Detection firm finds 82% of herbal remedy books on Amazon 'likely written' by AI (www.theguardian.com)

Originality.ai reported that it scanned 558 Amazon listings in the platform’s herbal remedies subcategory (January–September) and found 82% “likely written” by AI. The detection company — which sells tools to universities and businesses — flagged many entries with high confidence (some samples marked 100% AI-generated), pointed to red flags such as nature-themed pseudonyms and liberal use of leaf emojis, and singled out a No.1 bestseller, Natural Healing Handbook by a seemingly non‑existent “Luna Filby,” as probably machine‑generated. The study’s author called the findings a “damning revelation” of unlabelled, unchecked AI content proliferating on Amazon. For the AI/ML community this highlights two issues: the scale of synthetic content in consumer marketplaces and the real‑world risks when automated text touches health domains. Unverified, AI‑authored herbals and prior cases (e.g., chatbot‑written foraging guides) can spread dangerous misinformation; publishers and industry groups are urging Amazon to label or remove fully AI‑written books. Technically, the episode underscores limits of content provenance on distribution platforms, the arms race between generative models and detection tools, and the need for robust metadata, provenance standards and human verification for safety‑sensitive categories. Amazon says it enforces guidelines and uses detection methods, but the study suggests current controls may be insufficient.