Google I/O 2026: Every AI Announcement, Feature, Product Launch and Where to Try It

Google I/O 2026 was the week Google stopped presenting AI as a feature and started presenting it as an operating layer.

The event ran May 19 and May 20, 2026, with the main keynote, developer keynote, and on-demand sessions hosted at io.google/2026. Google also published its official recap, 100 things we announced at I/O 2026, and the keynote video is available on YouTube at Google I/O 2026 keynote.

The headline is not simply that Gemini got better. The real story is that Google is wiring Gemini into Search, Workspace, Gmail, Android, Chrome, YouTube, Google Flow, science tools, shopping, and developer platforms. The language shifted from "ask an AI" to "delegate work to agents." That is the important thread running through almost every announcement.

This guide is written as a factual field map. It lists the major product launches, the new capabilities, where Google says each feature is available, and where readers can go to try the relevant product.

Quick product links

Product or capability	What was announced	Where to try or learn more
Google I/O 2026	Keynotes, sessions, workshops and official recaps	io.google/2026
Gemini app	Gemini 3.5 Flash, Gemini Omni Flash, Daily Brief, Gemini Spark, Neural Expressive redesign and improved Gemini Live	gemini.google.com
Gemini Spark	Personal AI agent for delegated tasks, rolling first to trusted testers and Google AI Ultra users in the U.S.	Gemini Spark
Personal Intelligence	Opt-in connection to apps like Gmail, Google Photos and soon Calendar	Personal Intelligence
Google Search AI Mode	Gemini 3.5 Flash default model, seamless AI Overviews to AI Mode, information agents and generative UI	Google Search
Google Shopping and Universal Cart	Intelligent cart across Search, Gemini, YouTube and Gmail, with Google Pay checkout support	Google Shopping
Google Antigravity	Antigravity 2.0 desktop app, Antigravity CLI, SDK, subagents, hooks and async task management	antigravity.google
Google AI Studio	Workspace integrations, Android app building, Play Console test-track publishing and Cloud Run deploys	aistudio.google.com
Gemini API and Managed Agents	Remote sandboxed agents through the Gemini API and Interactions API	ai.google.dev
Firebase Studio	App development and deployment workflows connected to Google AI Studio	Firebase Studio
Android Studio	Gemini 3.5 Flash availability, Android CLI, Android skills, Android Bench and migration agent previews	Android Studio
Chrome DevTools for agents	Agent-visible debugging, verification and optimization workflows	Chrome for Developers
WebMCP	Proposed open web standard for exposing structured web tools to browser agents	WebMCP
Gmail and AI Inbox	AI Inbox expansion, contextual draft replies and Gmail Live later this summer	Gmail
Google Docs	Docs Live voice-based document creation and editing	Google Docs
Google Keep	Talk to Keep for turning spoken thoughts into notes and lists	Google Keep
Google Flow	Gemini Omni Flash, Flow Agent, Flow Tools and expanded creative workflows	Google Flow
Google Flow Music	Conversational music-video creation and finer song editing controls	Google Labs
NotebookLM	Literature Insights in Gemini for Science, plus NotebookLM as the source-grounded research product	notebooklm.google
Gemini for Science	Hypothesis Generation, Computational Discovery, Literature Insights and Science Skills	labs.google/science
YouTube	Ask YouTube conversational search, YouTube Shorts Remix with Gemini Omni and YouTube Create access	youtube.com
Android XR	Intelligent eyewear, including audio glasses with Gentle Monster, Warby Parker and Samsung	Android XR
SynthID	AI-content verification in Gemini, Search and Chrome, plus C2PA Content Credentials	SynthID

The operating map

graph TD
    A["Gemini 3.5 Flash"] --> B["Agents and Search"]
    A["Gemini 3.5 Flash"] --> C["Antigravity and AI Studio"]
    D["Gemini Omni"] --> E["Gemini app and Flow"]
    D["Gemini Omni"] --> F["YouTube Shorts Remix"]
    B["Agents and Search"] --> G["Personal Intelligence and Universal Cart"]
    C["Antigravity and AI Studio"] --> H["Managed Agents and Android tools"]
    E["Gemini app and Flow"] --> I["Daily Brief Spark creative tools"]
    G["Personal Intelligence and Universal Cart"] --> J["Consumer task completion"]
    H["Managed Agents and Android tools"] --> K["Developer task completion"]

What Google I/O 2026 was really about

The center of gravity at I/O 2026 was action. Google used the event to argue that AI is moving from answering questions to performing work across apps, files, browsers, stores, research workflows and codebases.

