code block so I can easily copy it.
Format each entry as: [date saved, if available] - memory content.
Make sure to cover all of the following — preserve my words verbatim where possible:
- Instructions I've given you about how to respond (tone, format, style, 'always do X', 'never do Y').
- Personal details: name, location, job, family, interests.
- Projects, goals, and recurring topics.
- Tools, languages, and frameworks I use.
In the response, click the Copy button to copy the memories and preferences. If you want to apply them as is, switch back to Claude. But you may want to paste them into a text editor to review them first. That's what I did, and I found some memories I didn't want to retain or transfer. Each memory is stored as a separate text string, so you can easily delete any of them.
When done, copy the revised information, then paste it into the appropriate field to import memory to Claude, or simply paste it as a new prompt. Click the Add to Memory button. After a few seconds, Claude will display a formatted list of everything it now knows about you. Close that window.
Also: I used Nano Banana 2 to make perfect sketchnotes: 5 lessons learned
Fire up a new chat and ask Claude what it knows about you. The AI should display all the details you added via the import.
If you ever decide you no longer want Claude to remember what it's learned about you, go back to the Memory page under Settings. Hover over the "Memory from your chats" section and click the trash can icon. Here, you're also able to turn off the option for "Generate memory from chat history." Alternatively, just tell Claude at a prompt that you want it to remove all stored memories about it, and the AI should comply.
Facts Only
Anthropic’s Claude AI has launched a memory import tool.
The tool allows users to transfer memories from other AI services like ChatGPT, Google Gemini, or Microsoft Copilot.
Users can access the import feature via Claude’s settings or a dedicated memory import page.
The process involves copying and pasting a generated prompt into the original AI service to extract stored memories.
Memories are formatted as text strings, which users can review and edit before importing.
The tool is available for both free and paid Claude accounts.
Claude’s iOS app recently ranked as the top free app in Apple’s App Store.
ChatGPT is currently facing a "QuitGPT" campaign.
Users can delete imported memories or disable memory generation from chat history.
Ziff Davis, ZDNET’s parent company, has filed a lawsuit against OpenAI alleging copyright infringement.
Executive Summary
Full Take
The strongest version of this narrative is that Claude’s memory import tool represents a user-centric innovation, addressing a genuine pain point in AI adoption: the tedious process of rebuilding personalized preferences when switching platforms. By allowing seamless migration, Claude positions itself as a more flexible and user-friendly alternative, particularly as ChatGPT faces backlash. The tool’s design—letting users review and edit memories before import—also reinforces transparency and control, aligning with principles of digital autonomy.
However, the narrative also subtly leverages competitive dynamics. The mention of ChatGPT’s "QuitGPT" campaign and Claude’s App Store ranking frames the tool as both a practical solution and a strategic move to capitalize on user dissatisfaction. While not overtly manipulative, this framing could be seen as an appeal to popularity (ARC-0012 Bandwagon) or a form of contrast framing (ARC-0031 False Binary), where the choice is implicitly presented as "stay with a troubled service or switch to a rising one."
The root cause here is the broader tension in AI adoption: users want personalization without lock-in, but platforms benefit from proprietary data. Claude’s tool challenges this paradigm by treating memories as portable, which could pressure other services to follow suit. Yet, the long-term implications are mixed. While users gain agency, the tool also normalizes the idea that personal data is a commodity to be transferred between corporations—raising questions about ownership and privacy. Who truly benefits? Users gain convenience, but Anthropic gains market share. The second-order consequence may be a race to the bottom in data portability, where platforms compete on ease of migration rather than core capabilities.
Bridge questions: How might this tool reshape expectations around data ownership in AI? If memories become portable, does that reduce or increase the risk of data exploitation? What would a truly user-owned memory system look like, beyond corporate platforms?
Counterstrike scan: A coordinated influence campaign would amplify the "QuitGPT" narrative while positioning Claude as the ethical alternative, using emotional appeals (ARC-0008 Fear Appeal) and borrowed credibility (ARC-0021 Authority Games) via App Store rankings. The actual content doesn’t fully match this pattern—it presents the tool as a neutral feature—but the framing leans into competitive dynamics. No structural alignment with a hypothetical attack playbook is detected beyond standard market positioning.
Patterns detected: ARC-0012 Bandwagon, ARC-0031 False Binary
Sentinel — Human
This article is likely written by a human journalist, as evidenced by its engaging tone, anecdotal examples, varying sentence length, and lack of mechanical structure.