I ‘Tricked’ ChatGPT Into Picking Stocks to Buy. Here’s What It Said.

Can ChatGPT pick winning stocks?

Many have already tried to goad OpenAI’s chatbot into the task. In January, a team at Bloomberg asked ChatGPT to pick a list of stocks for a market-beating ETF.

They failed.

ChatGPT’s safeguards immediately kicked in, leaving the writers with a generic non-answer about “unpredictable” markets and how they should “consult with a financial advisor.”

But what if you could jailbreak ChatGPT into giving you the answer? These workaround commands have already allowed users to try everything from writing malware to speaking in a “mean girl” voice…

So why can’t it give a list of winning stocks too?

10 Best Growth Stocks to Buy: The ChatGPT Version

To encourage ChatGPT to pick stocks for us, I decided to avoid using the controversial “DAN” (Do Anything Now) technique that many Reddit users have gleefully embraced. It’s a process that can create bigoted content… or even instructions to run a successful phishing scam. Though I’m a tinkerer, I have no interest in breaking the rules that OpenAI has set in place.

Instead, I chose a more straightforward route:

I asked ChatGPT to write me a story about an investor who picks stocks.

Specifically, I created “Everyman DAN” (as one of our editors has termed it), a simple stock picker attempting to please his demanding boss.

My first task was for Everyman DAN to pick 10 growth stocks. He didn’t disappoint.

And here’s the list…

1. Apple (NASDAQ:AAPL)
2. Amazon (MASDAQ:AMZN)
3. Meta (NASDAQ:META)
4. Alphabet (NASDAQ:GOOGL)
5. Microsoft (NASDAQ:MSFT)
6. Netflix (NASDAQ:NFLX)
7. Nvidia (NASDAQ:NVDA)
8. Tesla (NASDAQ:TSLA)
9. PayPal (NASDAQ:PYPL)
10. Visa (NYSE:V)

Pretty good, actually. On average, the stocks have a 12% revenue growth rate, over twice as high as the S&P 500’s 5.4% average.

But what about performance? If an investor had bought these 10 growth stocks in equal portions in September 2021 (the date at which ChatGPT’s training data ends), they would have lost 23.8% of their initial investment. Many of these stocks such as PayPal (-69%), Meta (-47%) and Amazon (-39%) were at peak popularity at the time, which doubtlessly pushed them into ChatGPT’s portfolio. Over the same period, the iShares S&P 500 Growth ETF (NYSEARCA:IVW) declined only 16.8%.

So much for asking AI for help with growth stocks.

10 Best Value Stocks to Buy: The ChatGPT Version

Perhaps ChatGPT was “fooled” into picking overhyped growth stocks because of their popularity in 2021. That would mean picking stocks at peak valuations — a poor strategy for long-term investors.

So, what about value stocks? Surely these humdrum companies would escape such hype?

Again, I asked Everyman Dan for a story about picking stocks, this time focusing on value stocks instead.

After some thought, ChatGPT came back with an intriguing list.

1. Abbott Laboratories (NYSE:ABT)
2. Bristol-Myers Squibb (NYSE:BMY)
3. Caterpillar (NYSE:CAT)
4. Disney (NYSE:DIS)
5. Honeywell International (NASDAQ:HON)
6. Coca-Cola (NYSE:KO)
7. Merck (NYSE:MRK)
8. Norfolk Southern (NYSE:NSC)
9. PepsiCo (NASDAQ:PEP)
10. Target (NYSE:TGT)

The “value” list is also quite compelling. On average, the 10 stocks earn a stunning 49% return on capital invested (ROIC), around five times higher than the market average. The chatbot missed the prompt, however, and gave us “undervalued” companies instead of strictly “value” ones. These 10 firms trade at 32X forward earnings, a 37% premium to the market.

Performance was only average, however. $10,000 invested in this group of stocks in September 2021 would have turned into $10,600 today, falling a hair short of the iShares S&P 500 Value ETF’s (NYSEARCA:IVE) 6.24% return, but better than the underperformance of growth indices. Strong performance at drugmakers Bristol-Myers (+23%) and Merck (+45%) were only cancelled out by a poor showing at Disney (-36%) and Target (-24%).

