Google I/O 2025: AI Ultra subscription, AI Search, and more unveiled

At its recent I/O developer conference, Google unveiled several new AI-powered tools and products, including a premium subscription service called Google AI Ultra. Here are some key announcements.  

Google AI Ultra Subscription

The subscription plan is priced at $249.99 per month in the US and targets professionals and power users requiring advanced artificial intelligence (AI) tools. It provides access to Google's most capable AI models, including Deep Think for complex reasoning tasks and Veo 3 for video generation. 

Subscribers also receive 30 terabytes of cloud storage, an ad-free YouTube Premium membership, and early access to experimental features such as Project Mariner, an AI assistant capable of managing multiple tasks simultaneously.  

Google also introduced Flow, an AI-powered filmmaking tool that enables video creation through text prompts. Ultra subscribers gain enhanced capabilities for this tool, including 1080p video output and advanced camera controls. Another tool, Whisk, which generates visual content and short animations, will be available with priority access for subscribers.  

Existing Google AI Pro subscribers will receive some new features at no additional cost, including early access to Gemini integration in Chrome and basic Flow functionality with Veo 2. Google plans to expand the Ultra service to additional markets following its US launch.  

AI Mode in Google Search

Google introduced a new AI Mode for Search, which provides AI-generated answers instead of traditional results for complex queries. The feature can process various requests, from analysing images to completing purchases, and may eventually handle tasks like making calls or creating study materials, says Google.

As per an official blog post by the company, AI Mode uses Google's query 'fan-out technology' to break down questions and run multiple searches simultaneously. It also allows follow-up questions and includes web links. The system runs on a customised version of Gemini 2.5.

The feature will appear as a new tab in Search and the Google app, rolling out first in the US. Google plans to gradually incorporate AI Mode features into regular Search based on user feedback. Traditional web results will remain available alongside the AI responses.

Android XR and smart glasses

Google introduced Android XR, a new operating system for extended reality devices developed in collaboration with Samsung and Qualcomm. According to Google, Android XR is the first Android platform built in the Gemini era and supports various hardware, including headsets and smart glasses.  

The system integrates the Gemini assistant to provide voice-based support and contextual awareness in mixed-reality environments, with features like real-time translation and object recognition. Google also announced AI-powered smart glasses based on Android XR, to be developed in partnership with eyeglass brands Warby Parker and Gentle Monster.  

Other announcements

Other conference highlights included updates to NotebookLM, Google's AI research assistant, and expanded integration of Gemini across productivity applications such as Gmail and Docs. The Chrome browser will also incorporate Gemini functionality to analyse webpage content.

Google Beam, also announced at the conference, is a new 3D video calling platform that converts 2D video into volumetric footage, simulating in-person communication with AI-rendered depth and facial cues.

The company also introduced SynthID Detector, a tool for verifying AI-generated content by detecting embedded watermarks in text, images, audio and video. While still in early testing, Google said it has already watermarked over 10 billion pieces of content using SynthID.

Gemini Code Assist, a new coding assistant powered by Gemini 2.5, offers workflow customisation, unit test generation and code standard enforcement. It integrates with GitHub and supports large-scale codebase analysis.

Jules, currently in public beta, is a newly announced autonomous coding agent that operates within a secure Google Cloud virtual machine. It is capable of writing tests, fixing bugs and updating dependencies. Users can preview changes before implementation.