How Traders Use AI: TU Research

The rapid rise of AI and algorithmic trading has reshaped financial markets. Today, automation is no longer a niche – it is the dominant mode of execution.

However, TU research suggests a critical paradox: AI is widely accessible, but not widely effective for retail traders. This study examines how traders actually use AI tools and whether those tools improve performance.

The study focuses on four key questions:

1. How widely is AI adopted by retail traders?

2. How is AI used in practice?

3. Does AI improve profitability?

4. How does trader behaviour affect AI results?

Findings

Institutional validation

Institutional and academic research strongly supports the patterns identified in TU research. It confirms that while AI and algorithmic trading are rapidly expanding, their effectiveness depends less on the tools themselves and more on infrastructure, execution quality, and access to high-quality data.

According to the Bank for International Settlements – “Intelligent financial system: how AI is transforming finance” (BIS Working Paper No. 1194, 2024), AI significantly enhances the financial system’s ability to process data, detect patterns, and automate decision-making. At the same time, the report highlights increasing complexity, reliance on large datasets, and systemic risks associated with AI-driven trading.

The IMF Global Financial Stability Report – Chapter 3 “Advances in Artificial Intelligence: Implications for Capital Markets” (2024) shows that AI adoption in trading is accelerating and already influencing pricing dynamics, market structure, and the speed of information incorporation. The report notes that AI-driven strategies are improving market efficiency but also increasing correlations and turnover.

Recent academic evidence from the National Bureau of Economic Research – “AI-Powered Trading, Algorithmic Collusion, and Price Efficiency” (2025) demonstrates that AI trading agents can significantly influence market behaviour. The study finds that reinforcement learning systems can independently develop coordinated trading behaviour, raising concerns about market efficiency and unintended outcomes.

A large-scale review published in ScienceDirect – “Artificial Intelligence Techniques in Financial Trading: A Systematic Literature Review” (2024), covering 143 studies, shows that deep learning models dominate modern trading systems. However, only about 16% of systems achieve full automation, indicating that most AI implementations remain partially dependent on human input.

Further research in ScienceDirect – “Deep Learning for Algorithmic Trading” (2025) highlights that while AI improves predictive capabilities and adaptability, performance is highly sensitive to data quality, overfitting, and model stability.

Additional insights from Springer – “AI-Powered Systems for Algorithmic Trading” (2025) emphasise that despite rapid technological progress, AI trading systems still face key limitations related to data availability, computational complexity, and regulatory constraints.

Key takeaways

Across institutional and academic sources, several consistent conclusions emerge:

• AI and algorithmic trading are rapidly expanding and reshaping financial markets;

• Institutional adoption is the primary driver of this growth;

• Deep learning and advanced models dominate modern trading systems;

• Only a minority of systems achieve full automation in real-world conditions;

• AI improves efficiency, speed, and price discovery – but also introduces new risks (e.g., collusion, instability, and opacity).

At the same time, these findings imply that:

• access to AI tools alone does not guarantee better trading performance;

• the execution environment (data quality, latency, infrastructure) is the key factor;

• the gap between institutional and retail traders remains structural rather than technological;

• AI is most effective when integrated into a complete system (data → model → execution), rather than used as a standalone tool.

Theoretical research

From a structural perspective, the use of AI in trading is shaped by three key factors:

1. Level of automation. While AI systems are widely used in trading, only ~16% operates fully autonomously. Most solutions still require human oversight, adjustment, or intervention at different stages of execution.

2. Real-world performance. Although many AI models demonstrate strong results in controlled or theoretical environments, their effectiveness often declines in live market conditions due to noise, changing dynamics, and execution constraints.

3. Risk and complexity. The integration of AI increases both system complexity and exposure to new types of risks, including model instability, overfitting, and unintended behaviour under stress conditions.

Academic and market studies confirm that:

• most AI trading systems are not fully automated and rely on human input;

• theoretical model performance does not consistently translate into real trading profitability;

• AI adoption improves efficiency but also introduces additional layers of risk and complexity;

• retail traders often underestimate these risks when applying AI-based tools in practice.

Survey data

To evaluate how effectively retail traders use AI and algorithmic tools in real trading conditions, we conducted a proprietary quantitative study focused on adoption, usage patterns, and performance outcomes.

