Experts urge caution about using ChatGPT to pick stocks (arstechnica.com)

AI chatbots are increasingly being used as DIY financial advisers: a Reuters summary of an eToro survey of 11,000 retail investors finds about 13% already use tools like ChatGPT or Gemini for stock selection and roughly half would consider doing so. Users typically query models for analysis and then manually place trades, rather than using automated algorithmic execution. Anecdotes include a former UBS analyst who relies on ChatGPT workflows in lieu of expensive data terminals, and a Finder experiment where a 38‑stock ChatGPT-picked portfolio reportedly rose ~55% since March 2023—outperforming popular UK funds—but that performance sits against a strong market backdrop that can inflate results. For the AI/ML community this trend highlights both opportunity and risk. Large language models can democratize access to analytic reasoning and screening workflows, but they also have well‑known limitations: they can hallucinate facts, lack access to real‑time market data, and aren’t optimized for financial backtesting or live execution. That means outputs may be brittle, poorly calibrated for tail risk, and prone to data-cutoff and survivorship biases. Practical implications include the need for specialized, updatable financial models, transparent uncertainty estimates, rigorous evaluation on forward‑looking returns, and user safeguards—especially as non‑expert investors increasingly treat conversational AIs as substitutes for licensed advice.