A new report reveals that hundreds of TikTok accounts are generating billions of views by posting AI-created content, including anti-immigrant material and sexualised imagery. Researchers from AI Forensics, a Paris-based non-profit, identified 354 accounts that collectively accumulated 4.5 billion views over a month, posting 43,000 pieces of AI-generated content.
The study, conducted from mid-August to mid-September, found that some accounts posted up to 70 times daily, indicating automation. Many were launched at the start of the year. Half of the most active accounts focused on content related to the female body, often depicting stereotypically attractive women in sexualised attire. Less than 2% of posts carried TikTok's AI label, raising concerns about deceptive potential.
Some content took the form of fake broadcast news segments featuring anti-immigrant narratives and sexualised depictions of girls appearing underage. These posts sometimes used known brands like Sky News and ABC. Researchers noted that accounts could evade moderation for months, though dozens were later deleted. TikTok stated the report's claims were unsubstantiated and highlighted its efforts to remove harmful AI content and block bot accounts.
AI Forensics acknowledged that some popular AI content, termed 'slop', was nonsensical or bizarre, such as animals in Olympic diving contests. The report also found accounts promoting health supplements and sponsorship schemes. Researchers welcomed TikTok's option to reduce AI content but stressed the need for better labelling, given the structural failure to identify such material.
