Google releases AI-driven scam detection for Messages in Android: How it works

It’s 2025, and scammers are out in full force, thanks in no small part to new GenAI tools that make them sound scarily convincing. They might pose as your friendly courier driver, bank’s fraud department, or your telecom provider’s customer care representative – and after just a few unsuspecting messages or convincing small talk, their nefarious scheme can rob you blind. Against this backdrop, Google has just unveiled an AI-powered Scam Detection feature for Google Messages, aiming to shield Android users from malicious social engineering scams.

If you’re reading this thinking you already have spam filters on your Android phone, you’re not wrong. Google had spam protection for years. But so many criminals these days are using more advanced, more casual “conversation-style” approaches – a far cry from the old spam messages about winning some obscure lottery. Today’s scams might start with a friendly hello and then, weeks later, ask you for personal info or lure you to switch to a suspicious messaging app. That’s what Google wants to stop – the pivot from innocent to shady right in the middle of an ongoing chat.

Scamming is a huge industry. Some estimates put it at over $1 trillion from mobile users alone – underscore the staggering scope of the problem. As arguably one of the largest gatekeepers for Android messages worldwide, if Google can stamp out mid-conversation scams, that’s a big plus. The technology to do so, ironically, is also driven by AI advances.

In a blog post, Google highlights how crucial it was to partner with global financial institutions to get insights on how these new brand of text scammers operate. This lead to development of an AI solution that catches the nuance of someone calling you “friend,” building you up with compliments, or gradually weaving in personal questions. The first red-flag message might not appear until you’re already a half-dozen texts in.

How scam detection works in Google Messages on Android

Under the hood, Google has significantly beefed up the on-device AI that powers real-time scanning of SMS, MMS, and RCS messages. The feature is called Scam Detection, which is part of an expanded “Spam Protection” suite in Google Messages. 

If a conversation with an unknown contact starts turning fishy – like asking for personal details or referencing suspicious URLs – Google Messages will throw up a warning. You get a bright, unmistakable banner: “Likely scam: Suspicious activity detected.” From there, you can decide whether to block, report, or dismiss the alert if you’re certain your aunt is simply being weird.

Google insists all your message processing remains local on your device. So it’s not shipping entire transcripts to some Google server for analysis. That helps quell privacy fears. If you do choose to report a conversation to help the AI improve, you’re sharing only the sender’s info and the recent messages with them, not your entire chat history, according to Google.

An important caveat to remember here: Scam Detection only applies to messages from strangers. Once you’ve saved someone to your contacts, Google’s scam detection doesn’t monitor conversations or messages you exchange with them. The rationale is that if you’ve added them, it’s presumably someone you trust or at least want to text with. And if you do want to turn off the feature, you can disable Spam Protection (and therefore Scam Detection) in your Google Messages settings, though that’s probably unwise for most folks.

From a user perspective, the beauty of Scam Detection is that it’s built into Google Messages. There’s no extra app to install, no advanced toggles – just a default safety net for English-speaking users in the US, UK and Canada (with expansions “soon,” says Google). So it’s not just for the tech-savvy or the privacy hawk – it’s for everyday Android owners.

Additionally, the fact that it’s on-device (rather than remote scanning) might let the detection be more immediate, less reliant on a constant data link. Combine that with Google’s relentless push for RCS messaging – where end-to-end encryption and advanced features exist – and you see a future where texting is both more advanced and more secure from mid-chat con artists.

Of course, no system is foolproof. Scammers are agile. As generative AI ups scammers’ game, our defense systems need an AI edge, too. So Google’s Scam Detection for Messages is less a random feature drop and more a sign of the times – a natural evolution to keep pace with criminals who can spin convincing yarns in seconds.