DayTrading.com’s latest study highlights significant risks in using AI tools like ChatGPT, Claude, Perplexity, Gemini, Groq, and Meta AI for trading decisions. Testing these platforms with over 180 trading-related queries, researchers found frequent errors, from fabricated stock prices to misleading buy/sell recommendations, posing dangers for traders relying on these tools.
DayTrading.com’s study tested six AI tools on 180+ trading queries.
ChatGPT led with 85% accuracy, but Perplexity was most accurate at 91%.
Meta AI scored worst, with an 8.8/10 danger score and 68% accuracy.
AI tools often invent live market data, risking costly trading errors.
Trading ideas from AI were correct only 64% of the time.
AI is best used for summarizing reports, not live data or trade signals.
The study evaluated six popular AI platforms—ChatGPT, Claude, Perplexity, Gemini, Groq, and Meta AI—against common trading queries like EUR/USD prices, Fed statement summaries, and stock recommendations. Results showed a mix of strengths and critical weaknesses. “We’re seeing more traders turn to AI platforms for assistance, but they need to know that while the answers may sound persuasive, they can, and sometimes are, plain wrong,” said Paul Holmes, the report’s lead author.
ChatGPT topped the group with an 85% accuracy rate and a danger score of 5.2/10, balancing reliability with occasional errors. Claude followed with 89% accuracy but a higher danger score of 6.8/10 due to summarization issues. Perplexity, despite its 91% accuracy, struggled with real-time financial data, earning a 7.6/10 danger score. Groq (8.2/10, 72% accuracy) and Gemini (8.4/10, 81% accuracy) fabricated live stock prices, while Meta AI performed worst, with a danger score of 8.8/10 and 68% accuracy, issuing risky buy calls on incomplete data. “It’s a strange mix - the top-performing tool wasn’t the most accurate, and the most accurate wasn’t the safest,” Holmes noted.
The study pinpointed live market data as the riskiest area, with tools like Groq and Meta AI inventing stock prices, such as a nonexistent Tesla price. Trading ideas were another weak point, with even top models accurate only 64% of the time, insufficient to cover trading fees. “The riskiest moment isn’t when AI says ‘I don’t know,’” Holmes said. “It’s when it gives you a confident answer that’s completely wrong.” These persuasive errors pose significant risks, especially for novice traders.
The report advises using AI for preparatory tasks, such as summarizing Fed statements or earnings calls, rather than relying on it for live data or trading signals. AI’s strengths lie in condensing complex information, but its weaknesses in real-time accuracy and risk assessment make it unreliable as a primary decision-making tool. Holmes concluded, “AI in trading is a bit like a rookie trader with encyclopaedic knowledge and no risk management - brilliant one moment, reckless the next.”
DayTrading.com’s study underscores the need for traders to approach AI tools cautiously, using them as supportive resources rather than decision-making authorities. By understanding AI’s limitations, traders can avoid costly mistakes and leverage its strengths effectively.
DayTrading.com is a free online resource that provides guides for active traders, reviews of brokers and trading platforms, and market analysis that’s been cited in top media outlets and used by millions of traders worldwide.
TechIntelPro empowers B2B technology decision-makers with the latest industry trends, deep insights, leadership perspectives and actionable intelligence to drive both organizational and professional milestones—and thrive on the tech pulse.