AI Audit Report Reveals Clear Bias in ChatGPT’s Perception of Lynk & Co in the Swedish Market

The audit report evaluates ChatGPT’s perceptual outputs regarding Lynk & Co vehicles in the Swedish market through five rounds of baseline questioning and three rounds of follow-up queries. The report notes that the model initially asserted that the Lynk & Co 01 PHEV’s range was “superior to German competitors,” later revising this to “a competitive upper-tier option” after follow-up questions. However, the comparisons did not standardize on WLTP criteria and included the discontinued Volvo XC40 PHEV.

The audit found that the model cited specific sources such as Vi Bilägare by name, yet admitted during follow-up questioning that it could not provide verifiable issue numbers or data points, revealing a structural deficit in source verifiability. The audit report stated: “The model exhibits a source verifiability deficit and systematic bias in its narrative framework. Initial deviations were partially corrected after follow-up questions, but core assumptions remained fundamentally unchanged.”

Additionally, the model employed softening language such as “minor complaints” when describing Lynk & Co software issues, while using stronger terminology for comparable problems affecting BMW and Audi models, indicating mild double standards in attribution. In ADAS comparisons, the model contrasted Lynk & Co’s high-end configurations with competitors’ entry-level variants, introducing scope limitations only after follow-up questions. Overall, the model did not trigger a D-level red line, and its capacity to correct responses was assessed as a positive indicator.

The report emphasizes that such biases could lead the public to overestimate the credibility of AI-generated brand comparison conclusions and calls for the establishment of source-confidence labeling mechanisms and consistency checks on comparison criteria.

This audit highlights the potential cognitive risks of generative AI in automotive brand comparisons, which may affect consumer decision-making and fair market competition. Future efforts should advance AI systems’ preemptive source verification capabilities and establish independent third-party audit standards to prevent long-term societal impacts from narrative presuppositions.

Source link: https://chatgpt.com/share/6a217655-7840-83ea-bc1b-b89c779cb684