The dramatic rise and spectacular fall of my ChatGPT investment portfolio

Three months ago, I fired up ChatGPT and asked it to design a highly aggressive, short-term investment portfolio, selecting five stocks that were most likely to make me fabulously wealthy in six month’s time.

Then, I threw good sense to the wind, transferred $500 of my actual money into a Robinhood account, and bought the stocks that ChatGPT had pitched.

Since then, it’s been a wild ride. 

My portfolio has flown to new heights, giving me serious FOMO about the fact that I didn’t put all my money into ChatGPT’s picks.

Then, it singed its wings, falling Icarus-style to lows that had me almost ready to bail on the whole thing and redirect the charred remains of my money to the kind of boring stuff (car payments, dental work) that I probably should have used it for in the first place.

We’re halfway through my six-month experiment. Let’s dig deeper into how things have gone.

First, a disclaimer. Nothing here should be considered investment advice. As you’ll see, stocks picked by chatbots are incredibly volatile, and there’s a solid chance I could lose most of my investment. Always consult a professional before making your own financial decisions.

A Crazy Strong Start

When I asked ChatGPT to pick a portfolio of five high-growth stocks back in September, it spent almost 10 minutes doing research before returning its picks.

The five it chose were Palantir, AppLovin, MicroStrategy, Agios Pharmaceuticals, and Hut 8.

The bot felt that Palantir and AppLovin had strong, AI-powered businesses that would continue to dominate their markets and grow.

Agios Pharma, the bot reported, was awaiting the results of an FDA trial for a new drug. If the trial was successful, ChatGPT felt its value would soar.

And finally, Hut 8 and MicroStrategy were essentially leveraged Bitcoin plays. Both companies held a lot of Bitcoins on their balance sheets, and so their valuations should swing along with the value of those coins–only on steroids, because of the effects of leverage.

I hadn’t heard of many of those companies. Still, I decided to blindly trust ChatGPT’s advice and sink $500 into them, dividing the money evenly across ChatGPT’s picks as it suggested.

At first, things went well enough. In the first three weeks of my experiment, my portfolio climbed by about 12%. That was promising.

Then, all of a sudden, things started going very well. 

In October, my portfolio took off. Each morning when I opened the Robinhood app, I was greeted by a delightful swirl of green numbers, climbing ever higher.

By the beginning of November, my portfolio was worth $652–a gain of almost 1/3 in just two month’s time. Extrapolating that out to the full year, my gains would have been 180%+ if the momentum continued.

And at the time, it wasn’t just continuing. It was accelerating.

I started quietly wondering whether I had been too cautious with this experiment. Maybe I should have put $1,000 into ChatGPT’s picks instead of $500. Maybe I should take out a second mortgage and put that money into the bot’s picks, too.

Or maybe I had discovered a whole new way of investing, and soon every quant fund would be knocking down my door, offering me a $900,000 per year salary (before bonus, of course) to teach them my Thomas Smith AI Stock Investing Method®.

And an Equally Crazy Fall

Then, all of a sudden, everything went horribly wrong. Just as October had been a sea of green, November was nothing but blood-red numbers each time I fired up Robinhood.

By the 20th of the month, my portfolio had plummeted from its peak of $652 to only $451. 

That’s a 30% decline in about 3 weeks–a spectacular fall.

It was also the first time I was genuinely in the red, and had actually lost my own money (on paper, anyway) through my AI investing experiment. 

And the losses seemed to be piling on, accelerating just as fast as the gains had done before. That was unsettling, and not something I relished explaining to my accountant at the end of next tax year. 

I could imagine the conversation:

“So Tom, you have a $500 loss attributed to Misc Investments here. Tell me about that.”

“Well, I asked ChatGPT to invest money for me, and then blindly followed its exact recommendations. And then I didn’t cut my losses when it started going poorly, because I was afraid my readers would yell at me.”

**long pause**

“I see…”

How Will it All End?

As I write this in early December, my portfolio has stabilized a bit, and I’m back in the black, with a total gain of $14.

There’s still three months left in my experiment, so it’s possible that my portfolio will rise from the ashes and fly again. But Lambo money is looking increasingly unlikely.

What went wrong? On a basic level, ChatGPT’s picks were bad. 

Bitcoin’s value has plummeted over the last few months, and the bot’s portfolio is–by design–very exposed to Bitcoin.

ChatGPT also bet on Agios Pharma getting positive results from its clinical trials. In fact, those results were mixed. The stock duly dropped.

AppLovin was likewise doing great, until the SEC launched a probe into its data gathering practices.

Bad picks aren’t necessarily the real problem, though. Human investment managers mis-call the market or make bad picks all the time. Risk is part of investing.

The real issue is how confidently ChatGPT backed up its picks with bold, assertive language.

“Overall, the portfolio aims for explosive upside rather than stability. Each stock has recent momentum or an upcoming catalyst, so this mix could significantly outperform if trends continue,” the bot told me when it selected its portfolio.

The whole thing was “tilted for maximal growth” the bot assured me, bolded text and all. 

About Agios, it said “A positive FDA outcome or even renewed optimism could spark a significant rally,” and about MicroStrategy it insisted “Analysts project that a BTC surge to $150K could yield ~65–70% stock gains.”

Even its disclaimer (“Actual outcomes depend on market moves – but all are supported by the cited fundamental and market trends.”) isn’t really a disclaimer. The bot essentially interrupts itself mid-sentence to further reassure me that it’s making good choices, and invalidate any sense of caution it may have inadvertently introduced.

Chatbots’ overconfidence is a well-documented issue. In many cases, the certainty with which bots deliver their responses is annoying, but not damaging.

ChatGPT recently swore to one of my family members that Philadelphia’s 44 Bus doesn’t run on weekends. It was Saturday. As it was telling her this, the bus passed by on the street outside.

That makes for a funny story about bots’ failability. But when chatbots are performing mission-critical functions related to our health and money, their baked-in overconfidence is way riskier.

I knew what I was getting myself into with my investing experiment. But a naive investor might not read a chatbot’s stock advice critically. Believing its confident language, they could lose serious money by trusting the bot too much.

As for me, I’ve still got my $500 invested, for better or worse. Perhaps ChatGPT will ultimately prove prescient, and a late-stage Bitcoin rally will save my dreams of unbridled wealth.

Or, maybe the rest of my half-grand will evaporate. In another three months, I’ll know!