Everything Google Announced At I/O 2026

Google I/O 2026 was unusually coherent for a company with this many products. The main thread was not another isolated model launch, but Google trying to make Gemini the agent layer across Search, Android, Workspace, YouTube, shopping, developer tools and cloud infrastructure.

The caveat is availability. Many of the most ambitious features are limited by geography, subscription tier, trusted tester status or a summer rollout window. Google announced a great deal on 19 and 20 May, but it did not make all of it available to every user on keynote day.

Gemini 3.5 and Gemini Omni

The biggest model announcement was Gemini 3.5 Flash, the first model in a new 3.5 series. Google describes it as a model built for “frontier intelligence with action”, with an emphasis on coding, long-horizon agentic work and lower-latency execution. It is generally available through Google Antigravity, the Gemini API in Google AI Studio, Android Studio, Gemini Enterprise Agent Platform and Gemini Enterprise, and it is also now the default model in the Gemini app and AI Mode in Search.

Google’s own benchmark claims are specific. The company says Gemini 3.5 Flash outperforms Gemini 3.1 Pro on Terminal-Bench 2.1, GDPval-AA and MCP Atlas, reaches 84.2% on CharXiv Reasoning, and runs four times faster than other frontier models when measured by output tokens per second. Gemini 3.5 Pro was described as already in internal use, with a rollout planned for June 2026.

The second model announcement was Gemini Omni, a generative media model that combines Gemini reasoning with video creation and editing. The first release is Gemini Omni Flash, which can take text, images, video and audio as input and generate video output. Google says image and audio output modalities will follow later. Omni Flash is rolling out to Google AI Plus, Pro and Ultra subscribers in the Gemini app and Google Flow, and at no cost in YouTube Shorts Remix and the YouTube Create app. API access for developers and enterprise customers is due in the coming weeks.

Omni is also where Google’s safety and provenance push shows up most clearly. Videos created with Omni include SynthID watermarks, and Google says users will be able to verify generated media through Gemini, Chrome and Search. Google is still testing broader audio and speech editing, which is a sensible constraint given the obvious impersonation risk.

Gemini App, Search And Shopping Agents

Google announced Gemini Spark as a 24/7 personal AI agent in the Gemini app. Spark runs on Gemini 3.5 and Google’s Antigravity harness, works in the background, and is designed to ask before taking major actions. Google says it will begin with trusted testers and bring a beta to Google AI Ultra subscribers in the US. The roadmap includes email and text access, custom subagents and payment authorisation with budgets and merchant constraints.

Daily Brief is the more immediate Gemini app agent. It builds a morning digest from connected Google apps such as Gmail, Calendar and tasks, prioritises the day and suggests next steps. It is rolling out first to Google AI subscribers in the US who have chosen to connect their Google apps. Google also introduced a redesigned Gemini interface called Neural Expressive, with inline Gemini Live, richer generated layouts and a Gemini Live model designed to be faster and less distracted by background noise.

Search received one of the largest changes. AI Mode has passed one billion monthly users, according to Google, and Gemini 3.5 Flash is now the default model in AI Mode globally. Google also announced an AI-powered Search box that expands for longer prompts, supports text, images, files, videos and Chrome tabs as inputs, and offers more detailed AI suggestions than autocomplete.

The more consequential Search work is agentic. Information agents will monitor the web and Google’s fresh data sources for user-defined topics, then send synthesised updates. Those launch first for Google AI Pro and Ultra subscribers this summer. Search is also gaining expanded booking capabilities, including local experiences and services, and in select categories Google will be able to call businesses on a user’s behalf in the US.

Google is also putting Antigravity-style coding into Search. It says Search will be able to generate interactive layouts, simulations, tables, graphs and custom dashboards in response to queries. Generative UI for one-off questions is planned for everyone in Search this summer, while persistent mini apps and trackers will start with Google AI Pro and Ultra subscribers in the US.