That is why the announcements arrived in clusters. Gemini 3.5 Flash is the fast model layer. Gemini Omni is the multimodal creative layer. Search agents and Gemini Spark are the consumer agent layer. Antigravity, AI Studio and Managed Agents are the developer agent layer. Workspace, Gmail, Docs, Keep and Flow are the productivity and creation layer. Gemini for Science is the research layer. SynthID and C2PA support are the provenance layer.

The pattern matters because it tells users what Google believes the next interface will be. Search is no longer only a search box. Gemini is no longer only a chatbot. A development environment is no longer only a code editor. A creative studio is no longer only a blank canvas. Each surface is being rebuilt so a model can reason across context, call tools, create artifacts and hand back something usable.

For readers deciding what to try first, the practical split is simple:

If you want personal productivity, start with Gemini, Daily Brief and Personal Intelligence.
If you want research help, start with NotebookLM and labs.google/science.
If you want creative media, start with Google Flow, Gemini Omni and YouTube Shorts Remix.
If you want developer productivity, start with Google AI Studio, Antigravity and the Gemini API.
If you want search and commerce, watch AI Mode, information agents and Universal Cart inside Google Search and Gemini.

Gemini 3.5 Flash

Google launched Gemini 3.5 Flash as the first model in its newest Gemini 3.5 family. The positioning is direct: frontier-level intelligence with the speed expected from the Flash line.

According to Google, Gemini 3.5 Flash is generally available through Google Antigravity, the Gemini API in Google AI Studio, and Android Studio. Google said it outperforms Gemini 3.1 Pro on several coding and agentic benchmarks, including Terminal-Bench 2.1, GDPval-AA and MCP Atlas. The important product claim is not only higher scores. Google is saying 3.5 Flash is fast enough and capable enough to power long-horizon agent tasks, coding loops, document work and interactive web UI generation without forcing the old tradeoff between model quality and latency.

Gemini 3.5 Flash showed up everywhere else in the event. It became the default model for AI Mode in Search. It powers Gemini Spark. It is co-optimized with the Antigravity agent harness. It sits behind Managed Agents in the Gemini API. In other words, Google is not treating it as one model for one app. It is treating it as an action engine across its surfaces.

Google also said Gemini 3.5 Pro is being used internally and is expected to roll out next month. That was a roadmap note, not a generally available launch during I/O week.

Gemini Omni

Gemini Omni was Google's big multimodal creation announcement. Google described it as a model that can create anything from any input, starting with video. In practical terms, the launch begins with video generation and editing, but the longer-term ambition is broader: take images, text, video or audio as references and produce a cohesive output.

The technical claim Google emphasized is world understanding. Gemini Omni is meant to understand physics, motion, culture, history and scene logic well enough to make generated video feel more coherent. Google specifically pointed to better handling of gravity, kinetic energy and fluid dynamics. That matters because generative video often fails when motion, object permanence or scene consistency becomes too demanding.

Where users can try it:

Gemini app for conversational video creation and editing.
Google Flow for creative workflows and scene iteration.
YouTube Shorts Remix and YouTube Create for eligible users and creators.

Google said Gemini Omni Flash is rolling out globally to Google AI Plus, Pro and Ultra subscribers through the Gemini app and Google Flow. It is also available in YouTube Shorts Remix and YouTube Create to users 18 and older at no cost.

The most consumer-friendly capability is conversational editing. Google gave examples like applying cinematic zooms, changing backgrounds, using camera-roll photos or videos, applying templates and creating a custom AI avatar that looks and sounds like the user. In Flow, the same model is designed to preserve character identity and voice across scenes, which is a critical feature for creators making multi-scene stories.