Methodology

The research was based on a structured online survey conducted among retail traders, using the CAWI (Computer-Assisted Web Interviewing) methodology. This approach ensured standardized data collection and consistency across different regions and respondent groups.

• Sample size: 1,020 retail traders

• Geography: global (North America, Europe, Asia)

• Experience level: beginner to intermediate (minimum 6 months of trading activity)

• Confidence level: 95%

• Margin of error: ±3.0%

Participants were selected based on active involvement in trading, with a focus on their use of AI tools, algorithmic strategies, and perceived impact on performance. The survey examined adoption rates, practical usage patterns, and the relationship between AI usage and trading outcomes.

Research team

The study was conducted by the analytical team at Traders Union:

• Anastasiia Chabaniuk (Author, TU Research) – research design and interpretation.

• Chinmay Soni (Fact-checker) – data validation and statistical verification.

• Dan Blystone (Editor-in-Chief) – editorial and methodological supervision.

• TU Research Team (Anton Kharitonov, Viktoras Karapetjanc) – data collection and analysis.

AI usage

To understand how widely AI tools are adopted in retail trading, the survey examined the level of engagement with AI-based solutions among participants.

Insight: The results indicate that AI adoption has already become mainstream among retail traders, with more than half of respondents using AI tools on a regular basis. However, a significant share of traders either use AI inconsistently or have only experimented with it, suggesting that widespread adoption does not necessarily translate into effective or systematic use.

Type of AI usage

To better understand how traders apply AI in practice, the survey examined the main types of AI-based tools and strategies used by respondents.

Types of AI usage in trading:

• Signal-based tools – 46%.

• Semi-automated bots – 32%.

• Fully automated systems – 22%.

Insight: The data shows that most traders use AI as a supporting tool rather than a fully automated solution. Signal-based tools dominate, indicating a preference for decision assistance, while fully automated systems remain less common, reflecting both higher complexity and lower accessibility for retail users.

Impact on profitability

To assess whether AI tools translate into real trading results, the survey examined how their use affects trader profitability.

Insight: The findings show that most traders do not achieve a measurable benefit from using AI tools. While a minority report improved performance, the majority see no meaningful change or even worse results, suggesting that access to AI alone is not sufficient to improve trading outcomes.

Behavioral factor

To understand how trader behaviour affects the effectiveness of AI tools, the survey examined common patterns in how users interact with algorithmic systems.

Behavioral impact on AI trading:

• Override AI decisions – 61%.

• Stop using AI after losses – 48%.

Insight: The results show that human behaviour significantly reduces the effectiveness of algorithmic trading. Frequent overrides and inconsistent usage patterns disrupt system logic, limiting the potential benefits of AI even when the underlying models are sound.

Practical implications for retail traders

To use AI effectively in trading, retail participants need to shift their approach from tool usage to system thinking. The following principles can help improve outcomes:

• Treat AI as a system, not a shortcut. AI-based trading requires structure, testing, and consistency. Institutional players integrate AI into full pipelines (data → model → execution), while retail traders often rely on isolated tools. Without a structured approach, AI becomes just another indicator rather than a performance driver.

• Avoid constant intervention. Frequent manual overrides disrupt the logic of algorithmic systems. Research shows that inconsistent human intervention reduces model efficiency and introduces behavioural bias, often turning statistically sound strategies into unstable ones.

• Validate strategies in real conditions. Backtesting is not enough. Many AI models perform well in simulations but degrade in live markets due to noise and changing conditions. Always test strategies with small capital and monitor real execution before scaling.

• Understand the limitations of AI. It cannot eliminate losses or predict markets with precision. Overreliance on AI tools often leads to overtrading and poor risk management, especially during volatile periods.

• Align AI usage with market conditions. AI models are sensitive to regime changes. Strategies that perform well in trending markets may fail in ranging or high-volatility environments. Continuous monitoring and adaptation are essential for maintaining performance.

• Prioritise risk management over optimisation. Institutional research consistently shows that risk control has a greater impact on long-term performance than model optimisation. Position sizing, drawdown limits, and discipline remain more important than the complexity of the algorithm.