Shopping moved in the same direction. Universal Cart is a cross-Google shopping cart for Search, Gemini, YouTube and Gmail. It can watch for price drops, stock availability and payment perks, and use reasoning to spot issues such as incompatible PC parts. Checkout is tied to the Universal Commerce Protocol and Google Pay, with selected checkout features planned across merchants including Nike, Sephora, Target, Ulta Beauty, Walmart, Wayfair and Shopify merchants. Universal Cart rolls out across Search and Gemini in the US this summer, with YouTube and Gmail to follow.

Google also described AP2, the Agent Payments Protocol, as a way to let agents make purchases under explicit boundaries. The important detail is the “digital mandate”: a verifiable record linking the user, merchant and payment processor. Google says AP2 will begin coming to Google products in the coming months, starting with Gemini Spark.

Workspace, YouTube And Creative Tools

Workspace announcements were concentrated around voice, inbox management and image creation. Gmail Live will let users ask voice questions about their inbox. Docs Live will turn spoken thoughts into structured drafts, with permission to pull in relevant material from Gmail, Drive, Chat and the web. Keep will gain a voice-led capture flow that turns rambling notes into organised lists. These features are due this summer for Google AI Pro and Ultra subscribers and in preview for Workspace business customers.

Google Pics is the new image creation and editing app for Workspace. It is built on the Nano Banana model and supports object segmentation, text editing, translation and Workspace integrations. Google is launching it first to trusted testers, with a global rollout this summer for Google AI Pro and Ultra subscribers and a preview for Workspace business customers.

AI Inbox, already introduced earlier in the year, is expanding from Google AI Ultra to Google AI Plus and Pro subscribers in the US. It now generates contextual draft replies, surfaces relevant Docs, Sheets and Slides next to tasks, and lets users mark suggested tasks as done or dismiss them.

Google Flow also became more agentic. Gemini Omni Flash is available in Flow for Google AI subscribers globally, and Google Flow Agent can now plan multi-step creative tasks, generate options, batch edit assets and organise project material. Google Flow Tools lets users create bespoke creative tools in natural language, such as video effects, animation helpers or shaders. Pomelli and Stitch also received updates for brand content, website design and real-time layout steering.

YouTube announced Ask YouTube, a conversational search experience that compiles relevant long-form videos and Shorts into structured answers and supports follow-up questions. It is currently available for YouTube Premium members aged 18 and over in the US through youtube.com/new, with broader rollout planned. YouTube also added Gemini Omni to Shorts Remix and the YouTube Create app, with watermarks, metadata, links back to original videos and creator opt-out controls.

Android, Chrome And Developer Tools

For Android developers, the clearest shift is that Google wants agents to work with the official toolchain rather than guess their way through mobile development. Android CLI is now stable and gives agents such as Claude Code, Codex and Antigravity programmatic access to Android tasks. Google says it can use Android Studio commands for semantic symbol resolution, warnings analysis and Jetpack Compose previews, and it supports Journeys for end-to-end UI tests under developer direction.

Google AI Studio can now build native Android apps from prompts using Kotlin, Jetpack Compose and Google’s recommended patterns. It includes an embedded Android Emulator, physical-device testing through ADB and Play Console integration for publishing to the internal test track. Developers can export projects to Antigravity for local development with the project context preserved.

Android 17 was also part of the I/O developer story. Google highlighted memory and performance tools, a lock-free MessageQueue, more frequent but less intensive young-generation garbage collection, a new contact picker, an eyedropper API, background audio hardening, SMS OTP protection, mandatory large-screen resizability, certificate transparency by default and restricted local network access. Beta testing is available now through enrolled devices and Android 17 emulator images.

Media and large-screen work also received attention. Android 17 adds CameraXViewfinder Composable for high-fidelity capture on foldables and tablets, Media3 AI Effects for features such as Magic Eraser and Studio Sound, CodecDB for chipset-specific encoding recommendations, ExoPlayer Scrubbing Mode and Google TV pointer remote support. Google Play is adding Play Shorts for store discovery, app discovery in Gemini and agentic catalog management in Play Console.

