The rapid rise of artificial intelligence (AI) has sparked intense debate about its economic and financial implications. A growing body of research has examined AI’s potential to boost productivity (Filippucci et al. 2024), transform labour markets (Hui et al. 2023), and drive technological innovation (Rodríguez-Pose and You 2024). Recent evidence also highlights its effects on financial markets: Eisfeldt et al. (2023) find that the release of ChatGPT significantly increased the value of firms whose workforces are more exposed to generative AI (GenAI). While these authors focused on workforce exposure to GenAI, in a recent paper (Ca’Zorzi et al. 2025) we offer a complementary perspective by examining how firms’ communication about GenAI influences stock markets.
A new framework for measuring GenAI exposure and sentiment
To study how corporate communication shapes investor responses, we draw on the emerging field of sentometrics, which applies econometric methods to quantitative information extracted from text. Recent contributions illustrate the value of sentiment analysis: it can help forecast economic activity from news and social media (Algaba et al. 2020), measure firm-level exposure to climate change from corporate disclosures (Sautner et al. 2023), and explain investment behaviour in fossil fuel firms (Adolfsen et al. 2024). These studies show that once qualitative information is quantified, it can provide explanatory power beyond traditional financial and macroeconomic indicators.
More recent applications highlight the policy relevance of this approach. Anastasiou et al. (2025) suggest that the emotional tone of Federal Reserve press conferences affects the probability of a fall in US bank stock prices. Culver et al. (2025) construct a sentiment-based index of geopolitical risk at the industry level, showing that industries with more negative sentiment experience larger stock price declines during periods of heightened geopolitical risk.
Building on this literature, we examine how firms communicate about new technologies in earnings calls and how investors price these discussions. We analyse more than 22,000 transcripts from S&P 500 firms covering 2014–2024. GenAI exposure is measured as the share of each call devoted to GenAI-related topics. Using GPT-4o, we further classify the content into three key themes, Opportunity, Adoption and Risk, allowing us to capture both the focus and tone of corporate discussions and link these dimensions systematically to stock market reactions.
Key findings
Our analysis yields three main insights:
First, there was a surge in GenAI exposure post-ChatGPT. Mentions of GenAI rose sharply beginning in late 2022, particularly in the information technology sector (from 1.3% in 2022 to 5.0% in 2023; see Figure 1a). Other sectors, such as Communication Services and Consumer Discretionary, also saw increases. Firms differed widely in engagement, even within the same sector. Over time, the focus shifted from opportunities to adoption, showing a stronger emphasis on implementation after the release of ChatGPT, particularly in the tech sector (Figure 1b).
Figure 1 Average GenAI exposure across the sample (left panel) and sentiment analysis breakdown (right panel)
Note: The vertical dashed red line indicates ChatGPT’s release in Q4 2022, coinciding with substantial increases in exposure, particularly in technology.
Second, early engagement paid off. Firms that discussed GenAI before the release of ChatGPT (“Early Exposed”) saw higher stock returns. Panel regressions indicate that a one percentage point increase in GenAI discussions was, on average, associated with a 0.62% rise in quarterly stock prices. Investors were eager to back firms that identified AI as a key factor for maintaining a competitive edge and driving future growth. We also find a positive interaction coefficient of 0.26 between being Early Exposed and GenAI engagement, indicating that firms with both early exposure and sustained focus on GenAI have benefited in terms of equity performance.
Figure 2 illustrates the estimated impact of GenAI on stock prices, based on regression coefficients and cumulative changes in exposure in 2023. This analysis suggests that Early Exposed firms significantly outperformed the broader market in equity returns, with roughly one-third of the gains attributable to their GenAI discussions.
Figure 2 Equity returns and estimated impact of GenAI, 2023
Note: Blue bars show actual 2023 equity returns, while red bars show the estimated GenAI contribution based on regression coefficients and exposure changes. Early Exposed firms show the strongest effects, with GenAI accounting for approximately 40% of their returns.
Third, tone matters. Investors responded differently depending on the theme of the discussion. Early Exposed firms outperformed lagging firms (“Laggards”) by 0.63% and 0.45% in quarterly excess returns for each 1% increase in adoption or opportunity-related exposure, respectively. The stronger effect for adoption suggests markets rewarded more discussions about implementation over speculative opportunities.
A generalisable approach
Our framework also captures corporate attention across major global events. Figure 3 illustrates firms’ focus on Brexit, COVID-19, the 2021-2023 inflation surge, and geopolitical and climate risks. These examples show how the approach, applied beyond GenAI, provides policymakers and analysts with a real-time tool to monitor emerging global changes and evaluate potential market and macroeconomic effects.
Figure 3 Firms’ exposure to a broad range of global topics
References
Adolfsen, J F, M Heissel, A-S Manu and F Vinci (2024), “Burn Now or Never? Climate Change Exposure and Investment of Fossil Fuel Firms”, ECB Working Paper Series, No. 2945.
Algaba, A, D Ardia, K Bluteau, S Borms and K Boudt (2020), “Econometrics Meets Sentiment: An Overview of Methodology and Applications”, Journal of Economic Surveys 34(3): 512-547.
Anastasiou, D, A Katsafados, S Ongena and C Tzomakas (2025), “Beyond Words: Fed Chair Voice Sentiments and US Bank Stock Price Crash Risk”, VoxEU.org, 19 June.
Ca’ Zorzi, M, G Lopardo and A-S Manu (2025), “Verba Volant, Transcripta Manent: What Corporate Earnings Calls Reveal About the AI Stock Rally”, ECB Working Paper Series No. 3093.
Culver, I, F Niepmann and L Shen (2025), “Measuring Geopolitical Risk Exposure Across Industries: A Firm-Centered Approach”, FEDS Notes, 29 August, Board of Governors of the Federal Reserve System.
Eisfeldt, A, G Schubert and M B Zhang (2023), “Generative AI and Firm Valuation”, VoxEU.org, 4 June.
Filippucci, F, P Gal and M Schief (2024), “Miracle or Myth: Assessing the Macroeconomic Productivity Gains from Artificial Intelligence”, VoxEU.org, 8 December.
Hui, X, O Reshef and L Zhou (2023), “Artificial Intelligence and Its Short-Term Effects on Employment”, VoxEU.org, 1 December.
Rodríguez-Pose, A and Z You (2024), “Bridging the Innovation Gap: AI and Robotics as Drivers of China’s Urban Innovation”, VoxEU.org, 5 June.
Sautner, Z, L Van Lent, G Vilkov and R Zhang (2023), “Firm-Level Climate Change Exposure”, The Journal of Finance 78(3): 1449–1498.