Earlier technologies of surveillance needed informers. Somebody had to write down where a man went on Tuesday, whom he met and what he ate. The new kind asks for nothing of the sort. The man writes it all down himself, in his email, his calendar and his camera roll, and is glad to have a place to keep it. Then, on an ordinary Tuesday in October, the machine offers to drive him there.
On 20 October 2026 Google started rolling out Personal Intelligence, the Gemini feature that draws on a user’s own data to shape its answers, to Android Auto. The feature had not been available in the car before, as Ryan Whitwam wrote for Ars Technica. Cars with Google built-in, which run Google’s software in the infotainment system itself instead of mirroring a phone, will get it over the next few months. It remains optional. Whitwam added a “thankfully” to that fact, and the word carries the whole mood of the announcement.
The archive on wheels
Personal Intelligence came at the start of the year as a paid extra for Gemini and later spread to every user. According to Thurrott’s report on the launch, free accounts got it in March; until then it belonged to Google AI Pro and Ultra subscribers. Its logic fits in one sentence. A model answers better the more it knows about you, and Google already knows a great deal. Google’s own phrase is that generative AI does better “when it knows the world and knows you.” Whitwam calls this correct “in an uncomfortable sort of way.”
Here the company sees its advantage over its rivals. Other assistants have to be told things. Gemini can look them up. According to 9to5Google, the sources it draws on are Workspace apps such as Gmail and Calendar, plus Photos, Maps, Search and YouTube, so a driver can skip details that already sit in an email, a picture or a “place you saved in Google Maps.”
The examples Google chose are deliberately small. Say “take me to my doctor’s appointment” and Gemini takes the destination from the calendar. Ask “What’s my flight gate? And is my flight on time?” and it finds the booking confirmation in Gmail and checks the status. Ask it to “find the car wash from the photo I took last week” and it reads the image to work out the place, then starts navigation. Another sample request asks for an indoor restaurant on the route that suits the driver’s “usual vibe”, so that a vegetarian need not explain the diet every time.
Google’s case is about attention. It argues that the car is a natural home for the feature because the driver can watch the road and stop dictating detailed instructions. “You don’t have to hunt for the confirmation email, then do a separate search for delays,” the company says. “You don’t have to remember the street address or look it up before you drive.”
That much is true. It is also true that the assistant needs no instructions because it already has the answers. Convenience and knowledge are one transaction here, seen from opposite ends.
The order of arrival
For once the sequence is reversed. Google usually gives new car features first to vehicles with Google built-in, a small fleet Whitwam describes as “very few vehicles”, and a label he calls questionable branding. This time phones projecting onto dashboards come first and the native systems wait. Those systems, in newer models from Ford, GM, Honda, Nissan, Volvo and Polestar according to launch reports, only got the Gemini assistant in April. Even after the update, 9to5Google reports, they will give personalised answers only “when you’re in the car with your phone connected.”
The map also has borders. Thurrott reports that Android Auto is getting the feature worldwide except in Switzerland, the European Economic Area, the United Kingdom and Nigeria. Only eligible personal accounts held by people aged 18 or older qualify. Workspace accounts for business, enterprise and education are excluded. Read in reverse, the list shows where regulators or corporate IT departments have made a company stop and wait.
The switch
The feature is off by default. Users can pick which apps Gemini may consult and can turn it off whenever they like. There is a catch, reported by How-To Geek and others: anyone who has already switched it on in the Gemini phone app will find it on in the car too, with no new question asked. That is why Ars Technica ran its piece under the advice that it is probably time to audit your Gemini settings.
Drivers who say no lose little that they can see. “If you don’t use Personal Intelligence, everything in the car still works,” Google says, “but the answers won’t be tailored to you.” Directions, music, calls and general questions all keep working.
This is how it tends to go now. Nobody is forced. The door is labelled, the switch is visible and nothing is hidden. Every reason to flip the switch is on one side: the gate number, the street address, the restaurant that already knows you skip meat. The only reason against it is an unease that the driver must supply alone. Google has spent the autumn sending its agents into the inbox and the office, down to giving one its own email address. The car was one of the few rooms left where a person still had to say out loud where they were going.