AI search engines can often resemble that overconfident friend who talks authoritatively on numerous subjects without real understanding. A recent study by the Columbia Journalism Review (CJR) highlights that AI models from companies like OpenAI and xAI frequently provide inaccurate information when asked about specific news events, with serious errors present in 60% of the queries tested. For instance, Perplexity returned incorrect information 37% of the time, while xAI's Grok fabricated details 97% of the time, sometimes even inventing fake URLs. Moreover, some AI search engines, such as Perplexity, have been known to bypass site paywalls despite publishers using do-not-crawl instructions, arguing that their practices fall under fair use. This controversy affects publishers’ control over how their content is presented, which can harm their brands if users receive incorrect information that they believe originates from reliable sources. Concerns about user gullibility are also raised, with Time magazine’s COO, Mark Howard, suggesting that consumers should be skeptical of the accuracy of free AI tools.

People’s tendency to seek quick answers without verifying them is evident, as many prefer immediate responses from AI models rather than clicking through to websites. Even before generative AI gained traction, over half of Google searches were "zero-click, " emphasizing the desire for accessible, straightforward information. The findings from CJR are not surprising, given the inherent limitations of language models that function as advanced autocomplete systems, often producing content that appears logical without fully understanding it. Despite these issues, Howard expresses optimism for future improvements in chatbot technology, though he acknowledges the risks of disseminating inaccurate information.