Gemini Live on Android Auto is brilliant — and very broken

Gemini Live has been one of the most helpful AI tools in my everyday life, ever since it launched two years ago on the Pixel 9 series. I use it to get answers while cooking, to check what’s wrong with my plants when something fails, and to ask questions about anything I’m not familiar with.

It’s an essential part of my Android experience and my Google Home smart speaker experience, too, but I recently tested it in a rental car with Android Auto, and the results left me a little… disappointed.

Where Gemini Live shines on Android Auto

Starting a Live chat in Android Auto is as easy as invoking the regular Gemini, then saying, “Let’s chat.” The interface switches from a simple overlay to a full-on side menu reminiscent of the old Gemini Live interface on phones, with black and blue gradients. It’s familiar and offers two very easy-to-access buttons: Hold and Stop. Once you’re in that Live mode, you can start talking, and Gemini will respond while keeping your ongoing navigation or music in the main Android Auto window. Simple as that.

I was navigating around Romania, so I used Live to discuss potential activities, stops, things I needed to know about towns and cities I was passing through or heading toward, and so much more. For example, on my way to Salina Turda, I asked about other activities I could do nearby, and it suggested a quick one-hour hike at the Turda Gorge, which ended up being an awesome experience. I could’ve researched that on my phone, but having a quick back-and-forth discussion to see how feasible it was, how far, how much time the hike takes, the terrain and elevation (I wasn’t wearing hiking gear), and whether the weather was going to change soon helped me make up my mind while I was on the way to Turda. No time lost.

I was partly expecting Gemini Live to only understand English, but, like Live on my phone, it supports multiple languages. I tried switching between English, French, Spanish, and Arabic mid-sentence, and it understood everything I said. I tried asking one question in English, then another in French, and it got that. Even throwing in one random Arabic word amid a bunch of English didn’t trip it. Gemini Live got it all. This ability keeps surprising me because I thought we were years away from AI understanding multiple languages like that, but no, it can do that already! And for someone who tends to think and speak in three or more languages, having Gemini answer my babble is one of the most impressive party tricks I’ve seen from tech in the last few years.

Gemini Live is also a good way to stay awake and focused while driving. Each time I felt a bit of a lull in my drive, I launched it and started blabbering. It kept me engaged on long stretches of highway and helped me pass the time.

The disappointment of using Gemini Live in a car

The negatives, however, far outweigh the positives in my Android Auto Gemini Live experience.

For one, it worked without issue when my Pixel 10 Pro XL was connected to the car, but whenever my husband’s Pixel 7 Pro was connected, it wouldn’t hear anything I said. I tried it repeatedly, closing apps, clearing cache and data, rebooting, to no avail. The Gemini Live interface would get triggered after “Let’s talk,” but no matter how much he or I spoke, how loud, or how slow or fast, it didn’t register a single word. It was unusable. There are many Reddit threads with Gemini and Gemini Live issues on Android Auto, so I’m not the only one to complain about it. What sucks is that I don’t know what the difference is, except that one phone is newer than the other.

Live also disappointed me with the lack of an end trigger. I can start a conversation with it with a “Let’s talk,” but there’s no way to end the chat unless I tap the red close button, like on my phone. I’ve tried dozens of variations of “I’m done,” “Stop,” “Cancel,” “End this chat,” and none worked. On my Google Nest Hub and Nest Audio, these work in Live mode, but not on my phone or in my car for some baffling, unrealistic reason. I should be able to say a couple of words to stop the blabbering, but I need to reach with my hand instead and tap a button while driving — why?!

Access to personal information is also quite restricted. It seemed Live had only cursory knowledge of my calendar and a couple of details, but it couldn’t control my smart home or do anything more advanced with my documents or emails. The more I used it, the more I realized it was basically a version of Live circa 2025, and not the more powerful Live experience we’ve had recently.

But the absolutely most disappointing aspect of using Live in Android Auto ended up being two ridiculously silly design issues: each time there was a Maps announcement coming from the car, it would stop Live (normal), instruct me about the next exit or turn or other navigation details (also normal), and then… just silence. Not normal. Vocal directions in Maps cause Live sessions to quit. Not pause, but completely quit. You have to launch Gemini again, then Live, and try to convince it to go back to the previous discussion, which is no easy feat.

If it were up to me, the two experiences would be completely separate, and Live would resume once the directions were done. It would also support personal information and properly end a chat with a vocal command. And finally, I would use the screen estate that Gemini Live occupies on an Android Auto head unit to display visuals and information related to the questions. So if someone asks about the Turda Gorge, show them photos; if they ask about the weather, show the forecast, and so on. Right now, that’s a wasted rectangle that does nothing except show a blue gradient animation.

Google still has a long way to go to ensure the Live experience on Android Auto is seamless, but I hope it keeps improving because it’s already off to a good start.