How AI Is Changing Market Volatility in 2025
Artificial intelligence now plays a central role in financial markets. Automated trading powered by large language models and predictive algorithms is reshaping volatility and liquidity patterns worldwide.
Artificial intelligence has moved beyond research labs and tech companies — it now drives a large share of financial market activity. From algorithmic trading to portfolio optimisation, AI is influencing how prices move, how fast markets react, and how liquidity behaves in 2025.
AI and the New Market Behaviour
The Bank for International Settlements (BIS) notes that more than 70 percent of global equity trades now involve algorithmic components, with AI systems increasingly integrated into decision-making processes. These models are not just faster — they are learning from real-time data and adapting to market patterns autonomously.
Source: BIS – Quarterly Review March 2025
This rapid adoption of generative and predictive AI has created both efficiency and fragility. While AI improves liquidity during stable periods, it can amplify volatility when many systems respond to the same triggers.
Herding Among Algorithms
In 2025, several sharp intraday swings across US and UK markets were traced to correlated AI trading models reacting to identical news sentiment data. A BIS study found that when machine learning models process similar training data, their decisions cluster — creating “synthetic herding.”
This effect was observed on 12 March 2025, when US tech shares experienced a 2.4 percent intraday reversal following a misinterpreted AI headline sentiment score related to cloud infrastructure.
Source: BIS – AI in Financial Markets Report 2025
The Rise of LLM-Driven Trading Signals
Large language models (LLMs) trained on financial data are now used to interpret earnings calls, regulatory filings, and social media sentiment. Firms including JPMorgan and Citadel have deployed AI systems capable of parsing thousands of documents per minute to detect tone, uncertainty, or forward-looking bias.
According to the CME Group, AI-driven sentiment analysis accounted for nearly 22 percent of futures trading volume growth between Q1 2024 and Q1 2025.
Source: CME Group – AI and Market Liquidity Insights 2025
New Volatility Patterns
AI systems process information in milliseconds, creating feedback loops between algorithms that operate faster than human oversight. This leads to:
- Shorter volatility spikes – markets adjust within minutes instead of hours.
- Shallower liquidity pools – liquidity appears abundant until multiple AIs withdraw simultaneously.
- False signal amplification – similar model architectures magnify sentiment errors.
The World Economic Forum (WEF) warns that this automation could make flash crashes more common, though shorter in duration. Regulators are studying circuit-breaker timing adjustments to address this.
Source: World Economic Forum – The Future of AI in Financial Systems 2025
Human Traders Still Matter
Despite automation, human oversight remains critical. The Financial Conduct Authority (FCA) reminds firms that accountability cannot be delegated to algorithms. Every AI-driven decision must have an auditable record linking back to a human supervisor.
Source: FCA – Algorithmic Trading Compliance Update 2025
Human traders are also adapting by using AI as a co-pilot — relying on systems for signal detection while exercising judgment in execution and portfolio management.
Risk and Regulation
The main risks identified by regulators include:
- Model concentration: too few vendors supplying AI engines.
- Data drift: outdated training sets producing flawed signals.
- Opacity: difficulty in auditing decision pathways of complex models.
The Bank of England’s Financial Policy Committee has launched an AI model register initiative for systemically important trading algorithms to improve transparency.
Source: Bank of England – Financial Stability Report 2025
The Bottom Line
AI is reshaping how volatility behaves — compressing reaction times, clustering decision-making, and creating new forms of systemic interdependence. While automation brings efficiency, it also introduces fragility when thousands of algorithms act on the same logic at once.
For investors, the takeaway is clear: technology may evolve, but diversification, human judgment, and regulatory oversight remain the best defences against the next AI-induced market shock.
References:
