Google I/O 2025 Key Highlights: Android XR glasses, AI Mode in Search, Gemini 2.5 updates, and other big announcements

Keep a tab on the tech section of The Indian Express for more AI news coming out of Google I/O 2025.

"More intelligence is available for everyone, everywhere. And the world is responding, adopting AI faster than ever before…,” Google CEO Sundar Pichai said, adding that the company has released over a dozen foundation models since the last edition of the conference. You can read the full on-ground, news report of The Indian Express here.

Here's an easy-to-digest list of everything that was announced during the opening keynote of Google I/O 2025:

- Updates to Gemini 2.5 Flash and Pro models, including native audio outputs

- Deep Think reasoning mode in Gemini 2.5 Pro

- Shopping in AI Mode, agentic checkout, and a virtual try-on tool

- New updates to Deep Research and Canvas

- Imagen 4

- Veo 3

- Synth ID detector

- Personalised smart replies in Gmail

- Speech translation in Meet

- Support for Anthropic's Model Context Protocol (MCP)

- Search Live in AI Mode and Lens

- Jules, an asynchronous AI coding agent

- Flow, a new AI filmmaking tool

- Beam, a 3D AI-first video communication platform

- Google AI Pro and AI Ultra subscription plans

- A look at Android XR-powered smart glasses

- Expanded access to AI Overviews

- Gemini in Google Chrome

- Expanded access to Gemini Live's camera and screen sharing capabilities

-

Sundar Pichai is back onstage talking about examples of how AI is helping people such as detecting wildfires and drone deliveries of critical medicine.

Android XR supports a broad spectrum of devices including headsets, smart glasses, and more. It is built in partnership with Samsung and optimised for Snapdragon Qualcomm. Samsung's Project Moohan will be available later this year.

Meanwhile, Glasses with Android XR will let Gemini see and hear the world. It works with your phone for hands-free access to apps.

Google AI Pro will be available globally with a full suite of AI products, higher limits, and Gemini Advanced. It is priced at $19.99/month. Meanwhile, Google AI Ultra is priced at $249.99/month AI subscription plan, offering higher usage limits and access to its top-tier AI models and features. AI Ultra is only available in the US for now.

Flow combines the best of Imagen, Veo, and Gemini. While the AI tool relies on the Veo and Imagen models for generating cinematic visuals, the Gemini integration is meant to make prompting more intuitive so that filmmakers can describe their vision in everyday language.

Google also launched a new Synth ID detector, a verification platform to help users identify whether a piece of content is AI-generated or now. Users have to just upload a piece of content and the SynthID Detector will identify if either the entire file or just a part of it has SynthID in it.

That means Veo 3, Google's AI video generator, can generate sound effects, background sounds, and dialogue.

The AI tool is available for Google Ultra subscribers in the US via the Gemini app and in Flow. It’s also available for enterprise users on Vertex AI.

Imagen 4 has been upgraded to generate 2k resolution images with clarity in fine details like intricate fabrics, water droplets, and animal fur. The AI-generated images are now available to download in a range of aspect ratios. Imagen 4 is also significantly better at spelling and typography, Google claimed. It is available today in the Gemini app, Whisk, Vertex AI and across Gemini Workspace.

Gemini is coming to the desktop version of Google Chrome but only for Google AI Pro and Google AI Ultra subscribers in the US. This enables users to ask Gemini to clarify or summarise any information on any webpage that users are reading. “In the future, Gemini will be able to work across multiple tabs and navigate websites on your behalf,” Google said.

Starting today, users can get a more customised Deep Research report by uploading their own private PDFs and images. "For instance, a market researcher can now upload internal sales figures (as PDFs) to cross-reference with public market trends, all within Deep Research," Google said.

On Canvas, users can use AI to generate interactive infographics, quizzes and podcast-style Audio Overviews in 45 languages.

The “try on” experimental feature is rolling out in Search Labs in the US today.

With AI Mode, users who search for products will be shown a browsable panel of images and product listings personalised to their taste. The righthand panel dynamically updates with relevant products and images, helping users pinpoint exactly what they’re looking for and discover new brands.

Google is bringing Project Astra’s live capabilities into Search. With Search Live, users can talk back-and-forth with Search about what they see in real-time, using their phone camera. For instance, if a student wants help with their project, they can tap the Live icon in AI Mode or Lens, point their camera, and ask a question.

AI Mode offers a more dynamic UI with a combination of text, images, and links. It uses a query fan-out technique to break down a user's question into subtopics and issuing a multitude of queries simultaneously. This enables Search to dive deeper into the web than a traditional search on Google, helping users discover more of what the web has to offer and find hyper-relevant content that matches a user's question.

AI Mode will soon be able to make its responses even more helpful with personalised suggestions based on past searches. Users can also opt-in to connect other apps such as Gmail with AI Mode. "Because your flight and hotel confirmations are in your inbox, it will sync when you actually be in Nashville (for instance)."

Personal Context in AI Mode is coming this summer. AI Mode will also get Deep Search to help users unpack a topic. It can issue hundreds of searches on a user's behalf to create a research report.

Complex data analysis and visualisations will also be possible via AI Mode. Project Mariner's agentic capabilities also coming to AI Mode to help users, for instance, find ticket options for sporting events.

AI Overviews is driving over 10% increase in usage of Google for the types of queries that show AI Overviews. People are coming to Google to ask more of their questions, including more complex, longer and multimodal questions, says Pichai.

"Deep Think uses our cutting edge research in thinking and reasoning including parallel techniques," says DeepMind's Demis Hassabis. Deep Think is currently being rolled out to Google's trusted testers.

Jules, an AI agent powered by Gemini 2.5 Pro and capable of autonomously reading and generating code, makes its debut at Google I/0 2025. Users can integrate Jules directly into their existing code repositories. The AI coding agent then makes a clone of the user’s codebase within Google Cloud virtual machine (VM) to ‘understand’ the context of the project and perform tasks such as writing tests, building new features, fixing bugs, etc.

Google is introducing a preview version of audio-visual input and native audio output dialogue for both Gemini 2.5 Flash and Pro. It allows the user to steer the AI-generated tone, accent and style of speaking. For example, you can tell the model to use a stronger voice or whisper when telling a story. It can also switch between languages (say from English to Hindi) mid-speech.

The text-to-speech capability is now available in the Gemini API.

Tulsee Doshi, Senior Director, Product Management at Google DeepMind, onstage talking about the updated versions of Gemini 2.5 Flash and Pro.

"AI is making an amazing new future possible. Gemini 2.5 Pro is our most intelligent model yet, and we’re truly impressed with what we’ve created. It can build entire apps from sketches. It also leads in learning and ranks number one across multiple benchmarks. I’m thrilled to announce Gemini 2.5 Flash, which comes with enhanced capabilities. Flash will be generally available in early June, and we’re making final adjustments based on your feedback," said Google DeepMind chief Demis Hassabis.

Google launches personalized smart replies in Gmail. It will help users draft emails that match their specific context and

tone. Available in Gmail this summer for subscribers. It pulls from a user's past emails and Google Drive to provide suggestions that are more relevant to the user.