Passer de ChatGPT à Gemini sans RIEN perdre (Tuto Complet)

This tutorial explains how to migrate from ChatGPT to Gemini, ensuring no data loss. Gemini has a new feature for importing memory from other AI tools, but the built-in methodology is limited. This guide provides a comprehensive method to recover all types of memory, including saved GPTs, projects, and conversations. First, you need to enable the memory setting for past discussions in Gemini. Go to your settings, then "personal context," and activate the memory option. This allows Gemini to remember information from your conversations, similar to ChatGPT. You can manage and delete this memory later if needed. Next, navigate to "import memory into Gemini" in your settings. Gemini offers a basic prompt to retrieve information from other AI tools. However, this prompt is not optimized for ChatGPT and may result in significant data loss. A more comprehensive prompt has been developed and is available for free in the AI resource base (link in description). Comparing the two prompts, Gemini's prompt uses persona prompting but is vague about the mission. It requests information in five broad categories: demographic information, interests and preferences, confirmed relationships, dated projects and plans, and instructions. The format requested is light. The custom prompt is far more detailed. It describes the task for ChatGPT more thoroughly, asking for all information, even sensitive or seemingly insignificant data. It includes strict rules and more segmented categories to ensure better exploitation by Gemini. It also provides precise instructions for searching memory, more detailed formatting requirements, writing rules, expected outputs, and final verifications, which the original prompt lacked. An infographic illustrates the benefits of using the custom prompt. Gemini's prompt is basic, while the custom prompt is structured around eight blocks: objective, strict rules, sources to search, data structure, memory data formatting for Gemini, detailed categories, a second pass for completeness, access and limits for sensitive information, and a final check before delivery. The advantages of using the custom prompt are numerous: you recover more information, avoid overly summarized data (Gemini's method aims to reduce token consumption, but users want maximum information), and reduce hallucinations. The custom prompt ensures exact formulations, captures weak signals to identify crucial, outdated, or irrelevant information for better transfer to Gemini, and provides a more transparent export. To use the custom prompt, copy it from the AI resource base and paste it into a new ChatGPT discussion. It is crucial to select the GPT 5.5 model and the "extended" reflection level to ensure all information is retrieved. Once pasted, instruct ChatGPT to "execute" the prompt. The process is quick; an export can be generated in about a minute, providing all memory information. This information is highly private, so it is not displayed in detail, but its structured format is shown. Copy this retrieved information and paste it into Gemini, then click "add memory." Gemini will then launch a discussion, confirming it has retained the most important information. This step highlights why activating conversation memory in Gemini's personal context settings is essential. However, importing memory is not enough. You also need to import other elements, such as conversations and instructions on how the AI should behave. Before moving on, the speaker introduces a private AI community. This community offers regular support, private live sessions, and a network of over 900 members for exchanging ideas, solving problems, and collaborating. It features sections for help, collaboration, debates, and exercises, helping users move beyond isolated AI tinkering. The community also includes a Classroom section with tutorials and training, a toolbox with prompts and tools, and a lab for sharing AI experiments. Comprehensive training is available for N8N, Claude, Gemini, and ChatGPT, including over two hours on Codex, an often-underestimated AI tool. The community also has a calendar of upcoming events and a hotline for questions. The community is currently available at a promotional price of $49 per month (excluding VAT, if applicable). This price will increase to $79 per month soon. Annual subscriptions offer two months free. Returning to the migration process, the next step is to migrate "saved memories." These are specific pieces of information you explicitly asked ChatGPT to remember. To retrieve them, go to ChatGPT, access your account, select "personalization," then "memory," and click "manage." This will display all the information ChatGPT has stored at your request. Copy this information. In Gemini, start a new discussion and paste the copied saved memories. Use a prompt instructing Gemini to "retain the information in its memory" and to "replace ChatGPT with Gemini where necessary." This is important because some saved memories might contain instructions specific to ChatGPT (e.g., "ChatGPT should do this"). Once the prompt and information are entered, launch the discussion. Gemini will process and integrate these memories. This step ensures that Gemini has a memory as rich as your previous ChatGPT setup. The next crucial step is to recover your conversations. In Gemini, under "import memory into Gemini," you will find an option to "import discussions" via a ZIP file up to 5GB. To obtain this ZIP file from ChatGPT, go to your ChatGPT settings, then "data management," and click "export data." Confirm the export. It may take several hours or even a day to receive the ZIP file via email. The download link expires in 24 hours, so retrieve it promptly. Once you have the ZIP file (which for intensive users can be around 3GB, well within the 5GB limit), return to Gemini's "import memory into Gemini" section, click "add," and upload your ZIP file. The discussions will begin importing into Gemini, allowing you to continue past conversations and further enrich Gemini's memory since conversation memory is active. Finally, you need to configure your custom instructions in Gemini. Go to settings, "personal context," and below the memory section, you will find "instructions for Gemini." Ensure this is activated. Fill the block with key information about your work style and how you prefer the AI to respond. If you previously set up custom instructions in ChatGPT, you can copy them. In ChatGPT, go to settings, "personalization," and access "more information about you." Copy these details. In Gemini's personal context section, you can modify existing instructions or add new ones. The speaker provides an example of structured custom instructions: profession/role, projects/challenges, specific interests/passions, values/principles, personal context, goals, preferences, language proficiency, specialized knowledge, training, and communication style. If you haven't created custom instructions before, a linked video tutorial on the "bricks method" can guide you. Be mindful of character limits; if the initial attempt fails due to too much information, reduce the content. Once accepted, these custom instructions ensure Gemini behaves as desired. The last part of the migration involves moving your projects and custom GPTs. This is a manual process. For custom GPTs, go to the GPT you wish to migrate in ChatGPT, click "modify GPT," and retrieve all instructions and any documents from its knowledge base. In Gemini, the equivalent feature is "Gems." Click "new gem," and you will be prompted to enter a name, description, instructions, tools, and a knowledge base. Copy the name, description, and instructions from your ChatGPT GPT. For knowledge base files, add them to the Gemini gem. If the GPT uses specific tools (e.g., image creation), select them in Gemini. After filling in the details, you can test the gem and then save it. Repeat this for all custom GPTs you have created. If you use a GPT created by someone else, you can ask it how it was designed and request its instructions to import into Gemini. For ChatGPT projects, which provide a contextualized workspace with specific files and instructions, there isn't a direct equivalent in Gemini. However, you can use Google's Notebook LM, an included tool with Gemini subscriptions. To create a project in Notebook LM, click "create a notebook" and import all project files. Then, add the project instructions by clicking "add sources," selecting "copy text," pasting the instructions, and naming the source "instruction." This way, when you want to work on a project in Gemini, you can select the relevant Notebook LM project via "more imports" and "notebook." The discussion will then be contextualized by the information in the notebook, mimicking the ChatGPT project functionality. This comprehensive process ensures a complete migration from ChatGPT to Gemini, recovering all memories, conversations, custom instructions, GPTs, and projects. While recovering GPTs and projects is the most time-consuming part, it offers an opportunity to clean up and organize them. The custom prompt for memory retrieval is available for free in the AI resource base. The private AI community offers further support and training to maximize the benefits of AI tools like Gemini. The community is commitment-free and cancellable anytime. The import of conversations may take time due to large data volumes, but it will eventually integrate into Gemini. This tutorial aims to provide actionable steps to leverage AI opportunities effectively.