Once upon a time, stock tips came from taxi drivers, colleagues at the pub, or the occasional market-obsessed friend. Today, a new source of investment advice is gaining attention: artificial intelligence chatbots such as ChatGPT and Claude.
These large language models are increasingly being used by both amateur and professional investors to generate stock ideas and analyse markets. Yet despite the enthusiasm, their ability to consistently outperform human judgement remains unproven.
The debate touches on a long-standing idea in finance. Economist Burton Malkiel famously argued in 1973 that a blindfolded monkey randomly selecting stocks could perform as well as professional investors, a cornerstone of the efficient market hypothesis. If AI systems can reliably beat markets, that theory would be seriously challenged.
Early evidence, however, is mixed. In a recent experiment conducted by US research lab Nof1, eight leading AI models were tasked with investing in US technology stocks. Six of them lost money. Anthropic’s Claude Sonnet recorded losses of nearly 60% on a $10,000 portfolio, while Google’s Gemini also posted steep declines. Only ChatGPT managed a modest gain of around $900, while Elon Musk’s Grok roughly broke even.
Even so, industry insiders argue the technology is still developing. Faizan Ahmad, co-founder of AI investment platform Rallies, says his experiments show promise in how models interpret markets. He points to instances where Claude shifted portfolios towards defence stocks during geopolitical tensions, while ChatGPT identified early growth potential in firms linked to artificial intelligence infrastructure.
Ahmad’s platform now manages around $24 million in retail capital across AI-driven portfolios, with a significant share tracking chatbot-generated strategies. He argues that AI’s ability to process filings, research reports and market data at speed gives it an analytical edge that humans struggle to match.
The institutional world is also paying attention. Hedge funds such as Man Group and Balyasny Asset Management are integrating large language models into their research and trading processes. At Man Group, AI-generated ideas still require human approval, but the systems are already accelerating idea generation and backtesting. Tasks that once took analysts days can now be completed in minutes.
Balyasny has reported widespread adoption of OpenAI tools across its investment teams, using AI to interpret earnings reports, central bank speeches and merger data. The firm says processing time for economic analysis has dropped from days to under an hour in some cases.
Despite these gains, experts caution that AI models lack access to proprietary trading systems and real-time institutional context, limiting their independence. As one executive noted, large language models remain strong at public information but are blind to internal strategies and execution constraints.
Some in the industry believe fully AI-run hedge funds may eventually emerge within a few years. Others remain sceptical, pointing to inconsistent performance and the continued need for human oversight.
For now, AI sits somewhere between assistant and advisor. Whether it becomes a reliable stock picker or just another modern version of the dart-throwing monkey remains an open question.
You must be logged in to post a comment Login