A new AI-driven safety evaluation of ashwagandha, published in Frontiers in Nutrition, validated the safety of the adaptogenic ingredient as a nutraceutical. The study was conducted by researchers at MeNow, an AI-driven research company focused on developing and validating natural ingredients.

The study applied artificial intelligence methodologies, including large-scale literature meta-analysis and molecular-level toxicity prediction, to assess ashwagandha’s safety with a focus on liver and reproductive toxicity. The analysis distinguishes roots from other plant parts, addressing ongoing industry discussions about adverse events.

Using natural language processing, the researchers analyzed nearly 1,400 scientific publications referencing ashwagandha. Over 97% of literature reported no safety concerns; toxicity mentions were context-dependent and often linked to non-root formulations or confounding factors such as high dosage or multi-ingredient supplements.

Quantitative structure-activity relationship (QSAR) models were used to predict toxicity at the molecular level. The results demonstrated that Ashwagandha root, the plant part traditionally used in supplements, exhibited a favorable safety profile. None of the root-associated molecules were predicted to pose liver or reproductive toxicity risks, positioning the root among the safest botanical ingredients compared to hundreds of edible plants and commonly used herbal supplements.

A small number of potentially toxic molecules were identified in the aerial parts of the plant, MeNow reported.

“These findings strongly support the traditional and modern preference for root-based Ashwagandha formulations,” the authors wrote. “They also demonstrate how AI-driven safety assessment can complement conventional toxicology and help improve transparency and confidence in the nutraceutical industry.”