Google Ads Let Deceptive Ads Reach 100 Million Views in Europe

Fake virus alerts and impossible promises

Many ads used scareware tactics, claiming a phone was hacked or a computer infected and urging an immediate app download. One Polish-language ad claimed an Android phone had 178 viruses after visiting adult sites and drew about 1.1 million views. Another promised to locate anyone from a phone number alone; it ran for more than two years and collected 14.5 million views. A third offered a last chance to recover 284 photos deleted seven years earlier and was seen 2.8 million times.

Removing one ad was not enough

The researchers also reported ads to Google. Most were not judged to violate its policies. In the case of a domain tied to the TamperedChef campaign, which spread malware disguised as a PDF editor, they found 42 ads pointing to it and reported one. Google removed that ad, but the other 41 were not withdrawn as part of that action. Of 147 apps promoted through suspicious ads, 34 were no longer in app stores while the ads stayed live.

AdLens works in two stages: it first compares ad content with known scam patterns, then more advanced language models, including Gemma and Qwen families, examine selected ads using text and visuals. It reached an F1 score of 0.922. The authors admit it has limits and cannot fully replace human review.

The tool has been released publicly so independent researchers and consumer groups can run similar analyses. The authors note they examined only a small slice of Google's ads, so the results cannot estimate the total number of fraudulent campaigns or user losses.