A study by Columbia Journalism Review’s Tow Center for Digital Journalism revealed significant accuracy issues with AI-driven news search tools.

The research tested eight AI models and found that over 60% of the queries about news sources were answered incorrectly.

The error rates varied, with Grok 3 having the highest at 94%.

The study highlighted a troubling trend where AI tools, instead of declining uncertain queries, often generated plausible-sounding but incorrect answers.

Premium versions of some tools fared worse in terms of errors, and many AI models bypassed Robot Exclusion Protocols, leading to unauthorized access to paywalled content.

These issues pose challenges for news publishers, who face decisions about whether to block or allow AI crawlers, ArsTechnica has reported.

Despite the problems, industry experts see room for improvement in future versions of the tools.

Written by B.C. Begley