A French investigative journalist has identified more than 15,000 websites that use generative AI to produce fake news content, primarily to capture digital advertising revenue and traffic from Google Discover. The research, conducted by Jean-Marc Manach for Next.ink, found that 75% of these sites are controlled by fewer than 300 operators, with some earning over $2 million in three months. The sites span multiple languages, including French, English, and German, and often feature bylines of non-existent journalists.

The investigation, which relied on digital forensics rather than automatic AI detectors, cross-referenced publication volume, author authenticity, and domain ownership. It also found that traditional media outlets have begun using AI to scale content production, with one fake author publishing up to 500 articles per day. The phenomenon is driven by search-engine optimization specialists who exploit algorithmic recommendation systems, creating what a Brazilian journalists' federation president described as “profoundly unfair competition for professional journalism.”

Separately, researchers at the Australian National University's Emotions and Faces Lab have developed a training method that significantly improves people's ability to distinguish AI-generated faces from real ones. Rather than searching for single flaws, participants were taught to assess six overlapping perceptual qualities: symmetry, proportionality, attractiveness, distinctiveness, expressiveness, and memorability. In the study, all participants improved, and high performers could spot fakes almost every time. The training addresses a growing threat: one in four Australians report encountering deepfake scams, and Interpol has noted the use of synthetic images in social-engineering schemes.

At an Italian conference on economic development and AI, Andrea Imperiali, a professor at Università Cattolica and member of Italy's communications authority AGCOM, argued that public debate on AI oscillates between apocalyptic and enthusiastic narratives, and called for a systemic analysis of its professional, ethical, and regulatory implications. An Argentine commentary, citing the economist Friedrich Hayek and a recent papal document, cautioned against hasty regulation, framing AI as a tool of decentralized cooperation that should not be stifled by centralized control.

The French investigation's findings far exceed previous tallies, such as NewsGuard's count of 3,749 similar sites, and underscore the challenge for platforms and regulators. The Australian team is now working to optimize the training—making it shorter and ensuring long-term retention—while the Italian conference continues to host discussions on energy, infrastructure, and the role of AI in industry. The next milestone to watch is whether platform algorithms adapt to the documented scale of AI-generated content farms.