Truace
Subscribe

Who Do We Trust to Correct Health Misinformation? Credibility of AI-Generated Fact-Checking Messages

As artificial intelligence (AI)-generated messages are increasingly used to correct health misinformation, it remains unclear when such messages are perceived as credible and how credibility shapes subsequent information behaviors. This study examines how source labeling (AI vs. human scientists) influences the perceived credibility of health-related fact-checking messages and whether these effects depend on perceived AI objectivity and trust in science. Using a 2 × 2 between-subjects experiment with U.S. adults…

By the Truvace desk1 min read1 sourcePeer-reviewedTRV-2026-2112· cite & verify
Who Do We Trust to Correct Health Misinformation? Credibility of AI-Generated Fact-Checking Messages
Facts for ladies : beauty, dining, by Kinsley's, house decoration, health of women and children by Gunn, Robert A. (Robert Alexander), 1844- author. Public domain

In brief

A 2 × 2 between-subjects experiment with 231 U.S. adults tested whether telling people a health fact-check came from AI or from human scientists changed how credible they found it. The label alone did not change credibility ratings, but credibility judgments depended on other factors such as perceived AI objectivity and trust in science.

The result matters because AI is increasingly used to correct health misinformation, and credibility shapes whether people share a correction or avoid the issue. The study measured intentions rather than actual behavior, so it cannot show whether AI-labeled corrections change real-world conduct or beliefs.

Main points

  1. Source labeling (AI vs. human scientists) alone did not affect perceived credibility of the health fact-checking messages.
  2. Credibility judgments were conditional, depending on perceived AI objectivity and trust in science.
  3. Higher perceived credibility was linked to more message sharing and less information avoidance.
  4. Credibility was not robustly related to accuracy-oriented information seeking.
  5. AI-labeled messages elicited greater intentions to seek additional information independent of credibility.

The gain

Fact-checking messages labeled as AI-generated prompted greater intentions to seek additional information than human-scientist-labeled ones, regardless of perceived credibility.

The rundown

The experiment used a 2 × 2 between-subjects design with U.S. adults (N = 231). Participants saw health-related fact-checking messages labeled as coming from AI or human scientists, and the study measured perceived credibility.

Three information responses were assessed: accuracy-oriented information seeking, fact-checking message sharing, and issue-relevant information avoidance. Credibility predicted more sharing and less avoidance, but not reliably more accuracy-oriented seeking.

The authors conclude that AI-based fact-checking should be aligned with audience trust orientations, since the credibility of AI in health communication is conditional rather than fixed.

Sources

  1. Peer-reviewedHealth Communication2026-10-20

The debate