What Google announced at I/O: the AI, search and hardware shifts that matter

Every year, Google I/O serves as a kind of weather report for the consumer internet. If you want to know what your phone, your search bar and your browser tabs are going to look like in twelve months, you watch the keynote. This year's edition was unusually dense — a sprawling pitch for an "agentic" future, a new flagship Gemini model, the biggest overhaul of Google Search in its history, and the long-promised return of smart glasses.

Stripping away the demos and the developer jargon, four things actually matter for ordinary users. Here's what they are, and why Australians in particular should pay attention.

1. Gemini 3.5 Flash: the model that will quietly run your day

The headline AI release was Gemini 3.5 Flash, Google's new mid-tier model. As Mashable reported, the model is being rolled out free to users through the Gemini app and AI Studio, with paid tiers offering higher usage limits.

The significance isn't that it beats every competitor on a benchmark chart. It's that "Flash" models are the ones Google actually deploys at scale — the engine behind summaries in Gmail, answers in Search, and the assistant baked into Android. When Google upgrades Flash, it's effectively upgrading the default AI experience for hundreds of millions of people who never type the word "Gemini" into anything.

For Australian users, that means the AI suggestions that pop up when you're drafting an email, searching for a tradie, or asking your phone to summarise a PDF are about to get noticeably faster and more accurate — without you doing anything.

2. Search, rebuilt around answers instead of links

The most consequential announcement wasn't a model at all. It was Search. The Verge described the changes as "the biggest ever" for Google's flagship product, and that framing is hard to argue with.

The shift is structural. Where Search has spent 25 years pointing you at other people's web pages, the new version increasingly tries to answer the question itself — synthesising sources, generating summaries, and building interactive results on the fly. AI Mode, AI Overviews and agentic features that can complete tasks (book a table, compare flights, draft a comparison) move Search from a directory to a destination.

This is a genuinely contested change. Publishers — including Australian ones — worry that fewer clicks will reach their sites. Users, meanwhile, get faster answers but lose some of the serendipity of stumbling onto a good blog or a niche forum. Either way, the muscle memory of "google it" is being rewired in real time.

What it means practically

3. The "agentic" pitch: software that does things, not just tells you things

The throughline across Google's developer sessions, captured in its own recap on blog.google, was the word agentic. The idea: instead of asking an AI a question and copy-pasting its answer somewhere useful, you give it a goal and let it act on your behalf — opening tabs, filling forms, calling APIs, comparing options, and reporting back.

Google highlighted new tools for developers to build these agents on top of Gemini, including better tool-use, longer context windows, and tighter integration with Workspace, Android and Chrome. In consumer-facing terms, that's the technology behind features like "help me plan a weekend in Hobart" producing not a list of links but a draft itinerary with bookings ready to confirm.

The honest caveat: agents are still early. They fail in unpredictable ways, they can be tricked, and giving software permission to act on your accounts is a meaningful security decision. The interesting question for the next year isn't whether agents are impressive in demos — they are — but whether ordinary people learn to trust them with real money and real calendars.

4. Android 17 and the return of smart glasses

On the hardware and OS side, PCMag's live coverage tracked updates to Android 17 and a renewed push into XR (extended reality), including Google's smart glasses platform built in partnership with eyewear makers.

Android 17 is largely a continuation of Google's "AI woven through the OS" strategy: Gemini deeper in the assistant, smarter notifications, on-device models for privacy-sensitive tasks, and tighter ties between your phone, your car (Android Auto) and your TV. None of it is revolutionary on its own; collectively, it's Google trying to make sure that when you reach for an AI feature, the Google one is already there.

The smart glasses story is more interesting because it's the first time since the original Google Glass debacle that the company looks serious about a face-worn computer. The pitch is different now: not a tiny screen demanding your attention, but a near-invisible assistant that can translate a menu, identify a landmark, or read directions when you glance at them. Google is leaning on partners for the frames (which is wise — fashion was a key failure of Glass) and on Gemini for the brain.

For Australian buyers, the practical question is timing and availability. Google hardware has historically arrived here later than in the US, sometimes in trimmed-down form. Expect smart glasses to follow a similar pattern — international launches first, local rollout to follow, and carrier deals to determine how widely they actually appear in the wild.

The bigger picture: Google is consolidating, not experimenting

The temptation with any I/O is to read it as a parade of cool demos. This year, the more accurate reading is consolidation. Google is taking the AI capabilities it has been building — Gemini models, agent frameworks, on-device inference — and pushing them into the products you already use every day: Search, Android, Chrome, Workspace, Maps.

That has two implications worth sitting with.

First, the competitive frame has shifted. The interesting fight isn't "Gemini vs ChatGPT" as standalone chatbots. It's whether Google can make AI feel like a default utility — like spellcheck or autocomplete — before rivals can convince users to leave the Google ecosystem entirely. Search getting its biggest overhaul ever is a defensive move as much as an offensive one.

Second, for users, the era of "trying out AI" is ending. You are not going to opt in to Gemini 3.5 Flash; it will simply be the thing answering your questions. You are not going to choose AI Mode in Search; it will be how Search works. The technology is becoming infrastructure, which means the questions worth asking are less about features and more about defaults: what data is being used, what's being summarised away, what's being decided for you, and whether you can turn it off when you want to.

I/O didn't answer those questions. But it made clear that they're the ones we'll all be living with for the next year.

Related on Bleen

Sources