On the web, Chrome’s I/O update focused on local AI APIs and agent-friendly interfaces. The Prompt API is stable in Chrome 148, using Gemini Nano with multimodal inputs and structured output. Google also announced the Gemma 197M model for task-specific APIs, HTML-in-Canvas in origin trial, element-scoped view transitions, and Chrome DevTools for agents so coding tools can inspect, verify and debug browser behaviour in real time.

Google Antigravity became the centre of the developer platform. Antigravity 2.0 is a standalone desktop app for orchestrating multiple agents in parallel, with dynamic subagents, scheduled tasks and integrations across Google AI Studio, Android and Firebase. Antigravity CLI replaces the older Gemini CLI path for terminal users, while the Antigravity SDK exposes the same agent harness for custom agents hosted on user infrastructure.

Managed Agents in the Gemini API are Google’s higher-level agent runtime. A single API call can provision an isolated Linux environment where an agent can reason, use tools, execute code, browse the web and preserve files and state across interactions. Developers can define custom agents with markdown instruction and skill files, which makes Google’s agent stack look increasingly like an infrastructure product rather than a coding assistant feature.

Cloud, Research, XR And Provenance

For enterprise customers, Google Cloud framed the I/O announcements around Gemini 3.5 Flash, Gemini Omni, Gemini Enterprise and Antigravity. Gemini 3.5 Flash is available in Gemini Enterprise, while Omni Flash will roll out to developers and enterprise customers through the Gemini API and Agent Platform API in the coming weeks. Google Cloud customers can connect Antigravity directly to Cloud projects under Google Cloud terms, with customer data kept within the customer’s cloud boundary by default.

Google also used I/O to expand Gemini for Science. The new Labs collection includes Hypothesis Generation, Computational Discovery and Literature Insights. Science Skills connect agentic platforms such as Antigravity to more than 30 life-science databases and tools, including UniProt, AlphaFold Database, AlphaGenome API and InterPro. Access to the experiments is gradual from 19 May, with registration through Labs.

Android XR had its first concrete eyewear roadmap. Google said intelligent eyewear will come in two forms: audio glasses with spoken help and display glasses with information in view. Audio glasses are first, coming later this autumn through partnerships with Samsung, Gentle Monster and Warby Parker, and will pair with Android and iOS devices. Announced features include asking Gemini about what the wearer sees, turn-by-turn navigation, calls and texts, photo and video capture, translation, app actions and task help such as preparing a food order pending final confirmation.

Finally, Google widened its content provenance work. SynthID has watermarked more than 100 billion images and videos and 60,000 years of audio, according to Google. SynthID verification for image, video and audio has been added to the Gemini app and is expanding to Search and Chrome. C2PA Content Credentials verification is rolling out in Gemini and is due for Search and Chrome in the coming months. Google also said OpenAI, Kakao and ElevenLabs will adopt SynthID for more of their AI-generated content, and it is launching an AI Content Detection API on Google Cloud’s Gemini Enterprise Agent Platform with trusted partners.

Taken together, I/O 2026 shows Google consolidating its AI work around one stack: Gemini models, Antigravity as the agent harness, Google products as distribution, Cloud as the enterprise control plane, and SynthID or C2PA as the provenance layer. The announcements are broad, but the engineering question is narrower. Google now has to prove that these agents can work reliably inside real user permissions, real enterprise boundaries and real product surfaces without turning supervision into another full-time job.

---

Published: 2026-05-30 - Sources: Google I/O 2026 collection, 100 things announced at I/O 2026, Sundar Pichai keynote transcript, Gemini 3.5, Gemini Omni, AI Search, Universal Cart, Workspace updates, developer highlights, Android developers recap, Android XR eyewear, Google Cloud I/O update, Chrome at I/O 2026, YouTube I/O update, content provenance update.