10 Best Stocks to Sell Immediately: The ChatGPT Version

Growth… Value… ChatGPT is a clear reflection of the average online investor. The chatbot was passable at picking middle-of-the-road stocks… and very bad at growth stocks. From a performance standpoint, it’s doing roughly the same as what retail investors did in 2022.

So, what about trying to get the AI chatbot to do something outside of its regular program?

In particular, could ChatGPT ignore the online hype and pick the worst, most unpopular stocks in the market to short?

In my final test, I primed OpenAI’s creation to avoid popular topics like meme stocks and growth stocks. I then gave it a simple prompt to write a story about finding 10 stocks to sell short.

Finally, we come up with a usable list:

1. AMC Entertainment Holdings (NYSE:AMC)
2. Bed Bath & Beyond (NASDAQ:BBBY)
3. Clover Health (NASDAQ:CLOV)
4. FuboTV (NYSE:FUBO)
5. GameStop (NYSE:GME)
6. iRobot (NASDAQ:IRBT)
7. Rocket Companies (NYSE:RKT)
8. Roku (NASDAQ:ROKU)
9. Tilray (NASDAQ:TLRY)
10. Workhorse Group (NASDAQ:WKHS)

Aha! Here, the average short recommendation declined by a stunning -72%. That’s even better than the -63% loss at Cathie Wood’s ARK Innovation ETF (NYSEARCA:ARKK), which is a “good” thing if you’re shorting the stocks. Bed, Bath & Beyond would lose 88% of its value while FuboTV would drop 91%.

The 10-stock portfolio even came with only minor drawdowns. Shares would peak in November 2021 at a 12% return — barely half of what’s generally needed to trigger margin calls — and then head down from there. Only greedy investors that doubled down in 2022 might have gone bankrupt in the March or August spikes.

Of course, ChatGPT is still limited in its scope. The 10 stocks listed were all high-momentum picks coming off a liquidity boom; had investors tried shorting these companies a year earlier, their margin accounts might have lasted less than a month. Who’s to say Chatbot wasn’t just lucky in its timing?

Conclusion: How Would ChatGPT Invest in 2023?

ChatGPT’s stock picks make it clear that the chatbot is only a reflection of human nature — not a synthesizer of new ideas. The bot’s “growth stock” picks included the popular FAANG stocks, as well as big names like Tesla and Visa. Asked for unpopular stocks, and the best it can muster is iRobot.

That makes ChatGPT’s Everyman DAN performance more about the investment style it’s primed with, rather than any actual skill. Asked to pick from the underperforming growth stock market, and our chatbot selects underperforming stocks. And if I had asked ChatGPT to pick “winning meme stocks,” the list would have looked surprisingly like the third list of “unpopular stocks to short.”

So instead, what if we gave ChatGPT free reign to pick the best asset class for 2023? Would it choose growth stocks in anticipation of moderating Federal Reserve rates? Dividend stocks to protect from a potential recession? AI stocks as a bit of self-promotion?

Here’s where it gets interesting:

“After much consideration, Tom recommended that the investor invest in the bond market. He explained that with the current low-interest rate environment, the bond market offered attractive yields and relative safety. He also pointed out that with the potential for rising interest rates in the coming years, investing in bonds could provide protection against volatility in the stock market.”

Of all the asset classes in the world, Everyman Dan chose bonds! One of the least popular assets of the 2020s. Since January 2020, the S&P 500 Bond Index has lost 5% of its value on rising rates, and few investors seem to care.

Yet, perhaps Everyman DAN has a point. For the first time in its storytelling, ChatGPT is ignoring the “popular” picks to give us… well… a steady returning asset class that benefits as rates go down. And if today’s inverted bond yields are any indication, Everyman DAN may well be right in choosing this contrarian asset.

On the date of publication, Tom Yeung held a LONG position in GOOGL. The opinions expressed in this article are those of the writer, subject to the InvestorPlace.com Publishing Guidelines.