To coincide with the Samsung Galaxy S26 launch, Google announced three Gemini-related upgrades for Android, coming first to both the Galaxy S26 series and the Google Pixel 10 range. In fact, one of these new features seems like a significant step forward for AI on smartphones: the ability for Gemini to handle multi-step tasks on an Android phone. The examples Google gives are booking a ride through Uber or reordering a meal on DoorDash, where the AI takes care of everything in the background, leaving you free to worry about, well, whether it’s going to work.
Google wants Gemini to perform tasks on your behalf
“When you’re working on a repetitive daily task, it’s easy to daydream about handing it off to someone else,” explains Google in the announcement blog post. “Now, we’re showing an early preview of what’s possible.” Understandably, this has a “beta” label as well as an “early preview” label attached, and it’s initially only going to be available on Galaxy S26 phones and Pixel 10 phones (the Pixel 10a misses out). It’ll also be limited to the U.S. and South Korea to begin with.
As per Google, the process will be tightly controlled: Automations begin with your voice command and end as soon as the task is finished. Live notifications will keep you updated throughout, and you can jump in and take manual control at any time. What’s more, the app is run in a “secure, virtual window” on your phone, with restricted app access.
It’s going to be interesting to see how well this works. If it works well, this feels like a notable moment in getting AI to actually do mobile tasks for us. It’s not clear if Google needs third-party app permission for this, but for now the feature will be available “for select apps” in the food, grocery, and rideshare categories.
Google is upgrading Circle to Search and Scam Detection
Next, Google is upgrading Circle to Search on Android, the feature that lets you circle anything on screen and run a search from it. Circle to Search can do everything from find you a lamp online to warn you about a phishing scam.
Now, Circle to Search will be able to search multiple elements on screen at once. It might be different pieces of furniture in a room, or different items of clothing in an outfit. You don’t actually make the selections yourself: You just circle the relevant area (or the entire screen), and Gemini picks out the various constituent parts.
What do you think so far?