• Focus on execution, not signals. The real institutional edge comes from infrastructure: low-latency execution, high-quality data, and stable environments. AI-generated signals alone do not provide a consistent advantage if execution conditions (slippage, spreads, delays) are poor.

From a practical standpoint, this means that the effectiveness of AI in trading is not determined by the algorithm alone, but by the conditions under which it operates. Even a well-designed model can underperform if execution quality is poor or market access is limited.

In retail trading, these conditions largely depend on the broker or platform used, including factors such as spreads, execution speed, slippage, and system stability during volatile periods.

Below is a comparison of the best Forex brokers that provide trading environments suitable for algorithmic and AI-based strategies:

Data sources and methodology references

Bank for International Settlements (BIS). (2024). Intelligent financial system: how AI is transforming finance (Working Paper No. 1194).

International Monetary Fund (IMF). (2024). Global Financial Stability Report – Chapter 3: Advances in Artificial Intelligence: Implications for Capital Markets.

National Bureau of Economic Research (NBER). (2025). AI-Powered Trading, Algorithmic Collusion, and Price Efficiency.

ScienceDirect (Elsevier). (2024). Artificial Intelligence Techniques in Financial Trading: A Systematic Literature Review.

ScienceDirect (Elsevier). (2025). Deep Learning for Algorithmic Trading: A Systematic Review of Predictive Models and Optimization Strategies.

Springer. (2025). AI-Powered Systems for Algorithmic Trading: Models, Data and Challenges.

OECD. (2024). Artificial Intelligence in Finance: Market Developments and Policy Considerations.

European Central Bank (ECB). (2024). Artificial Intelligence and the Future of Financial Markets.

Statista. (2025). Algorithmic trading market size and adoption trends.

IdSurvey. (2025). CAWI Methodology – Computer Assisted Web Interviewing.

Previous volumes in this series

• How Retail Traders Trade Gold by Time of Day: TU Research

• How Retail Investors Actually Trade Crypto: 2026 Research

Conclusion

Did you like the article?

Thank you for your feedback!

What could we improve?

send

Editors' Top Picks and Insights

2 hours ago Johnathan Maverick

How to build wealth from scratch in 3 practical steps

#Finance #Apple #NVDA

5 hours ago Eugene Komchuk

Kospi Index crash: Why South Korean market fell alongside AI stocks

#Finance #Stocks #Stock indexes

1 day ago Mira Kyivska

Bitcoin or Ferrari: Which investment is better?

#Crypto #Bitcoin

2 days ago Mira Kyivska

Strategy sells Bitcoin: Small sale tests market confidence

#Michael Saylor #Crypto #Strategy #Bitcoin

3 days ago Mikhail Vnuchkov

Ledger vs. Trezor: Search for ideal crypto wallet

#Ledger #Crypto #Trezor Wallet

4 days ago Oleg Tkachenko

Trading thin air: Why Binance is closing its NFT marketplace

#NFT #Binance

All news

Related Articles

What Is Algorithmic Trading? Definition And Key Concepts

Parshwa Turakhiya

4 days ago

AI Trading Bots: Do They Really Work?

Joshua Adakole

4 days ago

Best AI Trading Software (2026)

Rinat Gismatullin

4 days ago

Best Algorithmic Trading Strategies

Parshwa Turakhiya

4 days ago

Best Algorithmic Trading Software for Beginners in 2026

Mikhail Vnuchkov

4 days ago

Team that worked on the article

Anastasiia Chabaniuk

Educational Content Editor

Anastasiia has 17 years of experience in finance and content marketing. She believes that the support of information and expert opinion is very important for the success of investors and new traders.

Learn about our editorial policies

Dan Blystone

Senior English Editor

Dan Blystone began his trading career in 1998 as an arbitrage clerk on the floor of the Chicago Mercantile Exchange (CME). He later traded bond and Eurex futures at proprietary firms such as Altea Trading, gaining valuable experience in high-frequency trading and risk management.

Chinmay Soni

Head of Fact-Checking Department

Chinmay Soni is a financial analyst with more than 5 years of experience in working with stocks, Forex, derivatives, and other assets. As a founder of a boutique research firm and an active researcher, he covers various industries and fields, providing insights backed by statistical data.