At the Google I/O 2026 event, Google announced the overhaul of Search, information agents, and Gemini 3.5.

At times, Google will take users to an “intelligent box” where they will provide answers to questions. It is quite similar to the AI Overview, but more expansive, and users will be able to follow up with questions.

However, currently, standard Search with a mini-AI Overview still exists, at least from my testing. In other regions of the world, this update has been rolled out. If you don’t notice anything different, expect it soon.

Other features Google will be bringing soon is new agentic AI that can work even in the background. Google cites an example, if you’re looking for an apartment, you can provide all your requirements, and the agentic AI will keep watch for anything new. In theory, this would mean, if in your initial search nothing shows up, but later something does, Google’s agentic AI will notify you right away, saving you time to do another search in a few hours and the opportunity to immediately get that apartment. It would be interesting to read about users having similar searches, for instance, a rare item or, yes, an apartment, being suddenly available. It would no longer be a matter of luck when one user checks in again after a couple of hours, or days, since in theory, everyone searching for the same thing will get notified, assumingly, simultaneously, it would all be about the user’s speed in making that first phone call. It’d also be interesting if users can use this as a way to keep track of the volatility of stock markets or even crypto prices.

Apparently, it can also tune in on social media posts and websites. This would mean, Google’s new Search could just send users the summary of events, without the user actually visiting those sites. I can see how people might like that, saves time and all, as they say. But good luck on page views for everyone running a blog and a website.

However, the agentic features will roll out in the next couple of months, and will be available for Gemini AI Pro and Ultra users, for now.

In the future, Google will also be rolling out a little something for vibe coders. Through Google’s AI, agentic platform, Antigravity, users will be able to generate visual images such as graphs and tables, and generate apps like trackers. It should be noted that these generated apps wouldn’t be full fledged apps with a bunch of features. They’ll function more like things that you need at the moment and wouldn’t need for the long term. For instance, a specialized calculator that can calculate how many tiles you would need to fit in a room, or a compass app if you don’t have one.

Again, this vibe coding feature will roll out later and will be available only for Gemini AI Pro and Ultra subscribers.

All the new AI features are made possible with Google’s Gemini 3.5 rollout. As of now, Flash and Flash-Lite. It is designed more for agentic executions, such as the rollout of intelligent agents to keep searching in the background.

Omni is Google’s new video generation platform. Omni can generate and edit videos based on images and prompts in a conversational manner. Also, Google says you can upload videos you’ve taken yourself and can describe what changes to make.

Gemini Spark is described as a personal AI assistant. It can link up to Workplace apps and monitor activities. It can, for instance, monitor emails and payments. It can go through your Gmail and create a Doc based on certain information. For instance, if a user has frequent correspondence with a project manager, Spark can take all the information between those exchanges and create a Doc file as a draft. And because Google has a partnership with Canva, Spark will be able to create visual presentations as well.

So far, US users will get first dibs on Spark and only for Google AI Ultra subscribers, for now.

Gmail Live allows voice prompts to go through your email. It includes other features such as Doc Live that transcribes user speech into text.

With all of this AI implementation everywhere and the prevalent misinformation. Google has improved their AI identifications. Users will be able to Circle to Search images to verify if it was AI generated. However, this only works with AI content generated by Google’s own tools.