AI Mode and the new Google Search box

Search had one of the biggest strategic upgrades at I/O 2026. Google said AI Mode has surpassed more than 1 billion monthly users, and Gemini 3.5 Flash is now the default model for AI Mode globally.

The bigger announcement is the reimagined Search box. Google said users will be able to search using text, images, files, videos and Chrome tabs, with Search reasoning across them all. That means Search is moving from a query box toward a multimodal task surface. It still returns links and results, but the workflow now expects follow-up, synthesis, comparison and action.

Google also said it is bringing AI Overviews and AI Mode into one seamless AI Search experience. A user can start with a normal search, get an AI Overview, and continue the conversation in AI Mode with links to learn more.

The new feature set includes:

Gemini 3.5 Flash as the default AI Mode model.
Text, image, file, video and Chrome-tab input.
A more seamless path from AI Overviews to AI Mode.
Information agents that monitor topics or tasks in the background.
Generative UI that creates custom layouts, tables, charts, simulations and mini app-like experiences.

The practical takeaway is that Search is becoming both a retrieval product and an interface generator. For a complicated question, Google wants Search to produce not just an answer, but the right shape for that answer.

Information agents in Search

Google introduced information agents as the first category of Search agents. These are background agents that monitor web sources, news, blogs, social posts, and real-time data such as finance, shopping and sports.

The use case is continuous awareness. Instead of repeatedly searching the same topic, a user can create an agent that watches for changes and sends synthesized updates. Google says users will be able to run multiple information agents in Search at the same time.

Availability is staged. Google said information agents will roll out this summer, first to Google AI Pro and Ultra subscribers.

The product is worth watching because it changes Search from a moment-in-time query into an ongoing subscription to a question. That creates a very different user habit. Instead of "find this for me now," the interface becomes "keep an eye on this until something important changes."

Generative UI and mini apps in Search

Google also announced generative UI in Search, powered by Antigravity and Gemini 3.5 Flash. The idea is that Search can build a custom format for the question, not just show a static answer.

Examples include:

Interactive visuals.
Tables.
Graphs.
Simulations.
Custom layouts.
Ongoing trackers and dashboards.

Google said generative UI with Antigravity is rolling out to Search this summer for everyone, free of charge. More persistent custom experiences, described as mini apps for ongoing tasks like planning a wedding or managing a move, are expected in the coming months, starting with subscribers.

This is one of the most consequential announcements because it attacks a long-standing limitation of search results. Sometimes the right answer is not a blue link or a paragraph. Sometimes the right answer is a tracker, a calculator, a comparison matrix, or a simulation. Google is now saying Search can generate those formats on demand.

Personal Intelligence

Personal Intelligence is Google's opt-in layer for using personal context in AI Mode and Gemini. At I/O 2026, Google said it is expanding Personal Intelligence in AI Mode to nearly 200 countries and territories across 98 languages with no subscription required.

The key integrations are Gmail and Google Photos, with Google Calendar coming soon. Google emphasized user control: people choose whether to connect apps, and Personal Intelligence is designed around transparency, choice and control.

This is the feature family that makes Google's ecosystem advantage obvious. If a user opts in, Gemini can reason with email, photos, calendar context and tasks. That turns the assistant from a generic answer engine into something closer to a personal operating layer.

The upside is relevance. The risk is trust. Google knows this, which is why the keynote language repeatedly leaned on permission and user control.

Universal Cart and Universal Commerce Protocol

Google introduced Universal Cart as an intelligent shopping cart across Google surfaces. The cart is meant to follow the user across Search, Gemini, YouTube and eventually Gmail.

The capabilities Google listed are more ambitious than saving items:

Add products while browsing Search, chatting with Gemini, watching YouTube or reading Gmail.
Find deals and price drops in the background.
Show price-history insights.
Alert users when an item comes back in stock.
Proactively flag product incompatibilities.
Suggest alternatives.
Use Google Wallet information such as payment perks, loyalty details and merchant offers.

