I subscribe to Google alerts, which sends me an email whenever the phrase “neonatal research” appears on a new website or a new post. I was interested, therefore to receive an alert about an article which, according the blog “Bioengineer.org”, showed a major genetic contribution to the occurrence of Necrotising Enterocolitis.

The blog post includes the following quote “Bai et al.’s study represents a landmark in neonatal research by providing compelling evidence for the heritability of necrotizing enterocolitis in very preterm infants. The twin study design elegantly disentangles genetic predisposition from environmental influences and firmly establishes a genetic foundation for this complex disease”.

This was intriguing, so I checked on the original article. (Bai R, et al. Genetic susceptibility to necrotizing enterocolitis in very preterm infants: evidence from twin data. Pediatr Res. 2025). A nice study, from a group of authors in China, one of whom is my good friend Shoo Lee, working with the Chinese Neonatal Network. They collected data on NEC incidence and chorionicity of twin pairs of less than 32 weeks GA (or <1500g). They found no difference in the likelihood of a coherent diagnosis of NEC between monochorionic and dichorionic twins. They did further analysis restricting to surgical NEC, or comparing early and late onset NEC, and found no difference between mono- and di-chorionic twins.

In other words, the actual findings of the study are exactly the opposite of what the post on that blog stated. The conclusion of the Bai et al authors was : “heritability does not play a major role in the development of NEC”.

I don’t think an actual human being, reading the article, could possibly have misinterpreted the findings quite as dramatically as whatever generated the blog post. The post is accompanied by the following cute image, which they note was AI generated. My only explanation for this dramatic misinterpretation of the original research article is that the post itself is also AI generated, and that the AI engine just loaded the title and some sub-headings from the results (which are, indeed, misleadingly worded as if there were positive findings : “Heritability contributes to NEC” and “Heritability contributes to certain subgroups of NEC”), without being able to realize that the actual results show that the analysis, of what should have been sub-titled “Heritability contribution to NEC”, was actually zero.

At least this is on an obscure blog, and will probably not cause any harm. In contrast, actual primary publications are also being generated by AI, reporting research that never actually happened. Government policy is also being influenced by review articles written by AI, which include non-existent research, or research which has been misinterpreted, often purposefully so, for partisan ends. This is a major issue for the future of medical research.

Let’s be careful out there.