The company is expanding AI features across its platforms, with YouTube set to introduce a new search experience called Ask YouTube.

The tool will allow users to ask more complex questions and refine their searches through follow-ups. Again, it will combine Shorts and long-form videos, as well as surface specific parts of clips that match what people are looking for.

“With Ask YouTube, you can ask more complex search queries, such as wanting tips on how to teach your kid to ride a bike, or finding creator reviews of cosy games to play before bedtime,” the company explained. “You can even ask follow-up questions to continue refining what you’re looking for.”

The feature is currently being tested with YouTube Premium users in the United States on desktop, with a wider rollout planned for later this year.

YouTube also announced a new video creation feature linked to Gemini models, including Gemini Omni. The company said the system will be used in Shorts Remix and the YouTube Create app to help users generate and edit video content with more control over output.

“Remixing with Omni delivers a fresh way for users to create and build on each other’s imagination,” YouTube wrote in a press release. “The model better understands user intent creating more consistent and meaningful storytelling while also handling complex video and audio adjustments behind the scenes.”

Alongside this, YouTube is expanding its likeness detection system for creators aged 18 and above. The tool is designed to help identify when a creator’s face is used in AI-generated content without permission. Creators can request takedowns where misuse is detected.

Separately, Google chief executive Sundar Pichai used his keynote at I/O to outline the company’s progress and the scale of its AI systems.

“It’s been an extraordinary year since our last I/O, a period of relentless shipping, technology advances and hyper progress,” he said. “We’re now in the part of the AI cycle where people want to see the value in the products they use every day.”

He said Google is still focused on building across its hardware, research and product stack, with AI now central to Search, Android, cloud services and developer tools.

He also pointed to rapid growth in usage across Google systems, with the company now processing far more data through its models than in previous years. Usage is also surging across consumer and enterprise tools.

Search is the company’s largest distribution point for its AI features. AI Overviews now has more than 2.5 billion monthly users. AI Mode has also passed 1 billion monthly users.

The Gemini app has grown as well, now having more than 900 million monthly users, up from 400 million the previous year. Google said usage frequency has increased significantly as new features are added.

The company also highlighted adoption by developers and cloud customers. Millions of developers now build with its models, while enterprise usage continues to scale across industries.

Pichai said the company is seeing strong engagement with generative tools for study, work and creative tasks, including image and video generation.

On infrastructure, Google is expanding investment in custom chips known as TPUs to support growing demand. It also outlined newer generations of its hardware designed for both training and inference workloads.

These systems are aimed at improving speed, efficiency and scale across its services, particularly Search and Gemini.

Google also introduced Gemini 3.5 Flash at the I/O conference, a faster and more cost-efficient model designed for broad use across products and developer tools. The company said it improves performance across coding and real-world tasks while reducing operating costs for large-scale usage.

In internal testing, the model is already being used in developer systems to speed up work on new tools and services. Google is also expanding its agent-focused platform, Antigravity, which allows users to manage autonomous AI agents that can carry out tasks over time.

A new version, Antigravity 2.0, is being released as a desktop application for developers and enterprises.

Google also introduced Gemini Spark, a personal agent designed to work across apps and services on behalf of users. It can run tasks in the background and update users on progress.

Spark will integrate with Google products and third-party tools. It will first roll out to trusted testers before expanding to subscribers in the United States.

On the consumer side, Google said it is moving Search towards more interactive and task-based experiences. New features will allow users to set up agents that monitor information and deliver updates automatically.

The company also announced new tools for content creation, at the I/O conference, including Google Flow for planning and editing complex projects, and Google Pics for image generation and editing using its Nano Banana model.

It shared progress on wearable devices, such as audio glasses that can deliver spoken assistance and upcoming display glasses that show information in real time.

Google is also expanding tools for science and research, linking its Gemini systems with external databases to support scientific work and experimentation.

The company said the updates are aiming for systems that can handle tasks, not just respond to queries, as it expands its products across consumer, developer and enterprise markets.

The headline claim is striking: Gemini 3.5 Flash is not only faster than all comparable frontier models,  running at four times the output token speed of rivals, it also outperforms Google’s own previous generation, Gemini 3.1 Pro, across coding, agentic reasoning, and multimodal understanding benchmarks. That is a meaningful inversion.

Historically, speed and quality have been a trade-off in AI development. With Flash 3.5, Google insists it has collapsed the gap.

Under the hood, the model supports native processing of text, images, audio, and video simultaneously while maintaining near real-time conversational responses. Google says it is designed for “long-horizon workflows” the kind of multi-step tasks that require an AI to hold a complex problem in mind and act across several tools or documents over time. Developers building on the Gemini API, as well as Google’s own Antigravity coding platform, are already running on Flash 3.5.

“When compared to 3.1 Pro, Flash is better across the board, the new model has been a game-changer for us internally at Google.”
— Sundar Pichai, Google I/O Keynote

To demonstrate this, a DeepMind engineer named Varun Mohan walked the audience through a live demo: using Antigravity and Gemini 3.5 Flash together, a team built a functioning operating system from scratch,  in twelve hours. The crowd applauded. The AI community paid close attention.