Google's Circle to Search can now spot text scams

Got a text claiming you owe an unpaid toll fee that will double if you don’t click a sketchy link in thirty minutes? Don’t tap it. Circle it instead.

Google published official guidance detailing how Android users can leverage Circle to Search to instantly spot SMS phishing attacks, also known as smishing.

Check the link without ever opening it

The workflow takes literally two seconds. Long-press the home button or navigation handle to freeze the screen, then draw a circle around the suspicious message.

Android generates an AI Overview right over your app with analysis and suggested next steps. If red flags pop up, the feature tells you point-blank to trash the message and block the sender.

Google says the backend AI compares message patterns against known scam tactics and live web data. It looks for obvious pressure tactics like demanding quick cash, and checks the underlying link destination against reputable domains.

The biggest win here is zero risk. You get to inspect a shady URL thoroughly without ever touching a browser tab.

According to Google, the tool covers the most common fraud playbooks right now: fake bank alerts, phony package deliveries, unpaid toll fees, and wrong-number icebreakers designed to reel you into pig butchering investments.

Google made a point to clarify that merely reading a text won’t infect your phone. The trap only springs when you click links, download payloads, or reply. Hence their new playbook: stop, circle, search, and delete.

Google is quietly building an anti-scam moat

While Google highlighted the trick on October 15, the underlying capability has been building up for a while. The company first added scam-detection chops to Circle to Search and Google Lens back on December 2, 2025, surfacing warnings whenever system confidence ran high.

If your phone lacks Circle to Search or you are on an iPhone, Google says you can get the same analysis by taking a screenshot and feeding it to Google Lens.

For hands-free defense, Google also maintains notification-level Scam Detection across select devices. Running locally in an isolated sandbox across nine languages, it evaluates incoming message notifications without shipping raw text to Google servers.

The guidance landed directly after a broader security drop on October 14, where Google introduced Enhanced protection mode in Google Messages alongside call-forwarding and screen-sharing safeguards.

Google claims its AI defenses in the Phone and Messages apps already flag over 5 million suspicious calls and scam texts every single month, backing the claim with a third-party audit from Leviathan Security Group.

Why it matters

We think repurposing a screen-search gesture into an on-demand security scanner is a brilliantly pragmatic move.

For regular users and especially older family members who get paralyzed by fake bank warnings, this removes the guesswork. You do not need to teach someone how to inspect domain suffixes anymore; you just tell them to circle the screen before panicking.

From an industry standpoint, anti-phishing has spent decades playing catch-up with static blacklists and spam filters that collapse against freshly registered domains. By applying multimodal AI over the screen layer, Google effectively turns contextual search into an ambient antivirus shield.

The obvious catch is user initiative. This is an active defense tool: it only protects people who already feel suspicious enough to pause and draw a circle. If a frantic message successfully triggers panic, people might still tap the link before thinking about gestures. But as a zero-click sanity check, it is easily one of the most useful AI features currently sitting on an Android home screen.