Google I/O: Android Auto engages gear

A Volvo stood outside the keynote, engine off, systems on, and ready to talk.

There was no label on the windscreen. No velvet rope. Just a production car running the Gemini AI model on Google built-in, parked between the developer queues and the Shoreline Amphitheatre stage. It didn’t introduce itself. It waited for instructions.

That was the quiet announcement before the announcements at the Google I/O develop[er conference in Mountain View, California, last week.

Inside, the keynote made it official: Gemini has moved into the dashboard. Android Auto has been updated with new tools. Google built-in has been equipped with an assistant that holds a train of thought.

This is no longer about connecting a phone to a screen. The system knows who’s speaking, where they’re going, and what they’re trying to get done.

Unlike Android Auto, Gemini is not a car OS. It doesn’t handle driving. It doesn’t run maps. It doesn’t touch the vehicle’s control systems. It listens, interprets and responds.

The updates to Android Auto were simpler. The interface, which runs via a tethered phone – via cable or unwired – is gaining streaming apps, a web browser, and multiple weather tools. YouTube, Prime Video and Vivaldi will now show up on dashboard screens – but only when the car is parked. The additions are intended top be used for charging stops, waiting passengers, or those surreal moments when traffic and deadlines both vanish and you’re just sitting still.

More usefully, a new wave of weather integrations means that forecasts now appear without hunting through menus. The system checks your location and serves relevant data to the screen. The functionality no longer feels unfinished.

The real shift is happening inside Google built-in, the embedded Android platform that runs directly on a growing number of cars from manufacturers like Volvo and Renault. This is where Gemini is being trained to work.

“We’ve been working hard to bring Gemini to more surfaces,” said Shahram Izadi, Google’s vice president of engineering. “The car is one of those surfaces where context really matters, and where voice becomes the most natural interface.”

Gemini answer questions, but also responds to purpose. In the I/O, demo, the system was shown helping a driver meet a friend halfway between cities. It found a midpoint, suggested a café, considered traffic, and pushed the result to navigation – all in one exchange.

It was also asked to recommend scenic alternatives to a destination. It listed options, summarised differences, adjusted the map, and offered to reroute, without losing the original task.

Izadi described this process as “reasoning across context.” The assistant pulls from location data, calendars, traffic and maps, not just to reply, but to construct a response.

Are we there yet?

Android Auto isn’t there yet. The projected version remains a reflection of the phone. It doesn’t carry Gemini. It doesn’t hold memory between sessions. But the groundwork is being laid, and the behaviour is starting to change. What it now offers is consistency: a predictable interface, and better response timing.

Google built-in, on the other hand, is developing into an environment. The assistant moves within apps. It doesn’t reroute the driver unless asked, but it knows how to.

That approach avoids the theatrical and sticks to the useful. There is no claim that Gemini in the car will change the nature of driving. There’s no suggestion that it’s a co-pilot or navigator.

The vehicles that support Google built-in still represent a subset of the market. The availability of Gemini inside them is still emerging. But the intent is clear. Rather than a test run it’s an arrival.

* Arthur Goldstuck is CEO of World Wide Worx, editor-in-chief of Gadget.co.za, and author of “The Hitchhiker’s Guide to AI “.