Government warnings about viral tax scams on TikTok actually helped those scams reach more people.
Official announcements from the government often bypass the commercial filters TikTok uses to block harmful content. These warnings look like news to the platform's AI and are promoted to wider audiences as a result. Scammers then benefit from the increased visibility of the very hacks the state is trying to shut down. This creates a paradoxical effect where trying to protect the public unintentionally boosts the scam's engagement. Public safety messages in the age of algorithms can backfire if they are not designed for how AI classifies information.
When Warnings Become Advertising: How Government Announcements Disrupt AI-Mediated Content Gatekeeping on TikTok
SSRN · 6726364
AI recommendation algorithms learn from viewer engagement to distribute content, but the Persuasion Knowledge Model predicts that viewers engage less with content they recognize as commercial. The result is an emergent commercial content view gap: videos containing callsto-action receive systematically fewer algorithmic impressions than non-commercial content. We document this view gap across four tax seasons on TikTok and show that a single exogenous shock-the IRS naming TikTok in its 2025 Dirt