Google Unveils New Weapon to "Poach" Users from ChatGPT, Claude, and Others

Google released a new feature called “switching tools” in its AI chatbot, Gemini, on Thursday (26/3/2026). This feature allows users to transfer “memory” (personal information, preferences, etc.) as well as conversation histories from other AI chatbots (such as ChatGPT, Claude, etc.) to Gemini instantly. The feature is designed to make it easier for users who want to switch to Gemini without having to start from the beginning. Previously, users typically needed to “retrain” a new chatbot by re-entering preferences and personal information. Only then could the AI chatbot “connect” and “understand” the user. Google explained that users can transfer data in two ways. First, through a zip file containing conversation histories from the previous chatbot. This file can be downloaded from the settings of the old application, then uploaded to Gemini. The second method is by transferring “memory” using a special prompt. Users simply click “start” on the gemini.google.com page. Then, Gemini will provide instructions that can be copied to the old chatbot. As follows: You are helping me import context from one AI assistant to another AI assistant. Your task is to read our previous conversation and summarise what you know about me. In the output, avoid using first-person pronouns (I, my) and second-person pronouns (you, your). Instead, refer to the individual you have learned about as “the user” or use neutral phrases. Retain the user’s words as they are where possible, especially for instructions and preferences. Categories (output in this order): 1. Demographic Information: Preferred name, profession, education, and general location of residence. 2. Interests & Preferences: Ongoing active interactions (not just owning objects or one-time purchases). 3. Relationships: Confirmed and ongoing relationships. 4. Events, Projects & Plans by Date: Log of recent important activities. 5. Instructions: Rules that I have explicitly asked you to follow going forward, “always do X”, “never do Y”, and corrections to your behaviour. Only include rules from stored memory, not from conversation. Format: