A new study suggests AI may be behind much of the religious content currently for sale on Amazon — a finding with implications for content provenance and marketplace trust that should interest readers worried about authenticity in digital ecosystems. Key findings - Originality.ai analyzed 2,034 recently published religious books on Amazon and flagged 1,272 (63%) as “likely AI-written.” - The sample covered 14 religious and belief categories. Witchcraft (Amazon’s Wicca, Witchcraft & Paganism category) came in highest at 78%, with many titles focused on healing crystals, herbal remedies, and techniques to cleanse “negative energies.” - Hinduism and Taoism were flagged at 76% and 74%, respectively. Other faiths fell in between: Sikhism, Buddhism, Islam, Catholicism, Protestantism, Orthodox Christianity, and Judaism. Mormonism, atheism, and Satanism registered lower rates at 42%, 40%, and 22%. - Originality.ai’s author, Michael Fraiman, said many Wiccan books appear aimed at readers seeking self-help or “detox” solutions and “can address these concerns without being held to any scientific standard.” Methodology and caveats - Researchers reviewed book descriptions, author bios, and text samples. Content scoring 50 or higher on Originality.ai’s scale was classified as “Likely AI.” Fraiman emphasized this denotes likelihood, not conclusive proof. - The team also spot-checked flagged titles and believes it can distinguish likely AI-written content from human authorship. Still, AI-detector tools are imperfect and have produced conflicting and sometimes counterintuitive results in past tests. Detector reliability concerns - Detector inconsistency has been documented: in October 2024, four AI detectors gave wildly different readouts on the U.S. Declaration of Independence (ZeroGPT: 97.93% AI; QuillBot: 0% AI; GPTZero: 89% human). - In March, Colombia’s Supreme Court rejected a filing after a detector flagged it as AI-generated; a related submission was later scored 93% AI by the same tool. These examples illustrate false positives and cross-tool variability. Platform response and transparency - Originality.ai told Decrypt that Amazon is “aware of how common AI-written content is on their platform” and relies on an honor system for authors to self-identify AI-generated works. Originality.ai did not share its data with Amazon and said it has no information about how often authors disclose AI use. Amazon did not immediately respond to Decrypt’s request for comment. Broader concerns: training datasets and book sourcing - The debate extends beyond marketplace listings to how AI models are trained. Developers have purchased and scanned physical books to create training datasets. An August investigation traced a shipment of rare books to an Amazon facility in Las Vegas where workers reportedly removed bindings and scanned pages; Amazon confirmed it buys books through commercial channels for such operations. Why it matters - For readers and marketplaces, a high prevalence of AI-generated religious titles raises questions about quality, accountability, and disclosure. For platforms that monetize user-published content, it’s also a trust and moderation problem. - For the broader tech and crypto communities, the findings underscore ongoing challenges in content provenance, attribution, and the limits of automated detection—issues that are driving interest in cryptographic provenance, decentralized identity, and on-chain content verification as alternative approaches to establishing authenticity. Bottom line: The study signals that AI-written religious books are widespread on Amazon, but detection is probabilistic and imperfect. Whether platforms will adopt stronger verification, disclosure rules, or new provenance technologies remains to be seen. Read more AI-generated news on: undefined/news