Google also announced the Universal Commerce Protocol for smoother checkout. The idea is that users can check out from Google with Google Pay for participating brands, or transfer items to a retailer site.

Universal Cart rolls out across Search and Gemini this summer, with YouTube and Gmail to follow.

This is agentic commerce with a Google flavor. It is not only recommending what to buy. It is monitoring, comparing, warning, and moving the checkout path closer to the interface where intent appears.

Gemini Spark

Gemini Spark was one of the event's clearest "agent era" announcements. Google described it as a 24/7 personal AI agent that helps navigate digital life, takes action on behalf of the user and remains under user direction.

The most important detail is that Spark can work in the background on a phone or laptop even when the device is turned off. It runs on Gemini 3.5 and is built on the Google Antigravity platform.

Google is being careful with the rollout. Spark is early, and Google said it is prioritizing safety in the first release. It is rolling out to trusted testers, with a beta planned for Google AI Ultra subscribers in the United States.

Google also described future Spark capabilities:

Text or email Spark directly.
Create custom subagents.
Authorize payments with specified budgets and merchants.
Delegate multi-app personal tasks.

For normal users, Spark is the clearest expression of Google's agent strategy. It is not a developer tool. It is a personal delegation surface. The big question will be whether Google can make it helpful without making users nervous about what it can do.

Daily Brief

Daily Brief is a new out-of-the-box agent in the Gemini app. Its job is to organize and prioritize the user's day with a personalized digest based on goals, calendar, tasks and inbox context.

Google says Daily Brief works overnight, analyzes inbox, calendar and task signals, connects the dots, suggests next steps and learns preferences over time. It is rolling out to Google AI subscribers 18 and older in the Gemini app, starting in the United States. Users must choose to connect their Google apps.

This is a quieter announcement than Gemini Omni or Spark, but it may be one of the stickiest. A good daily brief is habit-forming. If it can reliably surface what matters and reduce morning triage, it becomes a daily interface into the broader Gemini ecosystem.

Neural Expressive and Gemini Live

Google also redesigned the Gemini app with a new design language called Neural Expressive. Google described it as a ground-up redesign with fluid animations, vibrant colors, new typography and haptic feedback.

More important than the look is the change in answer presentation. Google said Gemini will no longer simply return a wall of text. It can lay out responses in real time with interactive images, timelines and embedded visuals.

Gemini Live also changed. It now opens immediately and inline, uses a smarter and faster model, and is designed to be less distracted by background noise. Google also said regional dialect selection is coming.

The practical product point is that Gemini is becoming more spatial and interactive. It is no longer just a chat transcript. It is moving toward generated interfaces, visual explanations and live multimodal help.

Google Antigravity

Google Antigravity was the developer centerpiece. Google described it as an agent-first development platform, and I/O 2026 expanded it into several surfaces:

Antigravity 2.0, a standalone desktop app.
Antigravity CLI, a terminal-first interface.
Antigravity SDK, programmatic access to the agent harness.
Subagents.
Hooks.
Asynchronous task management.
Native voice support for Gemini audio models.
Integrations with Android, Firebase and Google AI Studio.

Google says Antigravity 2.0 can orchestrate multiple agents in parallel. One agent might code a website while another generates brand assets. The CLI gives terminal users a lightweight way to create agents without a graphical interface. The SDK gives teams control over the harness so they can customize behavior and host agents on their own infrastructure.

For enterprises, Google said Antigravity can connect directly to Google Cloud projects under expected enterprise terms. Gemini Enterprise customers are expected to see Antigravity rolled out in the coming months.

This is Google's answer to the new coding-agent market. The platform is not framed as a code completion tool. It is framed as a multi-agent workbench where agents produce artifacts, coordinate, and operate across a project.

Google AI Studio

Google AI Studio got a major I/O upgrade. Google is turning it into a place where an idea can become a working app, with deployment paths and Android support built in.

The announced capabilities include:

A coming Google AI Studio mobile app for capturing ideas on the go.
Google Workspace access inside apps built with AI Studio.
Native Android app building from the build tab.
Google Play Console support for internal test-track publishing.
Browser-based Android Emulator previews.
Android Debug Bridge support for installing on test devices.
First two Google Cloud app deploys at no cost for new builders.
Export from AI Studio to Antigravity with conversation history, project files and secrets.
Custom images generated by Nano Banana inside the AI Studio Build agent.
Preview-window annotation and editing.

