Responding to Generative AI Misinformation: Results from a Meta-Analysis of Scientific Evidence

This Summary for Policymakers provides a high-level précis of the synthesis report, "Confronting Misinformation Produced with Generative AI: A Meta-Analysis of Experimental Scientific Evidence."

Generative artificial intelligence (GenAI) can now rapidly produce large volumes of misleading text, images, audio, and video with readily accessible tools. This matters for policymakers because false or misleading content can be tailored to specific audiences. Such content can spread quickly and mimic authentic communication, making it difficult for people to judge what is true.

This Summary for Policymakers distills the main findings of the IPIE Synthesis Report (SR2026.2) on the effects of misinformation produced with GenAI and the measures most likely to reduce its influence. The assessment draws on a large-scale meta-analysis of experimental scientific evidence. It is based on 60 randomized controlled trial effect estimates from 24 peer-reviewed publications involving 33,801 participants, published between 2018 and 2025.

The report reaches four main conclusions:

1. Text-based GenAI misinformation currently poses greater persuasive risks than visual misinformation.
2. The current evidence base excludes most of the world. Research is concentrated in English-language and high-income countries, leavingsignificant gaps in our knowledge.
3. The most consistently effective intervention is to provide users with corrective information, preventively, so they can evaluate accuracy andcredibility themselves.
4. Labeling is effective when it is consistent. Content labeling generally decreases perceived credibility, but its impact varies greatly depending on how the firms create and apply labels.

‍

The policy implications are clear. Policymakers should prioritize addressing misinformation from text-based GenAI and support preventive and corrective information as a fundamental strategy. Labeling should be treated as a measure that requires careful design and testing. We must expand independent research well beyond English-language and high-income contexts.

Researcher access to platform and model data is crucial for enhancing public understanding of GenAI misinformation and for testing which safeguards are effective in practice.

‍

ISBN: 978-3-03983-016-9

DOI: 10.61452/QXAF2136

Citation: International Panel on the Information Environment [A. Herasimenka, S. Valenzuela, S. Boulianne, F. Esser, L. M. Given, S. Lewandowsky, E. M. NavarroLópez, P. N. Howard (eds.)], “Responding to Generative AI Misinformation: Results from a Scientific Meta-Analysis,” Zurich, Switzerland: IPIE, 2026. Summary for Policymakers, SFP2026.2, doi: 10.61452/QXAF2136.

Appendix: Download the Online Supplemental Appendices PDF.

‍