Anthropic has quietly shipped a feature that chips away at one of the strongest moats in consumer AI: it now lets you import your preferences and accumulated context from another assistant straight into Claude. The workflow is almost comically simple — copy a prompt, paste the result — and with it, users can carry years of refined AI training over from ChatGPT or another assistant instead of starting from a blank slate.
The mechanism matters less than the message. If the "memory" that used to lock you into one assistant can be exported and re-imported in two minutes, then the switching cost that kept users loyal starts to evaporate. For brands, that turns a subtle background fact into a strategic one: your customers are not committed to a single AI, and neither is their next question about you.
Key takeaways
- Anthropic's memory import lets users move preferences, working style, and context into Claude via a copy-paste prompt — no API integration or data export required. - The feature directly attacks switching costs, the "sunk cost of training" that kept users tied to one assistant. - Lower lock-in means users will move more fluidly among ChatGPT, Gemini, Claude, and others, and carry their context with them. - The GEO implication is the headline: if customers roam across assistants, single-platform AI visibility is a blind spot — you have to be findable and correctly represented on every engine, not just one. - Being strong in ChatGPT while invisible in Claude or Gemini is now a real and measurable gap, not a rounding error.
The feature: portability as a weapon
Claude's memory import targets the biggest barrier to switching assistants — the months users spend refining interactions, establishing preferences, and building context that a rival product would make them rebuild from zero. The process sidesteps all of that. You copy a prompt Anthropic provides into your current assistant, it generates a summary of your preferences and accumulated context, and you paste that summary into Claude's memory settings. Claude updates its memory and picks up roughly where you left off.
There are no API integrations and no formal data exports — just a clever use of a language model's own summarization ability to bridge two services. Per Anthropic, what carries over spans preferences (how you like information presented, tone, formatting), context (project knowledge, ongoing work, accumulated insights), and working style (communication patterns, typical requests, established workflows). Claude then maintains that context, keeps project details separated to avoid cross-contamination, and gives users visibility into what it has stored.
Why it matters: the lock-in war heats up
For most of the chatbot era, memory was the quiet reason people stayed put. The assistant that knew your projects, your preferences, and your history was worth more than a marginally smarter competitor, because leaving meant re-teaching a new system everything. Anthropic just reframed that memory as portable — something you own and can move — rather than something a provider holds hostage.
