AI Transparency Labels Paradoxically Reduce Trust in Truth While Boosting Credibility of Scientific Misinformation

A new JCOM study reveals AI labels on social media paradoxically increase the credibility of fake news while reducing trust in true scientific facts.

By: AXL Media

Published: Mar 9, 2026, 6:12 AM EDT

Source: The information in this article was sourced from Sissa Medialab

The Hidden Risks of Automated Science Communication

As artificial intelligence becomes a primary tool for generating scientific content on social media, regulators have rushed to implement transparency mandates to protect the public. However, the move to label AI-synthesized text may be creating a new set of psychological hurdles. AI-generated content is prone to two major risks: "hallucinations" where the model produces factually incorrect but plausible statements, and the deliberate prompting of systems to create persuasive fake news. According to researchers at the University of Chinese Academy of Social Sciences, the labels intended to mitigate these risks are failing to help audiences distinguish between fact and fiction.

The Experimental Design and the Weibo-Style Simulation

To test the efficacy of current labeling trends, researchers Teng Lin and Yiqing Zhang conducted an experimental study involving 433 participants. The team utilized GPT-4 to adapt verified information and debunked rumors from China’s Science Rumour Debunking Platform into Weibo-style social media posts. Participants were presented with four variations of content: accurate information with and without an AI label, and misinformation with and without an AI label. By asking participants to rate the credibility of these posts on a scale of 1 to 5, the researchers were able to measure exactly how the presence of a disclosure tag altered public perception.

Uncovering the Paradoxical Crossover Effect

The study’s most significant revelation is what the authors term the "truth-falsity crossover effect." In a paradoxical shift, the same AI disclosure label pushed credibility in opposite directions based on the underlying accuracy of the text. For factually correct messages, the label acted as a penalty, reducing the user's trust in the information. Conversely, for false messages, the AI label actually increased the perceived credibility. This suggests that instead of acting as a warning, the label may be redistributing trust in a way that elevates falsehoods to the same level of authority as verified science.