Back in June 2025 – a Reddit user by the name of Nathan Smith posted an experiment.
He asked ChatGPT to pick U.S. microcap stocks (stocks with a market cap below $300 million) for him.
His goal?
“See how a language model performs in picking small, under-covered stocks with a $100 budget.”
Over the first four weeks of his experiment – the account reportedly generated 24-25% return on the $100.
A 24-25% return…
Sounds like a success, right?
Not so fast.
Soon after posting those results… Smith wrote: “I no longer can get chat to provide insights. I’m not sure what has changed but it is making up data and agent mode won’t work anymore.”
Oops.
It’s a good reminder that leaving investing decisions solely up to AI is a BAD idea… for a number of reasons…
Problem No. 1. It’s not hands-free
While ChatGPT generated the stock picks, the user still had to monitor the trades every step of the way.
He had to adjust if ChatGPT contradicted itself or recommended trades that were impossible.
He also had to impose his own stop losses and manually place the orders.
So although he was technically using ChatGPT to trade, it was more of an assistant than a “done-for-you” trading service.
Problem No. 2. He had no access to realtime data
ChatGPT can’t produce live price charts, track volume, see liquidity, or react fast enough to catalyst news.
It’s a language model… meaning it’s picking stocks based on text patterns from user data, not financial reality.
This sets it up for all sorts of false or outdated information that could lead to disastrous trades that crater your account.
Problem No. 3. It encourages the worst mindset
The idea that “AI will trade for me” is not a strategy.
Proper trading involves rules, discipline, pattern recognition, position sizing, risk control – and above all – patience.
Those are skills you can’t outsource to a language model.
You can only acquire them through getting in the trenches and learning how to trade using a proven system.
Problem No. 4. No long-term track record
Anyone can look smart with a small handful of winning trades over a short time period… in a rip-roaring bull market like we’ve had since April.
But could this ChatGPT trading model hold up during earnings season? A Fed meeting? A Black Swan event like COVID or the Housing Crisis?
YOUR ACTION PLAN
While this redditor experiment is a fun example of how a robot could trade for you, NO ONE should see it as a viable long-term trading strategy.
While AI agents like ChatGPT can help with stock research…
I’ll always favor using realtime data, risk management, and proper timing for long-term trading success.
But when you combine the research power of an AI agent with the expertise of a veteran trader (like yours truly)… some real magic can happen.