Google still approves ads its own AI would ban
Original: Why is Google still serving dodgy ads?
Why This Matters
Ad platform integrity depends on enforcement — a gap between policy and practice erodes user trust at scale.
A developer found a deceptive iPhone storage-alert ad running on YouTube, reported it multiple times, and received identical responses each time: Google said it violated no policies. When the same ad was fed to Gemini, the model flagged it immediately for three distinct violations.
Chris from atomic14 noticed a suspicious YouTube ad mimicking an iOS system alert — the classic 'iPhone Storage is Full' popup — and accidentally clicked it. He reported it to Google. The response came back: no policy violations found. He reported it again. Same answer. Others reported it too, with identical results.
So he tried something else: he showed the ad to Google's own Gemini model. The AI flagged it immediately under three categories — Misleading Ad Design (fake iOS UI elements), Deceptive UI Components (non-functional 'Yes/No' buttons that just redirect users to an app store), and Fear-Based Tactics (fabricating an urgent device failure to drive installs). Gemini's verdict: disapprove immediately.
The author raises a pointed question: if Google's own AI can detect the violation in seconds, why aren't Google's ad review systems using it? He invokes Hanlon's Razor — never attribute to malice what stupidity explains — but also notes that high-click-rate scam ads are financially rewarding for Google, raising uncomfortable questions about incentive structures. The gap between Google's stated ad policies and its actual enforcement appears to be wide enough to drive a fake system dialog through.