A recent study from Just Facts provides a look into the accuracy and bias of leading artificial intelligence tools. Researchers tested four major chatbots including ChatGPT, Gemini, Claude, and Grok with 100 questions covering sensitive topics like crime, abortion, and immigration.
The data shows that many of these models are prone to repeating political falsehoods, particularly those originating from the left. While performance varied by platform, a significant portion of the answers provided by these systems were backed by nonexistent or irrelevant sources.
Jim Agresti, president of Just Facts, noted that nearly half of the sources cited by the chatbots were illegitimate. The study highlights a pattern where these systems frequently fabricate references, a problem that experts warn is dangerous when applied to public policy and health decisions.
While some models showed different patterns of bias, the general findings suggest that users should approach these tools with caution. The researchers emphasized that these AI models do not function as neutral experts and often repeat inaccuracies with high levels of confidence.
For those who rely on these tools, the takeaway is clear. Instead of trusting the output, users must verify the information against credible, independent data. The convenience of these platforms does not replace the necessity of critical thinking.