When AI reads the fine print, markets begin to shift
For years, financial markets have reflected a quiet imbalance rooted not in access to information, but in the ability to interpret it.
Institutional investors rely on teams and analytical tools to process complex disclosures, while retail investors often face the same documents with fewer resources to extract meaningful insight. Even when information is widely available, differences in interpretation can shape trading behavior in significant ways.
New research from the UB School of Management examines how generative AI is beginning to change that dynamic. Conducted by Toghrul Aghbabali, PhD student, Kee Chung, Louis M. Jacobs Professor of Financial Planning and Control, and Sahn-Wook Huh, Associate Professor of Finance, the study explores how tools like ChatGPT influence the way investors process public information and respond to it in financial markets.
At the center of the research is a new metric, the Coverage Score Average, or CSA. The measure captures how effectively an AI system can interpret a company’s financial disclosures and translate them into useful insight for a nonexpert user.
Firms with higher CSA scores tend to have more detailed and complex filings, where AI-assisted interpretation can add value. Firms with lower scores offer less opportunity for additional insight. This distinction allows researchers to examine where AI has the greatest potential impact.
The findings point to a consistent pattern. Following the public release of ChatGPT, firms with higher CSA scores experienced a measurable decline in informed trading. Indicators tied to information asymmetry and trading costs also decreased, while short-term return volatility became more stable.
These changes suggest that AI tools can help reduce the informational advantages traditionally held by more sophisticated investors. By improving the ability to interpret public disclosures, AI supports a more balanced understanding of available information across different types of market participants.
The study also finds that trading behavior becomes more aligned, particularly among retail investors. Measures of order imbalance increase, indicating that investors are more likely to respond to information in similar ways.
This shift reflects a change in how information is processed rather than how it is accessed. As more investors draw comparable conclusions from the same disclosures, variation in interpretation begins to narrow.
At the same time, the effects are not uniform across all situations. The research shows that AI has a stronger influence on trading tied to positive news than negative news. When evaluating favorable developments, investors are more likely to engage with analytical tools and incorporate structured insights into their decisions. In contrast, responses to negative information tend to be more immediate and less dependent on detailed analysis.
This distinction highlights the role of human behavior in shaping how AI is used. While generative AI can support interpretation and reduce cognitive barriers, it does not fully replace the behavioral factors that influence decision-making under uncertainty.
Overall, the findings suggest that generative AI contributes to improved market efficiency by enhancing the interpretation of existing information. Rather than introducing new data or private signals, these tools help a broader group of investors better understand what is already available.
As AI continues to evolve, its role in financial markets is likely to remain closely tied to this function. Improving how information is interpreted may prove as important as expanding access to it, particularly in environments where complexity has long limited participation.
This story was written by AI and edited by a member of the UB School of Management Marketing and Communications Office.