• by Laurie Sullivan  @lauriesullivan, May 8, 2025

Google has expanded the ways it uses artificial intelligence (AI) to prevent scams in Search, Chrome, and Android, and on Thursday began sharing ways it has implemented the technology to keep users safer.

Classifiers using machine-learning algorithms now identify patterns, anomalies, and linguistic cues indicative of fraudulent activity.

As tactics change, advancements in AI and large-language models (LLMs) now analyze vast quantities of text to identify subtle linguistic patterns and thematic connections that might indicate coordinated scam campaigns or emerging fraudulent narratives. 

Through this method, Google has identified interconnected networks of deceptive websites that might appear legitimate in isolation, but are dangerous when connected.

LLMs also helped Google scale protections across languages. Whether identifying a scam in English, Hindi, German, Spanish, or another language, LLMs have allowed Google to train its systems to catch those scams in multiple other languages as well thus limiting user exposure to scams globally.

AI in search can detect and block hundreds of millions of results based on scams that can impact a brand's reputation through lookalike web listings and sites.

Google’s Fighting Scams in Search report shows how investments have enabled it to identify 20 times the number of scam-type pages.

“For more than a decade Google has used advancements in AI to protect you from online scams where malicious actors deceive users to gain access to money, personal information, or both,” Jasika Bawa, group product manager and senior director of engineering at Google, wrote in a blog post.

AI enables Google to analyze larger quantities of Web text to identify coordinated scam campaigns and emerging threats. The technology ensures that results are legitimate, as well as protect users from harmful sites trying to steal sensitive data. 

For example, Google noticed a significant increase in bad actors on the web impersonating airline customer service providers and scamming people in need of help. Using AI, the company reduced these types of scams related to phone numbers by more than 80% in Search.

There had been a rise in misleading pages that mimic official resources such as those with information on visas or other government services. In 2024, Google implemented new protections that decreased scams that impersonate official sites by more than 70%. 

Gemini Nano, an on-device large language model (LLM) for desktop, also provides Enhanced Protection for users with another layer to defend against online scams. It identifies insights into risky websites and enables Google to offer protection, even against scams that have not previously been identified.

The same approach is being used to protect users from remote tech support scams, but the goal is to expand this protection to Android devices and more types of scams in the future.

Google has implemented AI to stop malicious sites that are attempting to defraud people through notifications with an AI-powered warnings for Chrome on Android. When Chrome’s on-device machine-learning model flags a notification, the user receives a warning with an option to unsubscribe or view the content that was blocked. If the warning serves up incorrectly, the user can choose to allow future notifications from that website. 

Android phones now notify the user when messages or phone calls are a likely scam, through an on-device AI-powered Scam Detection in Google Messages and Phone by Google to protect Android users from these types of sophisticated scams. 

Google also offered tips to look for in lookalike domains in search results. Bad actors often use similar domains that trick users. For example, instead of the domain @thisisgoodlink.com a bad actor may use the similar domain name “@thisisagoodlink.support.”

Another way to gain more information about an online source is by using About this result -- the three dots next to a search result that enables users to learn more about sources like an online store before clicking into its page.

Users should beware of strange formatting, unusual fonts, or unexpected symbols or emojis, which could indicate a spoof site.