构建个人智能：迈向真正个性化AI的一步

A step towardstruly personal AI January 2026 8 min. Page 03Introduction1 Page 05 How Personal Intelligence works2 Page 08Responsible Development3 Page 12Research Directions4 Making our products more helpfulby making them personal At Google, we’ve spent yearsrefining our products to be notjust helpful, but helpful to you. We learned early on that personal context matters: whensomeone searches for “running shoes”, they aren’t usuallylooking for a list of generic bestsellers, but for the specificbrands and styles they prefer. That insight — that your historyand preferences should inform your results — enhancedGoogle Search, and in the years since, we’ve brought thatsame level of personalization to many of our products. Interacting with your personal information has historicallymeant navigating independent product experiences:searching for your flight reservation required opening Gmail,while finding a memory meant searching or scrolling throughGoogle Photos. In the last few years, we began bridging these gaps withfeatures that allowed for explicit information retrieval acrossproducts with your permission — like tagging @Gmail to findflight details in the Gemini app. We also introduced featuresthat personalize your experience in a specific product —like using your past chats in the Gemini app to give you amore tailored response. But these still don’t provide a fullypersonalized experience. Introducing Personal Intelligence Personal Intelligence marksthe shift to AI that can trulyunderstand your personal context. It brings together the information you share and theactivity within the products you use, while connectingthe dots across your Google apps to to make AIuniquely helpful for you. With Personal Intelligence, you can choose to connectcertain Google apps to get more tailored responses— all while your information remains safeguarded. Today, you can use a beta version of PersonalIntelligence in the Gemini app. It securely connectsinformation from apps like Gmail and Google Photosto make Gemini uniquely helpful. Personal Intelligenceis coming soon to AI Mode in Search. That means the Gemini appcan help with moretailored discovery and complex planning — likehelping you plan for spring break based on placesyou’ve already been or proactively finding the perfectset of tires for your specific car make and model. If you turn this beta feature on, you control whichapps to link, and each one supercharges theexperience. This is a foundational step toward moving beyondgeneric assistance toAI that works for you. It’s stillearly days for this technology, and we’re continuing towork through known technical issues and limitations.As we continue to test and learn, we are eager to hearyour feedback. The Technical Challenge:Solving the Context Packing Problem Personal Intelligence unlocks new utility by solvingfor the context packing problem: enabling ourGemini models to safely and accurately reason overdisparate and vast amounts of personal data sourcesin real-time without compromising user privacy. Search(coming soon) Gemini app Gemini model enabling Personal Intelligence Personal Intelligence Engine Gmail More Photos Advanced Gemini Models &A New Personal Intelligence Engine To handle the complexity of retrieving data from multiple sourcessimultaneously and provide significantly more helpful responsesto you, we built a new engine for Personal Intelligence. Personal Intelligence has two corestrengths: tool calls to retrieve specificdetails and reasoning across manycomplex sources. It often combinesboth approaches and can work acrosstext, photos, and video to give youone-of-a-kind answers. Advanced Reasoning Gemini 3, our most intelligent model series to date,is better at general understanding and decipheringmore depth and nuance — capabilities that are criticalfor understanding complex personal context, suchas mapping familial relationships or recognizing yourspecific aesthetic preferences. It enables Gemini models to securelyretrieve relevant context, leveragingadvances in tool use, dense retrieval,and long context capabilities to reasonacross your personal data from Googleproducts in real-time. Advanced Tool Use The Gemini model also has significant advances in tooluse capabilities which means it can understand your goaland retrieve more information related to your preferencesfrom the Personal Intelligence engine. This retrieval alsobuilds on the foundation of research we’ve done onsearch and dense retrieval, such asGemini Embeddings. Long Context Gemini 3 has a 1 million token context window whichenables reasoning across a vast amount of data.However, truly helpful personalization requires processingat a much larger scale, as a user’s accumulated contextacross emails and photos alone often exceeds thiswindow by orders of magnitude. To bridge this gap, we utilize “context packing”— a technique that helps us dynamically identify andsynthesize appropriate pieces of information into thew