Artificial intelligence is making financial markets more efficient by accelerating research, improving price discovery and narrowing information gaps, but its growing influence could also increase the risk of sharp market swings during periods of stress, according to a research note by Bernstein.
The research said AI-powered tools are transforming how investors analyse markets by processing corporate earnings reports, regulatory filings and alternative datasets at speeds far beyond traditional human analysis. As a result, analyst forecasts are converging more quickly with actual company performance, reducing earnings surprises and improving overall market efficiency.
Bernstein noted that AI is also reshaping the investment research process by automating routine tasks such as earnings reviews, financial model updates and the initiation of stock coverage. This has enabled analysts to monitor a broader range of companies, particularly smaller firms in emerging markets that have historically received limited attention.
The expansion in coverage of emerging market small-cap companies over the past year could gradually reduce the valuation premium often enjoyed by under-researched stocks, as investors gain access to more timely and comprehensive information.
Despite these benefits, Bernstein cautioned that the widespread adoption of similar AI models could introduce new systemic risks. As investors increasingly rely on comparable algorithms trained on overlapping datasets, trading strategies may become more closely aligned, leading to crowded positions across financial markets.
Such synchronised trading behaviour, the report warned, could amplify market reversals during periods of uncertainty, making volatility more severe when investors attempt to exit similar positions simultaneously.
Bernstein pointed to the unwinding of the yen carry trade in August 2024 and episodes of AI-generated misinformation that temporarily moved U.S. equity markets as examples of how automated trading systems and synthetic content can intensify market volatility before investors have sufficient time to verify information.
The report also highlighted what it described as the “reflexivity problem”, whereby AI-generated investment recommendations begin to shape investor behaviour, influencing asset prices in ways that reinforce future outputs from the same models.
According to Bernstein, this feedback loop could strengthen momentum-driven trading, increase market concentration and contribute to more extreme asset valuations over time.
While AI is expected to continue reducing average market inefficiencies through faster research, broader market coverage and improved trade execution, Bernstein believes the technology could simultaneously increase the likelihood of larger market dislocations during periods of financial stress.
The research concluded that although AI is making markets more efficient on average, investors and regulators should remain alert to the greater tail risks that may emerge as automated systems become more deeply embedded in global financial markets.
