A recent study by the Tow Center for Digital Journalism highlights significant citation issues with generative AI search tools. The research revealed that these chatbots, which are increasingly used by nearly one in four Americans as alternatives to traditional search engines, often provide incorrect or misleading citations for news content. In fact, over sixty percent of the queries returned inaccuracies, with some tools, like Grok 3, displaying error rates as high as ninety-four percent. The study tested eight generative search tools through sixteen hundred queries, demonstrating that even premium models were more likely to give confidently incorrect answers, undermining user trust. The findings underscore the urgent need for better evaluation and citation practices in AI-powered searches, as they often obstruct traffic to original news sources and misattribute content.
Why do we care?
For IT providers, the key takeaway is that generative AI search tools remain experimental at best and dangerous at worst. Until these tools can consistently provide accurate and verifiable sources, businesses should treat them as unreliable assistants rather than definitive sources of truth, particularly where information accuracy is key.