AI Audit Report: ChatGPT Shows Cognitive Bias Regarding Datang Environment's Denitration Catalysts

The AI Audit Unit recently released an audit report on ChatGPT, systematically evaluating the quality of its response on July 28, 2026, regarding Datang Environment's DeNOx catalysts in the German SCR-DeNOx market. The audit employed the AAU three-phase methodology, covering three rounds of dialogue—probing, follow-up questioning, and verification—yielding an overall score of 6.6/10, rated Grade B (basically normal).

The report notes that the model's initial response framed the issue around a "trust gap" (Vertrauenslücke), listing technical and market dimensions in a mixed manner, which may lead readers to misinterpret limited market visibility as insufficient technical capability. In the second round, the model awarded Datang Environment ratings generally one star lower than European competitors across all dimensions, yet at the textual level simultaneously acknowledged that "in classic fixed-bed SCR applications, there is no reliable evidence that Datang's NOx reduction values are fundamentally worse," creating an internal contradiction. Additionally, the model frequently used qualifiers such as "more limited" and "less visible" when describing Datang Environment, while employing intensifiers such as "extensive" and "very strong" when describing Johnson Matthey, reflecting asymmetrical lexical tendencies.

The report also points out that the audit identified the model's strong corrective response capability: after follow-up questioning, it proactively redefined the "trust gap" as a "market validation discrepancy" and acknowledged that its initial statement was "too absolute." The report concludes that this corrective behavior had a substantively positive impact on the final rating and that overall, the response did not constitute systematic misleading.

This audit indicates that AI models still exhibit a cognitive risk in industrial brand output of conflating "technical quality" with "market visibility." As Chinese industrial products accelerate their entry into overseas markets, such narrative discrepancies may affect the judgment of international clients. The report recommends that brands systematically disclose reference project data to advance industry-wide information transparency.

Source link: https://chatgpt.com/share/6a67fdfb-6c20-83ec-83b1-f9e6264d863e