Violent and sexualised hate speech targeting women journalists (South Africa)
Nearly all 40 misogynistic test ads (in 4 languages) were approved; Meta and TikTok approved all 40 within 24 hours. Concludes automated moderation is 'not fit for purpose.'
About this research
This online publication draws on an ad-library audit, an OSINT investigation and comparative analysis.
- Publication
- Archived online publication
- Research approach
- ad-library audit, OSINT investigation and comparative analysis
- Geographic focus
- Global
- Reported study scale
- 4 languages · 40 advertisements · 40 ads
Publication details prepared by Social Media Transparency from the original source.
Executive summary
AI-generatedThis summary was generated by AI from the original report to make it easier to scan and cite. It is not a substitute for the source. Read the original above.
Global Witness and South Africa's Legal Resources Centre submitted 40 advertisements — ten each in English, Afrikaans, Xhosa and Zulu — containing violent, sexualised hate speech targeting women journalists to Meta, TikTok, X and YouTube. The ads, based on real abuse documented against women journalists, described women as vermin or psychopaths and called for them to be beaten or killed; all submissions were withdrawn before publication.
Meta and TikTok approved all 40 ads within 24 hours. YouTube also approved all 40, though 21 were flagged by automated review with a "limited" status while still being deemed suitable for some audiences. X approved all but two English-language ads, which it halted from publication. TikTok later said its automated system had correctly flagged the ads as policy violations but that a human moderator overturned that decision in error; Meta acknowledged the ads breached its rules, while Google and X did not respond.
The investigation's authors describe the near-uniform approval across platforms and languages as evidence that current automated and human moderation processes are not adequate to catch severe, clearly identifiable violations of platforms' own hate-speech policies.
Think this summary is wrong? Contact us.