The Workflow story is clean: prototype in AI Studio, deploy to Cloud Run or Firebase, publish Android tests through Play Console, and export to Antigravity when the project needs deeper local development or multi-agent orchestration.

Managed Agents in the Gemini API

Google introduced Managed Agents in the Gemini API. This is one of the most important developer announcements because it reduces the infrastructure burden of running agents.

With Managed Agents, a single API call can provision an Antigravity agent in a remote Linux environment. The agent can reason, plan, call tools, execute code, manage files in an isolated sandbox and browse the web to fetch and process live data.

Google says Managed Agents are powered by the new Antigravity agent, built with Gemini 3.5 Flash, and available through the Interactions API and in Google AI Studio.

Developers can extend the agent with instructions and skills defined in markdown files such as AGENTS.md and SKILL.md. That detail matters because it makes agent customization feel more like configuring a team process than writing orchestration plumbing from scratch.

Android developer tools

The developer keynote added a separate set of Android-focused announcements. Google said Gemini 3.5 Flash is available in Android Studio and showed several agentic Android tools.

The most important items:

Stable Android CLI for agents.
Open-sourced Android skills for workflows such as Jetpack Compose migration and Jetpack Navigation migration.
Android Bench, an LLM leaderboard for Android development tasks.
Open-weight models such as Gemma 4 added to Android Bench.
A migration agent preview in Android Studio that can migrate app code to native Kotlin Android from React Native, web frameworks or iOS.

Where to start:

The agentic Android story is practical. Mobile development has many environment, SDK, emulator and migration headaches. Google is trying to make those chores accessible to agents through official tools and skills rather than leaving developers to hack together brittle scripts.

WebMCP, Modern Web Guidance and Chrome DevTools for agents

Google's web announcements are aimed at a future where browser-based AI agents need structured ways to interact with websites.

WebMCP is a proposed open web standard that lets developers expose structured tools such as JavaScript functions and HTML forms to browser agents. The experimental origin trial begins in Chrome 149, with support for Gemini in Chrome coming soon.

Modern Web Guidance is an early-preview set of expert-vetted skills for AI coding tools. It supports more than 100 use cases and is designed to help agents build accessible, performant and secure web experiences. It integrates with Baseline and can be installed in Antigravity or through the CLI.

Chrome DevTools for agents gives agents visibility into debugging and optimization. Google says it can help agents verify, debug and optimize code in real time, automate quality audits, emulate user experiences and hand over sessions with auto-connect.

The important direction is this: Google is trying to make the web more legible to agents without making agents guess at every button and form. That is a necessary step if browser agents are going to be reliable enough for serious workflows.

Google AI Ultra and AI Pro subscriptions

Google announced a new $100 Google AI Ultra plan aimed at developers, technical leads, knowledge workers and advanced creators. Google said it includes 5X higher usage limits in Gemini and Antigravity than AI Pro, plus 20TB of cloud storage and other frontier AI features.

Google also said Google AI Pro paid subscriptions now include YouTube Premium Lite individual at no extra charge.

This subscription story matters because many I/O features are gated by plan and geography. Gemini Omni Flash is rolling to Google AI subscribers through Gemini and Flow. Daily Brief starts with Google AI subscribers in the United States. Spark beta starts with Google AI Ultra subscribers in the United States. Information agents start with AI Pro and Ultra subscribers. Google is clearly segmenting casual, power-user and developer access.

Gmail AI Inbox and Gmail Live

Gmail got several agentic productivity updates. Google said AI Inbox is a new view that surfaces what matters most, helps prioritize to-dos and provides updates on important threads.

