Labeling content as AI-generated or AI-enhanced reduces user engagement, particularly for emotional posts; delaying disclosure helps only for AI-enhanced content, not fully for AI-generated pieces.

Main Finding

Labeling social media posts as AI-enhanced or AI-generated reduces both affective and behavioral user engagement relative to human-created labels—especially for emotionally framed content. Affective engagement mediates the drop in behavioral engagement. Timing of disclosure matters: late disclosure improves affective engagement for AI-enhanced (hybrid) content but not for fully AI-generated content.

Reference: Seeger, Wessel & Lehrer (2026), Electronic Markets. DOI: 10.1007/s12525-026-00883-2.

Key Points

• Continuum of AI involvement: human-created → AI-enhanced (human + AI) → AI-generated (fully AI).
• Algorithmic aversion and human favoritism both shape user responses: users penalize AI-labeled content and prefer human authorship.
• Effects are content-dependent: emotional (affect-driven) posts suffer larger engagement declines from AI labels than rational (utilitarian/informational) posts.
• Disclosure timing moderates reactions:
  • Late disclosure raises affective engagement for AI-enhanced content (suggesting preserved sense of human agency), but
  • Late disclosure does not improve engagement for fully AI-generated content.
• Psychological mechanism: AI labeling lowers affective engagement (emotional connection), which in turn reduces behavioral engagement (liking, commenting, sharing intentions).

Data & Methods

• Design: Two online experimental studies using simulated Instagram profiles.
• Samples: Study 1 n = 325; Study 2 n = 371. Recruited via Prolific. Participant mean age ≈ 35 years; 55% female.
• Manipulations:
  • AI level: human-created vs. AI-enhanced vs. AI-generated.
  • Content type: emotional vs. rational visual posts.
  • Disclosure timing: early vs. late disclosure of AI involvement.
• Measures: Self-reported affective engagement and behavioral engagement (intentions to like/comment/share). Cognitive engagement was not measured.
• Analysis: Between-subjects comparisons, mediation tests to assess affective engagement as mediator, and moderation analyses for content type and disclosure timing.

Implications for AI Economics

• Platform revenue and attention economics:
  • Reduced engagement from AI labeling (especially for emotional content) can lower metrics that drive ad revenue and network effects. Platforms must account for label-driven engagement declines in revenue forecasts and platform-design choices.
• Incentives for creators and content supply:
  • Creators relying on generative AI may face lower engagement returns for emotionally driven posts, shifting incentives toward human-authored or hybrid outputs for such content, or toward AI use in rational/content-light contexts where penalties are smaller.
• Labeling policy and compliance trade-offs:
  • Mandatory, early, or prominent AI disclosure (as in some regulatory regimes) can undermine engagement. Regulators and platforms need to balance transparency goals with economic impacts on attention markets. The finding that late disclosure mitigates harm for AI-enhanced but not fully AI-generated content highlights the possibility of nuanced disclosure requirements (level-of-AI granularity, timing rules) rather than one-size-fits-all mandates.
• Product design and market segmentation:
  • Platforms can segment features: encourage transparent AI use in informational/rational content (where engagement loss is limited), and support verified human-authorship signals or hybrid-attribution conventions for emotional content to preserve engagement.
• Moral hazard and reporting behavior:
  • If disclosure reduces engagement, creators may underreport AI involvement (adverse selection). Markets may evolve verification/credential systems (auditing, provenance tech) that influence trust and pricing of attention.
• Modeling consequences:
  • Economists modeling platform competition, creator incentives, and ad markets should incorporate label-driven demand elasticity for content, heterogeneity by content type, and timing-sensitive disclosure effects.
• Research and policy priorities:
  • Need for field/longitudinal studies to measure real-world effects on actual posting behavior, platform metrics, and creators’ monetization. Cost–benefit analyses of disclosure regulations should factor in these engagement externalities.

Limitations to consider: lab/online-experimental setting with simulated profiles and self-reported engagement; Prolific sample may not fully represent platform user populations; short-term exposure—long-run adaptation and learning effects are untested.

Practical takeaway: For platforms and creators navigating AI integration, preserve emotional authenticity signals (human authorship or clear human-AI collaboration), be mindful that disclosure timing and label granularity materially affect engagement outcomes, and anticipate differential economic impacts across content types.