Hype about AI widens corporate technology gaps at first but can trigger catch-up: modest 'AI washing' suppresses innovation, yet beyond a tipping point firms ramp R&D and narrow the gap; participatory learning eases the harm while high investor sentiment deepens it.

Main Finding

There is a significant inverted U‑shaped relationship between “AI washing” (the gap between AI narratives and actual capabilities) and a firm’s technological gap: at low-to-moderate levels, AI washing widens the technological gap (stifling real innovation), but beyond a threshold firms respond/adapt and the gap narrows. This nonlinear effect is mediated by both firm-level and industry-level R&D investment, moderated by firm absorptive/participatory learning capability and investor sentiment, and attenuated in national AI pilot zones and high‑technology‑intensive industries.

Key Points

• The study applies Cognitive Dissonance Theory to explain firm behavior when AI narratives diverge from capabilities: initial dissonance encourages posturing rather than genuine investment, but sustained dissonance spurs corrective investment past a tipping point.
• Inverted U‑shape: AI washing → (initially) increases technological gap → (after threshold) decreases technological gap as firms adapt.
• Mediation: internal R&D investment and industry‑level R&D investment are the primary channels through which AI washing affects the technological gap.
• Moderation:
  • Strong AI‑enabled participatory learning and absorptive capability flatten the inverted U (firms correct earlier, reducing the peak gap).
  • High investor sentiment steepens the inverted U (amplifies the early widening of the gap).
• Heterogeneity: the nonlinear effect is weaker for firms located in national AI pilot zones and for firms in high‑technology‑intensive industries.
• Contributions: provides quantitative evidence on “AI washing,” extends Cognitive Dissonance Theory to technology strategy, and informs responsible AI governance.

Data & Methods

• Sample: panel data of China’s A‑share listed firms spanning 2007–2022.
• Empirical approach: panel econometric analyses testing a nonlinear (quadratic) relationship between AI washing and technological gap; mediation analysis to test R&D channels; moderation tests for firm capabilities and market sentiment; subgroup analyses for policy and industry contexts.
• Variables (as described): AI washing measured as discrepancy between AI narratives and actual AI capability; technological gap measured at the firm level; mediators—firm R&D investment and industry R&D investment; moderators—AI participatory learning/absorptive capability and investor sentiment; controls for firm and industry traits and likely fixed effects/time controls.
• Robustness: analyses include moderation and heterogeneity checks across regulatory (national AI pilot zones) and industry‑technology intensity dimensions (as reported).

Implications for AI Economics

• Resource allocation and market dynamics: AI washing can misallocate capital and managerial attention, creating short‑term valuation effects and raising the risk of mispricing until firms or markets adjust—this explains some boom‑and‑bust dynamics around AI narratives.
• Firm strategy: firms should invest in genuine absorptive capacity and participatory learning to detect and correct AI washing early, flattening and shortening the period of harmful divergence.
• Investor behavior: high investor sentiment can exacerbate AI washing’s harm; investors need better informational standards and due diligence to avoid funding narrative over substance.
• Policy & governance: regulatory interventions (e.g., pilot zones, disclosure standards, R&D incentives) can dampen the adverse effects of AI washing and speed corrective adaptation; transparency and verification of AI claims reduce cognitive dissonance costs.
• Research directions: quantify welfare and macroeconomic consequences of AI washing-driven misallocation, extend cross‑country comparisons, and evaluate regulatory tools (disclosures, audits) for mitigating narrative‑driven distortions.