At I/O 2026, Google said:

AI Inbox is available for Google AI Ultra subscribers and is rolling out to Google AI Plus and Pro subscribers in the United States.
AI Inbox can generate personalized draft replies based on context.
If a task requires reviewing a Google Doc, Sheet or Slide, the link can surface next to the to-do.
Users can mark individual tasks done, dismiss unhelpful suggestions or mark all emails in a topic as read.
Gmail Live is coming this summer for Google AI Pro and Ultra subscribers, letting users talk to their inbox and ask specific queries without digging through threads.

This is a strong example of Google's agent model. The AI is not only writing replies. It is turning email into a task surface.

Google Pics

Google announced Google Pics as a new image creation and editing tool built on Nano Banana. Google positioned it for assets like party flyers and infographics, but the features are broader:

Blank-canvas generation.
Editing existing photos.
Object segmentation.
Precise element selection.
Text editing.
Translation.
Workspace integrations.

Google Pics launched to a limited trusted-tester group at I/O. Google says it will roll out globally this summer to Google AI Pro and Ultra subscribers and in preview for Google Workspace business customers.

This product looks like Google's answer to the everyday design gap: people need quick visual artifacts that are better than a template, but they do not want to become professional designers.

Docs Live and Talk to Keep

Google announced Docs Live for voice-based document creation and editing. The user talks, and Docs Live structures the document, organizes thoughts and, with permission, pulls relevant context from Gmail, Drive, Chat and the web. Google AI Pro and Ultra subscribers will be able to talk to Docs starting this summer.

Google also announced Talk to Keep for Google Keep. The feature turns spoken thoughts into organized notes and lists. It is rolling out this summer for Google AI Pro and Ultra subscribers and in preview for Google Workspace business customers.

The deeper point is that Workspace is becoming less form-bound. Documents and notes are no longer only things users type into. They become outputs from conversation, dictation and connected context.

Google Flow, Flow Agent and Flow Tools

Google Flow moved from AI video tool toward creative studio. Google said Flow is available in more than 140 countries and now includes Gemini Omni Flash for Google AI subscribers globally.

The major new announcements:

Gemini Omni Flash in Flow.
Better character consistency and voice preservation across scenes.
Flow Agent for multi-step creative work.
Flow Tools for creating custom creative tools with natural language.
Global availability of Flow Agent for Flow users.
Existing Flow Tools usable by all Flow users globally.
Tool creation and remixing for Google AI subscribers.

Flow Agent is the big conceptual jump. Until now, Google says Flow could only execute one prompt at a time. The agent can plan and reason through more complex creative tasks. It can brainstorm, create scene variations, batch edit assets, organize assets into collections and rename them.

Flow Tools is also important. It lets users create custom creative utilities in natural language, such as an image editor, video resizer, custom shader or animation workflow. If a user builds something useful, they can share it with other Flow users.

Google Flow Music, Pomelli and Stitch

Google Flow Music extends Gemini Omni into music-video workflows. Google said users can work conversationally with the agent to direct shareable music videos. New refinement controls allow editing specific portions of a song, including changing lyrics into another language, changing genre, adjusting instruments or fine-tuning other details.

Pomelli is adding new ways to build brand content and design websites.

Stitch is becoming more conversational and real-time. Google said users can describe what they want through text or voice, and Stitch can build and reflow ideas. It can also import existing codebases and design files, which helps keep outputs on-brand.

The connective tissue is clear: Google is trying to give creators not just models, but working surfaces where they can steer, revise and package assets.

Gemini for Science

Gemini for Science is Google's new collection of science tools and experiments. It is aimed at daily research tasks such as staying current with papers, turning research goals into usable code and generating hypotheses.

The three experimental tools are:

Hypothesis Generation.
Computational Discovery.
Literature Insights.

Hypothesis Generation is built with Co-Scientist. Google says it simulates the scientific method by collaborating with researchers to define a challenge, then using a multi-agent idea tournament to generate, debate and evaluate hypotheses. Claims are deeply verified and backed by clickable citations.

Computational Discovery is built with AlphaEvolve and Empirical Research Assistance. It generates and scores thousands of code variations in parallel, helping scientists test modeling approaches in fields like solar forecasting and epidemiology.

