The Verification Crisis: Expert Perceptions of GenAI Disinformation and the Case for Reproducible Provenance

TL;DR

This paper investigates a verification crisis in GenAI disinformation by presenting Wave 1 expert perceptions (N=$21$) from AI researchers, policymakers, and disinformation specialists. It identifies a multimodal threat landscape, with text posing a systemic risk of epistemic fragmentation and deepfake video delivering high shock value, while detecting AI-generated content remains unreliable. The authors argue for reproducible provenance as the infrastructural solution, integrating frameworks such as the Momeni‑Khan checklists, the C2PA standard, and the Methods Hub, complemented by knowledge graphs and longitudinal data to track narratives. Their proposed metrics—Manifest Robustness, Verification Latency, Chain Continuity, and the TOM Score—aim to render provenance and methods auditable, enabling policy and research to be grounded in reproducible evidence. The work offers a blueprint for operationalizing provenance as core defense against disinformation and calls for Wave 2 participation to advance an open, verifiable ecosystem for truth verification.

Abstract

The growth of Generative Artificial Intelligence (GenAI) has shifted disinformation production from manual fabrication to automated, large-scale manipulation. This article presents findings from the first wave of a longitudinal expert perception survey (N=21) involving AI researchers, policymakers, and disinformation specialists. It examines the perceived severity of multimodal threats -- text, image, audio, and video -- and evaluates current mitigation strategies. Results indicate that while deepfake video presents immediate "shock" value, large-scale text generation poses a systemic risk of "epistemic fragmentation" and "synthetic consensus," particularly in the political domain. The survey reveals skepticism about technical detection tools, with experts favoring provenance standards and regulatory frameworks despite implementation barriers. GenAI disinformation research requires reproducible methods. The current challenge is measurement: without standardized benchmarks and reproducibility checklists, tracking or countering synthetic media remains difficult. We propose treating information integrity as an infrastructure with rigor in data provenance and methodological reproducibility.

Figures (3)

• Figure 1: Threat Perception Heatmap. Heatmap of expert threat ratings showing the concentration of high-severity concerns in the Political domain compared to the Social domain. Note the skew towards "Severe Threat" (7) for Political Text and Video.
• Figure 2: Urgency Ranking. Top-3 Urgent Risks selected by experts, highlighting Election Interference (90.5%) and Text Spambots (61.9%) as the dominant concerns, far outpacing Phishing or Financial Fraud.
• Figure 3: The Mitigation Gap. Best-Worst Scaling results indicating that Technical Detection Tools are viewed as the least effective strategy (38% rated "Least Effective"), contradicting the current industry focus on "AI catchers." Provenance and Literacy are favored but face implementation hurdles.