Google's AI search is recommending malware and scam sites

If you have started relying on Google's new AI summaries to browse the web, you might want to look a little closer before clicking.

Earlier this month, Google began expanding its Search Generative Experience, known as SGE. The feature serves conversational, AI-generated answers right at the top of search results, along with embedded links to helpful sites.

However, SEO consultant Lily Ray noticed something alarming: the AI was actively recommending scammy and malicious websites right inside those answers, making it dangerously easy for everyday searchers to stumble into traps.

Redirection chains and fake giveaways

Tests run by BleepingComputer revealed that these shady sites share remarkably consistent fingerprints. Most rely on .online domains, share the same HTML page templates, and bounce visitors through identical redirection hubs.

That pattern strongly suggests a coordinated SEO poisoning campaign designed to sneak into Google's core index.

Once a user clicks on one of these AI-backed links, they are bounced through a rapid series of redirects. Most dead-end at fake verification pages or spoofed YouTube clones that relentlessly push you to enable browser notifications.

If you click accept, scammers gain permission to blast spam and intrusive ads directly to your desktop, even when your browser is closed. In testing, these alerts promoted tech support scams and fake antivirus warnings claiming a machine had ten viruses, urging users to buy McAfee licenses so the fraudsters could pocket affiliate commissions.

Other links routed visitors to fake Amazon loyalty giveaways offering an iPhone 15 Pro to harvest personal data, while some pushed malicious extensions designed to hijack browser search queries.

The cat-and-mouse game

Google told BleepingComputer that it continuously updates its ranking algorithms and applies standard spam-fighting systems directly to SGE. The company also confirmed it removed the flagged examples, noting they primarily showed up on uncommon queries.

Even so, spammers are evolving just as fast, keeping both sides locked in a relentless cat-and-mouse cycle.

If you accidentally granted permissions to one of these sites in Chrome, head to Settings, open Content, click Notifications, and simply remove the suspicious domain from your allowed list.

Why it matters

The real danger here is conversational trust. When a standard search result looks sketchy, experienced users usually skip it. But when an AI assistant wraps that exact same link into a polished, authoritative summary, it inadvertently stamps Google's credibility onto a malicious operation.

For regular searchers, this is a clear signal that AI overviews are not pre-vetted safe zones. If an AI snippet sends you to a site demanding notification access or plugin downloads, close the tab immediately.

We think this incident highlights a critical flaw in the race to deploy AI search. Feeding unvetted web indices into conversational engines creates an entirely new attack surface. Until search engines can reliably sanitize their training and retrieval pipelines, turning search into an AI dialogue might create more security headaches than it solves.