AI Search Engines Struggle with Citation Quality, Study Reveals

This Columbia Journalism Review article examines how several AI search engines handle news citations and finds that citation quality is a serious problem. The research team from the Tow Center for Digital Journalism tested eight generative search tools with live search features to see whether they could accurately retrieve and cite news content. Their methodology focused on evaluating not only whether the systems could answer questions, but whether they could attribute information correctly and reliably. This is an important distinction because AI search tools are increasingly being used as substitutes for traditional search, yet users need trustworthy sourcing if they are going to rely on the answers.

One of the article’s main findings is that licensing deals with news organizations did not guarantee accurate citations. In other words, even when a platform had some form of content agreement with publishers, the resulting chatbot answers could still misattribute, distort, or mishandle the source material. That insight is especially important for publishers and information professionals who may assume that licensing automatically solves citation quality. The article suggests that the problem is deeper than access rights; it concerns how AI systems retrieve, format, and present information after search.

The article’s practical value lies in showing why users should be skeptical of citations produced by AI search tools, especially for news and current events. It also implies that organizations should not rely on AI answers alone for editorial, research, or monitoring tasks. Instead, practitioners should verify source links directly and examine whether the text cited by the tool actually matches the claim being made. The article is relevant to journalists, researchers, and SEO teams because it shows that visibility in AI search is not enough; accurate and faithful citation is a separate quality that can fail even when content is technically accessible. Overall, the piece provides a concrete, research-based warning about the reliability of AI search engines in real-world information workflows.