Experiences with AI-Generated Pornography (link.springer.com)

Researchers analyzed 390 English-language Reddit posts (collected August 2024) to map everyday experiences with AI-generated pornography across a communication-process framework: production, content, use, effects, and ethical–legal implications. Using a reliability-tested manual content analysis (mean Cohen’s κ = 0.88), they found that discussions most often centered on production (59.5%) and content (60.8%), while firsthand use reports were relatively rare (12.8%); effects (37.2%) and ethical–legal concerns (35.1%) appeared in roughly one-third of posts. Redditors’ accounts ranged from outrage and concern—especially about non-consensual deepfakes—to curiosity, sexual enjoyment, and perceived economic opportunities tied to AI tools. Technically, the study highlights two distinct phenomena: entirely fictional, multimodal AI pornography (text, audio, images, agents) and deepfake pornography that repurposes real people’s likenesses via deep learning (often requiring only a few photos). It documents a split ecosystem: mainstream generative models (e.g., ChatGPT, Gemini, DALL‑E) with explicit moderation versus an emergent market of “unmoderated” web/apps and local models that can bypass safeguards. The findings underscore the field’s research gap beyond non-consensual deepfakes and call for nuanced responses from policymakers, platform designers, educators, and mental-health professionals to address consent, moderation, detection, and user harms.