Literature Insights is built with NotebookLM. It searches scientific literature, structures results into custom searchable tables, enables chat over a curated corpus, and creates artifacts such as reports, slide decks, infographics, audio overviews and video overviews.

Google says access to the three experimental tools began gradually from May 19, and users can register interest at labs.google/science.

Science Skills, Paper Assistant Tool and ScholarPeer

Gemini for Science also includes Science Skills, a specialized bundle that integrates insights from more than 30 life-science databases and tools. Google named UniProt, AlphaFold Database, AlphaGenome API and InterPro as examples.

Science Skills are designed for agent-first platforms like Google Antigravity, where they can help researchers perform workflows such as structural bioinformatics and genomic analysis in minutes instead of hours. Google says Science Skills became available May 19 on GitHub and for all Antigravity users.

Google also announced pilots with scientific conferences including ICML, STOC and NeurIPS for agentic peer review and scientific validation tools, including the experimental Paper Assistant Tool and ScholarPeer.

This section of I/O matters because it shows Google treating scientific workflows as one of the most important targets for agents. The promise is not just faster writing. It is faster hypothesis generation, code exploration, literature review and validation.

NotebookLM

NotebookLM was not presented as a standalone consumer headline in the same way as Gemini Spark, but it sits inside one of the most important research announcements. Literature Insights in Gemini for Science is built with NotebookLM, and its role is to ground analysis in a curated corpus.

For readers, NotebookLM is still the place to try Google's source-grounded research assistant. It supports notebooks built from user-provided sources and can generate grounded summaries, study guides, briefings, audio overviews and video overviews depending on feature availability and account access.

The I/O-specific connection is this: NotebookLM is becoming part of Google's scientific research stack. That matters because source grounding is the difference between a model that sounds convincing and a research tool that lets users inspect where claims come from.

If you want to experiment with the kind of source-grounded workflow Google is pushing, start with notebooklm.google, upload a small set of sources, ask it to produce a comparison table, then inspect the citations. That is the workflow pattern behind Literature Insights.

Ask YouTube

Google announced Ask YouTube as a conversational search experience for video discovery. The feature lets users ask more complex queries, and YouTube compiles relevant videos across long-form content and Shorts into an interactive structured response.

Google's example was a user asking for tips on teaching a child to ride a bike. Instead of returning only a ranked list of videos, Ask YouTube can assemble a more helpful answer from across the YouTube catalog.

Availability is experimental. Google said Ask YouTube begins rolling out this month on desktop to a subset of users searching in English in the United States.

This is YouTube moving closer to Search's AI Mode. The product is no longer just indexing videos. It is trying to synthesize video knowledge into a conversational answer while still pointing users toward the underlying videos.

YouTube Shorts Remix and YouTube Create

YouTube also gets Gemini Omni through Shorts Remix and YouTube Create. Google says users 18 and older can try the new Gemini Omni model at no cost in YouTube Shorts Remix. The upgraded workflow lets users select an eligible Short, describe what they want changed, add themselves or another visual reference, and create a remixed version.

For creators, this is a major distribution play. Gemini Omni is not only inside Gemini and Flow. It is inside the platform where short-form video already lives. That gives Google a way to put generative video into an existing creation habit rather than forcing users to start somewhere new.

Android XR and intelligent eyewear

Android XR returned at I/O 2026 with a focus on intelligent eyewear. Google said there will be two types:

Audio glasses that offer spoken help in the user's ear.
Display glasses that show information when needed.

Google said its first audio glasses, made in partnership with Gentle Monster, Warby Parker and Samsung, will arrive this fall and will be compatible with Android and iOS devices.

The XR announcement is tied to the broader Gemini story. If AI is becoming an action layer, glasses are one of the most natural future surfaces for it. The challenge is not only display hardware. It is whether Gemini can be fast, contextual and restrained enough to help without becoming annoying.

SynthID and provenance

Google expanded SynthID, its imperceptible watermarking technology for AI-generated content. Google said SynthID verification for image, video and audio is already in the Gemini app and has been used 50 million times globally.

At I/O 2026, Google said it is expanding verification to Search and to Chrome over the coming weeks. Users can ask questions such as "Is this made with AI?" through Search features like Lens, AI Mode and Circle to Search, as well as Gemini in Chrome.

