Abstract
Artificial Intelligence (AI) has emerged as a General Purpose Technology (GPT), yet significant uncertainty around firms’ AI investment remains because the complementary investments required to realize its returns are largely intangible and difficult to verify. We examine the characteristics of firms that disclose AI activities, and whether these disclosures provide credible signals of firms’ underlying AI investments and their future economic outcomes. We document several key findings. First, firms that disclose AI activities are generally innovative, operate in AI-intensive industries, and face greater investor scrutiny. Second, AI disclosures are positively associated with subsequent operational efficiency and AI patent filings, even after accounting for firms’ investment in AI human capital. This indicates that disclosure not only provides a signal about a firm’s AI investment levels, but also about managers’ confidence in the expected returns of the investment. In cross-sectional tests, we find that AI disclosure appears to provide stronger signals when disclosure is about the use of AI in a firm’s products and services and when firms operate in industries that are less expected to use AI, and weaker signals when firms appear to be outsourcing their AI capabilities. Overall, our findings suggest that AI disclosures are informative about firms’ underlying AI investments and managers’ confidence in their payoffs, and that financial statement users could benefit from AI human capital disclosure.
Key Findings
Who Discloses AI
Across 11,927 firm-years (2016–2023), firms that disclose AI are innovative, R&D-intensive, and face greater investor scrutiny. A Shapley decomposition attributes 83–87% of the explained variation in AI disclosure to industry-level factors.
Disclosures Track Real Investment
AI disclosures are positively associated with firms’ subsequent operational efficiency and AI patent filings—even after controlling for AI employment (measured from Revelio workforce data)—signaling both AI investment levels and managers’ confidence in its expected returns.
When Signals Are Stronger
The signal is roughly twice as strong for product- and service-related AI disclosure as for generic operational disclosure, and stronger in industries less expected to use AI; it weakens, but stays positive, when firms appear to outsource their AI.
Complements, and Informative
Disclosure and AI hiring act as complements: firms high on both show the largest efficiency and patenting gains. Overall, AI disclosures are informative, and financial-statement users would benefit from AI human-capital disclosure.
Research Contribution
AI is a general-purpose technology, but the complementary investments needed to realize its returns are largely intangible and hard to verify—leaving investors uncertain about which firms are genuinely investing. The paper examines who discloses AI activity and whether those disclosures are credible signals of underlying investment and future performance.
The evidence indicates that AI disclosures carry real information: they are positively associated with subsequent operational efficiency and AI patent filings, and their signal strength varies with what is disclosed and the firm’s industry. The results speak to investors assessing firms’ AI claims, to standard-setters weighing AI and human-capital disclosure, and to research on how firms communicate about emerging technologies.