Google IO 2025 AI Innovations Unveiled

Image credit: Maria Diaz/ZDNET

Google I/O, a landmark event for developers, recently concluded, and the message was clear: AI is at the forefront of Google's strategy. While the tech giant continues to build on its foundations in Android, Chrome, Search, and Workspaces, this year's conference heavily emphasized advancements in artificial intelligence, promising to reshape how we interact with technology.

Millions of developers worldwide tune into Google I/O as it dictates the future roadmap for Google's vast ecosystem. This year was no exception, with a slew of announcements impacting everything from core AI models to user-facing applications. For more perspectives, CNET also covered the latest AI upgrades from the event.

Let's dive into the most significant news from Google I/O 2025, exploring upgrades to Gemini, new AI-driven services, and innovations that could transform communication and productivity.

Gemini AI Models Evolve Further

Ahead of I/O 2025, Google had already released Gemini 2.5 Pro to positive feedback from beta testers. Building on this, Google announced several enhancements for both Gemini 2.5 Pro and Flash models. Gemini 2.5 Pro now features "Deep Think" support, enabling the AI to perform high-level research by considering multiple hypotheses before generating a response.

According to Google's evaluations, Gemini 2.5 Flash has seen improvements across reasoning, multimodality, coding, and response efficiency, now requiring 20% to 30% fewer tokens. Furthermore, Gemini 2.5 models will support audio-visual input and native audio output dialogue through a preview version in the Live API. This allows developers to customize the tone, accent, and speaking style for more natural conversational AI experiences. ZDNET highlighted some of the best AI features and tools from the event.

Building a "World Model" with Advanced Agentic AI

Google is laying the groundwork for what it calls a "world model" – an AI with a deep understanding of the world. Examples include Veo, an AI video generator with an understanding of physics, and Gemini Robotics, which enables robots to adapt to various environments.

Leveraging Gemini 2.5 Pro, Google aims to create an AI that understands context, plans, and acts on a user's behalf across any device. We've seen early demonstrations of this through Gemini Live, which incorporates aspects of Project Astra, a system designed for contextual understanding and assistance. One particularly cool feature of Gemini Live is now available for free to all Android and iOS users.

Taking this a step further, Google introduced Project Mariner, a browser-based agentic AI capable of handling up to 10 tasks simultaneously, such as booking flights, researching, and shopping. The latest research prototype will first be available to Google AI Ultra subscribers in the US.

New Google AI Subscription Tiers Unveiled

Google is introducing two new subscription plans for its AI features: Google AI Pro and Google AI Ultra. The Pro plan offers access to Google's full suite of AI products, including the Gemini app (with Advanced capabilities), Flow, and NotebookLM, with higher rate limits and special features compared to free versions.

The Google AI Ultra plan, detailed further by ZDNET regarding what's included in the $250 subscription, provides the highest rate limits and early access to experimental AI products like Project Mariner, Veo 3, and Gemini 2.5 Pro with Deep Think mode. This premium subscription costs $250 per month (with a 50% discount for the first three months for new users) and includes early access to Agent Mode. Agent Mode allows users to issue prompts at a desktop level for tasks like live web browsing and research, with Google's AI models handling the execution.

Innovations in Generative Media Models

Google showcased new versions of its generative media models: Veo 3, Imagen 4, and a new tool called Flow. Veo 3 now supports audio prompt generation, allowing it to create sounds like traffic noise or dialogue between characters, adding a new dimension to AI-generated videos that were previously silent.

To give creators more control, Google introduced Flow, an AI filmmaking tool. Flow enables adjustments to video production aspects such as camera angles, motion, cast, and location. Additionally, Imagen 4 has been upgraded for better accuracy and clarity, especially with fine details like fabric textures, water droplets, and animal fur. It can now produce images in various aspect ratios and up to 2K resolution. For those interested in AI tools on the go, Google's NotebookLM now has its own Android app.

AI Transforms the Google Search Experience

AI Mode in Search, introduced earlier this year, is now becoming widely available in the US without requiring a Labs sign-up, and it comes with new capabilities. "Deep Search" in AI Mode expands background queries significantly to compile more robust and well-researched responses, delivering fully cited reports in minutes. Indeed, your Google Search experience is set for a major change with new AI features.

For visual queries, Google Search will integrate Project Astra's multimodal capabilities, allowing users to point their camera at an object and ask questions, similar to Google Image Search but more interactive. A new AI Mode shopping experience will help users find inspiration, narrow down choices, and even visualize themselves in an outfit using an image generator by uploading a photo.

Project Mariner's agentic functions will also be incorporated into AI Mode, enabling users to prompt Google to find event ticket deals or book restaurant appointments. While the AI won't complete purchases, it will present options for user approval. Lastly, AI Overviews have been expanded to over 200 countries and territories and support more than 40 languages.

AI Enhancements for Google Workspace

Google's AI integration extends deeply into its Workspace services, including Gmail, Meet, and Docs. Gmail will feature personalized smart replies that adapt to your communication style with specific contacts or in particular threads, generating responses that match the tone and context. An inbox cleanup feature will allow users to prompt Gemini to delete emails from a specific sender within a set timeframe. ZDNET has detailed 7 new AI features coming to Google Workspace and how they can improve daily workflows.

Gemini will also proactively suggest meeting booking windows in Gmail if it detects an attempt to schedule a meeting, streamlining the process. Google Meet is introducing near-real-time speech translation, converting spoken words into the listener's preferred language, a beta feature rolling out first to Google AI Pro and Ultra subscribers. In Docs, users can enable source-grounded writing assistance, ensuring Gemini only pulls information from specified source links, a feature expected to be generally available next quarter.

Introducing Google Beam for Immersive Video Conferencing

Project Starline, which previously impressed with its ability to turn 2D video calls into lifelike 3D simulations without bulky hardware, has been officially rebranded as Google Beam. Google Beam uses AI to transform standard 2D video into more realistic 3D experiences, akin to the spatial video processing seen from Apple and Meta.

The objective is to create a communication platform that fosters trust and deeper understanding through realistic body movements, cues, and eye contact. As ZDNET notes, Google Beam is poised to bring 3D video conferencing mainstream. To aid its enterprise expansion, Google is partnering with HP, which recently made news by acquiring Humane for its AI hardware expertise, to create communication devices running Google Beam.

The Future is Wearable with Android XR Smart Glasses

While Android XR was initially unveiled last December with a focus on XR and VR headsets, Google I/O 2025 shifted attention to Android XR glasses. These everyday wearables will use cameras, microphones, and speakers to interpret the user's surroundings and provide assistance via Gemini. This vision for wearables is shared by other tech giants like Meta and Apple, but Google's ecosystem integration could give it an edge. One promising example in this space is Xreal's Project Aura, which looks like the Google smart glasses many have awaited.

Google demonstrated how Android XR's Gemini capabilities, combined with a configurable in-lens display, can provide a directional system for navigating city streets, much like a car's HUD. Users will also be able to visualize and respond to text messages, translate conversations in real time, and capture photos using voice commands, minimizing the need to pull out a phone.

To appeal to a broader audience, Google is collaborating with popular eyewear brands such as Gentle Monster and Warby Parker to develop more stylish smart glasses. Google expects these Android XR glasses to be available later this year.

These announcements from Google I/O 2025 paint a clear picture of a future deeply intertwined with artificial intelligence, impacting nearly every facet of Google's offerings. For more insights into AI developments, consider exploring further resources on innovation in technology.