Google makes it easier to switch from ChatGPT to Gemini: you can now import memory and chats

Google has effectively started a fight over the portability of AI context: Gemini can now pull in not just individual prompts from ChatGPT and other services, but also memory, user preferences and full conversation history. For people who spent months building up work habits, personal details and long project threads in one bot, this removes one of the main barriers to changing assistants. The new service no longer starts from a blank slate and doesn't force you to explain again who you are, what you work on and in what style you want answers.

Google announced the launch on March 26, 2026. The tool is rolling out to consumer Gemini accounts and appears directly in settings.

In essence, the company has split the transfer into two scenarios. The first covers memory: lasting facts about the user, their interests, relationships, recurring tasks, preferred tone and rules they previously set for another assistant. The second is importing the full chat archive, so old threads can be searched and continued inside Gemini. At the same time, Google is renaming the past chats section to memory, signaling that chat history is now treated as part of a broader personal context.

Memory import works without any direct sync between platforms. Gemini offers a ready-made prompt that you paste into your current AI service, for example ChatGPT. The prompt asks it to compile a structured summary of the user: basic details, interests, important relationships, recent projects and plans, and the explicit instructions the person has given the bot. The user then copies the response and pastes it back into Gemini. The system analyzes the text and saves the relevant details to its own memory. This approach is slightly less seamless than an automatic transfer via API, but it gives the user a clear point of control: you see exactly which information will go into the new service.

Full chat history works differently. In ChatGPT, you open your profile, go to Settings, then Data Controls, and request Export Data. OpenAI sends an email with a link to a ZIP archive containing the conversation history and related account data. That file is then uploaded to Gemini through the Chat History Import section. According to Google, the service accepts ZIP files up to 5 GB. Imported conversations appear in the regular chat list, where you can search them, open them and use them to continue earlier work. But not everything transfers: Gemini imports the text of conversations, while images, attachments and project files from the other assistant are not imported.

The feature also has important limitations. It is designed for free and paid personal accounts and is not supported for work, school, enterprise or child profiles. In addition, the launch does not yet extend to the United Kingdom, Switzerland and the countries of the European Economic Area. To use import, the user must be at least 18 years old, and personalization in Gemini must be turned on. After the transfer, you can delete individual conversations or an entire imported batch.

But the very purpose of the feature raises an obvious privacy question: if old chats contain sensitive data about work, family, health or finances, transferring the whole archive indiscriminately is probably not a good idea. Less friction when moving between bots also means users will more often move large chunks of their digital biography between platforms.

The main takeaway is simple: competition between AI assistants is shifting from raw model power to a fight over accumulated user context. Previously, choosing a new bot almost always meant losing your memory, habits and long interaction history, which strengthened lock-in to a single platform. Now Google is trying to loosen that lock-in and make Gemini a more convenient entry point for those who already lived inside ChatGPT or another service.

For the market, this is a signal that data portability is becoming a product feature in its own right; for users, it means they can experiment with different models without wiping their experience clean. But the easier such a transfer gets, the more important filters, deletion settings and a deliberate attitude toward what personal context an AI should remember at all.