The Switching Cost Problem, Solved on Purpose
One of the quietest moats in the consumer AI market has nothing to do with model quality. It's the accumulated weight of everything your chatbot already knows about you — your preferences, your projects, the context you've painstakingly built up over months of use. Starting over with a new AI means losing all of that. Until now, that friction has quietly worked in favor of whoever got to a user first.
Google just decided to dismantle that advantage — and in doing so, revealed exactly how it thinks about where it stands in the chatbot race.
What Gemini Is Actually Launching
Google is rolling out two new features on desktop under the banner of "switching tools." The first, Import Memory, works through a prompt-based handshake: Gemini generates a suggested prompt you paste into your current chatbot, which produces a summary of what it knows about you. You paste that back into Gemini, and the transfer is done. Key preferences, personal context, relationship details — all migrated without manual re-entry.
The second feature, Import Chat History, is more ambitious. Users export their full chat logs from their previous AI — ChatGPT and Claude both support zip exports — and upload them directly to Gemini, up to 5GB. From there, Gemini can search and reference those old conversations, giving users the ability to continue threads that started elsewhere.
Google is also renaming what was previously called "past chats" to "memory," a rebranding that signals it's taking the persistent-context angle seriously as a product category, not just a feature.
Both tools are available to free and paid consumer accounts on desktop. Business, enterprise, and under-18 accounts are excluded for now.
This Is a Direct Response to the Numbers
The competitive context here is important. OpenAI announced last month that ChatGPT has reached 900 million weekly active users. Google, despite controlling the default AI placement across Android and Chrome — distribution advantages that most companies would consider unassailable — disclosed 750 million monthly active users for Gemini during its Q4 earnings call. Monthly versus weekly is a meaningful difference in how those figures are constructed.
Gemini has not converted its distribution dominance into mindshare dominance. The switching tools are a frank acknowledgment of that gap and a direct attempt to lower the barrier for users who've built their AI habits around ChatGPT or Claude.
This puts pressure on OpenAI and Anthropic in an interesting way. Both companies benefit when users feel locked into their platforms by accumulated context. Google is betting that if it makes the transfer easy enough, enough users will try Gemini — and that Gemini's quality is sufficient to retain them once they arrive. The bet only works if Google is confident in what happens after the import.
Anthropic Moved First
It's worth noting that Anthropic updated its own memory import tool earlier this month — this is Google responding to a move that's already in play, not inventing the category. The fact that both companies are now competing on ease-of-switching suggests the industry has implicitly concluded that lock-in via accumulated context isn't a sustainable strategy, or at least not the only one worth pursuing.
The more interesting question is whether OpenAI follows. ChatGPT's memory features are more mature than either Claude's or Gemini's, which means OpenAI has the most to lose if switching tools become standard. Making it easy to leave ChatGPT is not obviously in OpenAI's interest — but making it hard to leave while competitors make it easy to arrive could become a liability if users start to notice the asymmetry.
The Architecture of Trust
There's a subtler dimension to these features worth unpacking. When Gemini generates a prompt designed to extract what another AI knows about you, it's essentially coaching users on what personal information to hand over. The output of that prompt — a structured summary of your preferences, relationships, and history — gets stored in Gemini's memory system.
For users who are already comfortable sharing this kind of information with AI assistants, that's a convenience. For users who haven't thought carefully about what they've disclosed to ChatGPT or Claude over time, the import process makes the scope of that data suddenly visible and portable. That's not a criticism of the feature — it's arguably a step toward user awareness — but it's worth flagging that the data being moved around here is genuinely personal.
Google says users can delete specific imported chat histories or entire zip imports from the settings menu, which is the right design choice. Whether that granularity is sufficient, and how long imported data persists in Google's systems, are questions users should ask before uploading years of AI conversations.
What This Means
- For users currently on ChatGPT or Claude: The friction of switching to Gemini just dropped significantly. If you've been curious about Gemini but dreaded starting over, now is the time to run the experiment. The memory import won't be perfect — context loses texture in transit — but it's a reasonable starting point.
- For Google and Gemini's product team: The switching tools buy attention, not loyalty. Users who import their ChatGPT history and find Gemini's responses inferior will leave with less friction than before — the same tools that ease arrival also ease departure. Quality still has to close the deal.
- For OpenAI: The pressure here is asymmetric. Google is making it easy to arrive at Gemini; OpenAI's interest is in making ChatGPT worth staying in. If the memory import becomes a standard feature across the industry, the implicit lock-in of accumulated context evaporates, and model quality and product experience become the only durable differentiators.
- For Anthropic: Claude's earlier move on memory import puts it on the right side of the portability trend. The risk is that by making it easy to leave Claude, Anthropic accelerates churn to Gemini among users who were already considering the switch. Portability is good for users; it's a double-edged sword for any individual platform.
- For enterprise and developers: These features are explicitly consumer-only for now. But the underlying infrastructure — portable memory, structured personal context, importable chat history — is exactly what enterprise AI deployments will eventually demand. Watch this space for a business-tier version.
The chatbot market is slowly evolving from a competition over who can acquire users to a competition over who can retain them. Switching tools accelerate that evolution by making acquisition easier for everyone — and retention harder for anyone coasting on inertia.