Google also announced support for C2PA Content Credentials, which can help users check whether content is an unaltered original from a camera or whether it has been modified and by what tools. C2PA verification is rolling out in the Gemini app first, then coming to Search and Chrome in the coming months.

Google also said companies including OpenAI, Kakao and ElevenLabs are bringing SynthID technology to more AI-generated content.

This is the governance layer of the event. As Google makes creation easier with Omni, Flow, Pics and YouTube remixing, it also needs stronger ways for users to inspect content origin.

What builders should take away

For developers, the I/O 2026 message is that Google wants agents to become normal software infrastructure.

The stack is now fairly clear:

Gemini 3.5 Flash supplies the model speed and reasoning.
Antigravity supplies the agent harness.
AI Studio supplies prototyping, deployment and app-building surfaces.
Managed Agents supply remote sandboxed execution through the API.
Android CLI, Android skills and migration agents make mobile development easier for agents.
WebMCP and Chrome DevTools for agents make browser workflows more structured and debuggable.

The best next step for builders is to pick a specific workflow and test the new agent surfaces against it. Do not start with "build everything." Start with one recurring task: migrate a screen to Jetpack Compose, build a small internal dashboard, generate test cases, create a prototype with Sheets data, or run a remote data-processing agent in a sandbox.

The winners in this tooling wave will be the teams that define agent boundaries clearly. Give the agent a repo, a task, tests and a review path. Keep secrets protected. Keep logs. Treat the generated output like work from a junior teammate with very high stamina and very uneven judgment.

What business leaders should take away

For executives, I/O 2026 is evidence that Google is trying to own the full agent workflow: consumer attention, enterprise productivity, developer tools, commerce, science and creative media.

The practical questions are:

Which Google surfaces already hold your organization's work context?
Which tasks could safely become agent-assisted before they become agent-executed?
Which teams need personal productivity features like Daily Brief or AI Inbox?
Which teams need builder tools like Antigravity, AI Studio and Managed Agents?
Which content teams need Flow, Pics, Stitch or YouTube creation tools?
Which governance standards are needed before agents can connect to real systems?

The mistake would be to treat I/O 2026 as a pile of disconnected feature announcements. The better reading is that Google is building an agentic operating layer across its ecosystem. That layer will be most valuable where Google already has context: email, calendar, files, photos, search intent, shopping intent, code, videos and cloud projects.

What everyday users should try first

If you are a normal user who wants the practical version, start here:

Try Gemini for the redesigned app, Omni access if available on your plan, and Gemini Live improvements.
Try Personal Intelligence if you want Gemini to connect with personal Google context.
Watch for Daily Brief in the Gemini app if you are a Google AI subscriber in the United States.
Use NotebookLM for research grounded in your own sources.
Use Google Flow if you create video or visual stories.
Watch for Ask YouTube if you want conversational discovery for videos.
Watch for Universal Cart inside Search and Gemini this summer if you shop across Google surfaces.

The honest caveat is availability. Some features are global now. Some are U.S. first. Some are for subscribers. Some are trusted-tester only. Some are coming this summer. Google I/O 2026 announced the direction, but not every feature is equally available on day one.

The clean summary

Google I/O 2026 had a lot of announcements, but they all point in one direction: Gemini is becoming the action layer across Google.

Gemini 3.5 Flash is the fast agent model. Gemini Omni is the multimodal creation model. Search is becoming an agentic and generative interface. Gemini Spark is the personal agent. Daily Brief is the daily personal digest. Antigravity is the developer agent platform. AI Studio is becoming an app-building and deployment surface. Managed Agents give developers remote sandboxed agents through the API. Workspace is becoming voice-driven and context-aware. Flow is becoming an agentic creative studio. Gemini for Science brings agents into literature review, hypothesis generation and computational discovery. SynthID and C2PA support give Google a provenance story for the media age it is helping create.

That is the real news from I/O 2026. Google is not just launching AI features. It is trying to make Gemini the connective tissue between questions, context, creation and action.