Free tools Windows power users keep installed
One-click scans. No signup required.
Nathan Smith reported that a portfolio selected with ChatGPT was up roughly 24–25% after four weeks—but that was an interim, self-reported result from a narrow micro-cap experiment, not proof that AI can reliably beat the market. Smith supplied information and executed the trades; ChatGPT did not operate a brokerage account on its own.
What was the $100 AI stock experiment?
In a July 29, 2025 post, Nathan Smith described an experiment that began with a $100 budget and was planned to run for six months, from June 27 through December 27, 2025. He used ChatGPT to select under-covered U.S.-listed micro-cap stocks, with a stated market-capitalization ceiling of $300 million and a goal of maximizing returns over the experiment period. Smith’s original Reddit post contains his description and interim chart.
The setup matters: this was not a general test of ChatGPT across the stock market. It focused on a small, volatile segment of U.S. equities, and the findings cannot automatically be applied to large companies, diversified portfolios, or other time periods.
What happened after four weeks?
Smith reported that the portfolio had gained roughly 24–25% after four weeks. That is the experiment owner’s reported interim result, not an independently audited performance figure. A secondary article repeated the approximate gain, but repetition does not independently verify the account or its return. Indian Defence Review’s coverage is a report of the claim, not an audit.
#1 Best Overall
The headline framing that the result left Wall Street “stunned” is not established by the available reporting. The evidence supports a more limited description: Smith posted a strong early result for his experiment.
Did ChatGPT trade the stocks by itself?
No. Smith said ChatGPT made selections using information he supplied, while he executed the trades. He also described code and data processing, daily updates, and periodic reassessment. This was an AI-assisted experiment operated by a person, not an autonomous trading service making and routing trades independently.
Rank #2
Does a 24–25% first-month gain prove AI can beat the market?
No. A short, self-reported result cannot establish a repeatable trading edge. Smith compared the experiment with the Russell 2000 and the SPDR S&P Biotech ETF (XBI), which he selected as comparisons for the portfolio’s micro-cap focus. A raw percentage return over a few weeks does not account for differences in holdings, concentration, volatility, liquidity, transaction costs, or trade execution. The result also says nothing by itself about whether the same approach would work in another period.
To assess a trading strategy more fully, a reader would need a complete, verifiable record of holdings and trades, a clearly defined comparison period, and enough observations to evaluate risk and repeatability—not just an early gain chart.
Recommended Free Tools
What was the experiment’s final six-month result?
The planned experiment window ended December 27, 2025, but a verified final result is not established by the sources cited here. The reported four-week gain should not be presented as the six-month outcome.
Quick Recap
Best Value
What should investors take from the story?
- Treat the reported 24–25% as an interim figure attributed to Smith, not a verified performance guarantee.
- Recognize that the experiment concerned U.S.-listed micro-caps, not a broad or diversified market portfolio.
- Keep the human role in view: Smith supplied information and placed the trades.
- Do not use a single short experiment as evidence that ChatGPT—or any AI tool—can reliably select winning stocks.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
