A mandatory AI label on Kickstarter slashes project funding by nearly 40% and backer numbers by roughly 24%; transparent, authentic disclosures soften the blow, but overstated positive emotion makes investors more wary.

Main Finding

Mandatory AI disclosure on Kickstarter substantially reduced crowdfunding performance for AI-involved projects (−39.8% in funds raised; −23.9% in backer counts). However, the magnitude of this adverse effect depends on disclosure strategy: greater AI involvement amplifies the negative impact, whereas high explicitness (logos) and high authenticity (ethos) mitigate it. Excessively positive emotional tone (pathos) backfires and further worsens outcomes. Experimental evidence shows these effects operate mainly through two mechanisms: perceived creator competence and AI-washing concerns.

Key Points

• Natural experiment: Kickstarter’s August 2023 mandatory AI-disclosure policy provides causal leverage.
• Average treatment effect: AI disclosure → −39.8% funds raised; −23.9% backers for AI projects.
• Substantive signal (degree of AI involvement):
  • Higher AI involvement increases negative funding effects.
  • Works primarily by lowering perceived creator competence.
• Rhetorical signals (Aristotelian triangle):
  • Logos (explicitness/detail): mitigates negative effects by increasing competence perceptions and reducing AI-washing concerns.
  • Ethos (authenticity/credibility): mitigates negative effects mainly by reducing AI-washing concerns, which then increases perceived competence (serial mediation).
  • Pathos (positive emotional tone): too much positivity increases AI-washing concerns, reduces perceived competence, and lowers pledge intention (serial mediation).
• Four preregistered between-subjects experiments on Prolific causally confirm moderators and identify mediation pathways.
• Overall conceptual advance: moves beyond binary “disclose vs. not” to “how to disclose,” distinguishing substantive vs. rhetorical signals and their distinct psychological pathways.

Data & Methods

• Primary empirical strategy:
  • Difference-in-differences (DID) leveraging Kickstarter’s mandatory AI disclosure rollout (August 2023) as a natural experiment.
  • Identification of AI-involved projects via keyword-based text classification following prior work.
  • Comparison: AI-flagged projects (treatment) vs. other projects (control) before vs. after policy.
• Main outcome measures: funds raised and number of backers.
• Supplementary causal tests:
  • Four randomized online experiments (Prolific), each manipulating one disclosure dimension (AI involvement, explicitness, authenticity, emotional tone) in a between-subjects design.
  • Manipulations built from real Kickstarter disclosure text; non-focal content minimally altered using LLMs to invert the focal dimension.
  • Measures: pledge intention, perceived creator competence, AI-washing concerns; mediation analyses (including serial mediation) to trace mechanisms.
• Robustness: combined observational DID with randomized experiments to establish external validity (platform outcomes) and internal causal pathways (psychological mediators).

Implications for AI Economics

• Policy trade-offs: Mandatory transparency can reduce funding for AI-related entrepreneurial projects—regulators should weigh oversight benefits against potential dampening of early-stage financing, especially for projects with high AI integration.
• Disclosure design matters more than disclosure existence:
  • Platforms and regulators should require structured, specific disclosure fields (encouraging logos) and discourage vague or marketing-heavy labels that fuel AI-washing suspicions.
  • Encouraging or standardizing authenticity cues (e.g., provenance of content, concrete role of human creators, verifiable development milestones) can mitigate investor skepticism.
  • Guidance should warn against overpositive emotional framing in required disclosures, which may increase perceptions of deception.
• For entrepreneurs:
  • When AI plays a large role, provide concrete, technical, and verifiable descriptions of the AI’s role and safeguards to preserve perceived competence.
  • Emphasize authenticity (evidence, process transparency, human oversight) rather than excessive cheerleading.
• Platform design and market infrastructure:
  • Develop template-based, standardized AI-disclosure forms that collect substantive details (degree of automation, tasks performed by AI vs. humans) and permit third-party attestations—this reduces ambiguity and AI-washing concerns.
  • Consider verification/credentialing mechanisms (audit trails, model provenance, independent attestations) to counterbalance the negative funding effects of disclosure and support innovation.
• Theoretical contribution:
  • Extends signaling theory in digital markets by showing AI disclosures simultaneously activate competence and deception (AI-washing) inferences; substantive and rhetorical signals affect these channels differently.
• Future research/policy needs:
  • Investigate long-term effects on innovation rates and entrepreneurial entry.
  • Explore sector-specific heterogeneity (creative vs. technical projects) and the role of complementary verifiable signals (third-party audits, certifications).