TikTok's AI Avatar Problem: How Fake Black Women Are Being Used to Sell Explicit Content

The Investigation That Forced TikTok to Act

Here is a sentence you should not have to read in 2026: dozens of social media accounts have been using AI-generated avatars of sexualised black women to funnel users towards paid explicit content. And yet, according to a BBC News Arabic investigation published this week, that is precisely what has been happening across TikTok and Instagram, largely unchecked, until journalists started asking questions.

The BBC, working alongside independent AI publication Riddance and analysts Jeremy Carrasco and Angel Nulani, reportedly identified approximately 60 accounts featuring AI-generated black female avatars across both platforms. These were not subtle deepfakes tucked away in dark corners of the internet. They were racking up views in the millions.

Stolen Content, Synthetic Faces

The investigation highlights a particularly grim technique. According to the BBC, a Malaysia-based model identified as Riya Ulan allegedly had her video content stolen and manipulated, with an AI-generated dark-skinned face overlaid onto her original movements. The result? A fabricated person performing in a real creator's body.

The numbers reported by the BBC are staggering, if difficult to independently verify. One manipulated video allegedly reached 35 million views on TikTok and 173 million on Instagram, roughly 47 times the engagement on Riya's original post. One account reportedly amassed 3 million TikTok followers within weeks of its creation in December. That is not organic growth. That is algorithmic rocket fuel.

A Naming Convention That Says It All

The accounts were not even trying to be coy about what they were doing. According to the investigation, many incorporated racial terms such as 'black', 'noir', 'dark', and 'ebony' directly into their names. This is consistent with broader research showing that phrases like 'AI Generated Black Women' and 'Black Women Baddies AI' have become active search categories on TikTok's own discovery pages.

Links from these accounts led to paid sexually explicit content sites. Crucially, the destination sites labelled the material as AI-generated, but the social media platforms hosting the promotional content did not. A convenient omission, one might say.

Platform Responses: The Usual Script

TikTok banned 20 accounts after the BBC made contact. Not before. After. The platform stated that it prohibits AI-generated content depicting individuals without their permission and has 'zero tolerance' for content promoting off-platform sexual services. Multiple investigations by outlets including The Bureau of Investigative Journalism and Media Matters have documented the rather wide gap between TikTok's stated policies and what actually gets enforced.

Meta's response was even less convincing. The BBC reports that the company said it was investigating but took no immediate action. Nine Instagram accounts reportedly disappeared later, though the timeline and reasoning remain unclear. Riddance's own earlier investigation found that of 23 reported videos on Instagram, Meta removed just one.

More Than a Content Problem

Moroccan model and content creator Houda Fonone described the trend as an 'erasure' of authentic black women's representation online. She is not wrong. This is not simply about dodgy AI content. It sits within a much larger documented pattern stretching back to at least mid-2025, from Google's Veo 3 enabling the so-called 'Bigfoot Baddie' trend to AI-generated fake videos that Fox News mistakenly reported as real.

As Brookings Institution scholar Nicol Turner Lee has noted, there is a grim historical precedent here. During slavery, black people were deliberately exaggerated in illustrations to emphasise supposed primal traits. A review of 5,000 AI prompts found that 68% of outputs leaned on stereotypes. The technology is new. The racism is not.

The Bottom Line

Social media platforms have the tools to detect and remove this content proactively. They simply choose not to until a journalist rings them up. That is not moderation. That is public relations.

Read the original article at source.