Enrique Dans's article discusses a study evaluating the accuracy of generative AI algorithms in providing bibliographic citations and URLs. The research tested well-known algorithms like ChatGPT and Gemini, finding a high error rate in their responses. 

The algorithms often fabricated links and cited incorrect or non-existent sources, even with premium versions performing worse than free ones. Dans suggests that these tools are unreliable for bibliographic tasks, despite their potential in other areas. 

He emphasizes the need for careful verification and recommends using traditional search engines for citations.

This article is also available in English on my Medium page, «When AI flips its wig: the problem with generative algorithm citations«