Google’s 2025 in AI: the 40 features it wants users to remember
Google closed 2025 with a list of 40 AI uses across Gemini, Search, Photos, Maps, Workspace and Pixel. Here is what the list says about where the company put its effort.
Google closed 2025 with a list of 40 AI uses across Gemini, Search, Photos, Maps, Workspace and Pixel. Here is what the list says about where the company put its effort.
Speaking at the Axios AI+ Summit, the Google DeepMind chief predicted that models uniting text, images, video, audio, 3D and robotics will define the next year of AI.
A beta in the Google Translate app for Android brings real-time spoken translation in more than 70 languages to any headphones, with Gemini models improving text results.
Structured prompts, one model learned deeply, rich context and routine verification: seven practical rules for getting better results from ChatGPT, Gemini or Claude.
Google's guidance for its Gemini 3 based image model covers scene-setting, camera and lighting detail, multilingual text, image blending, editing commands and brand consistency.
A plain-language explainer on Google Gemini: its multimodal design, the model sizes Google offers, and how it is built into Gmail, Docs, Sheets, Drive and Android.
A NotebookLM update that uses Google's Nano Banana Pro image model to generate infographics and slide decks has found an early audience among people redesigning their resumes.
A guide to Google AI Studio, the free web platform where users can test Gemini models, generate and edit images, share a screen in real time and build apps from plain-language prompts.
Google's Guided Learning mode turns Gemini into a step-by-step tutor that asks questions, builds quizzes and works from your own notes. Here is how to use it.
Google has introduced Gemini-powered tools for Maps developers, including a Builder Agent that turns plain-language prompts into prototypes and an MCP-based Grounding Lite service.
New users of ChatGPT, Gemini and Claude often treat them like a search box. One short opening prompt, followed by a request for a demonstration, shows what they can do.
NotebookLM answers questions only from the files you upload, which makes it useful for staff training, onboarding and classroom study. Here is how it works.
Google's prompt engineering whitepaper, written by Lee Boonstra and hosted on Kaggle, lays out zero-shot, few-shot, chain-of-thought and ReAct prompting along with practical rules for getting better answers from large language models.
From Meta AI and ChatGPT to Gemini, Claude, Perplexity and DeepSeek, a guide to the leading AI assistants, how many people use them and what each is designed to do well.