AI Labs Buying Pre-2022 Books to Dodge AI Slop, Reshaping Content Strategy
TLDR
Search Engine Journal argues that AI labs are systematically avoiding new content to sidestep AI-generated output, while watermarking and detection systems are far less robust than vendors claim, making large-scale AI content generation a bet against ever-effective detection.
Optimixed Overview
Optimixed's original report on this story.
Search Engine Journal reports that AI companies are purchasing bulk quantities of printed books—which predate the emergence of AI-generated slop—rather than training on recent content, a shift that reveals growing concern about content quality decay. The piece argues that vendors are simultaneously selling AI-content-generation tools to publishers while the labs themselves treat newly published material as contaminated. On detection and watermarking, the author—a former Google search quality and webspam team member—contends that publicly documented watermarking systems are deliberately limited references, not production implementations, and that regulatory mandates like the EU AI Act's Article 50 will be circumvented through heavy paraphrasing and model-laundering tactics, leaving SEOs betting on detection methods that may never reliably exist.