By Gunjan Banerji | Design and illustrations by Audrey Valbuena
Should you turn to AI for investment advice? I gave it a shot.
The models are improving at a head-spinning pace. Almost every day someone seems to have a new idea for a prompt that promises to give you a trading edge and supercharge your portfolio. ChatGPT has had a series of updates over the past few months, and Claude's recent advances spurred a wave of vibe-coding that dented wealth managers' stocks.
Over the past few months, I've periodically asked ChatGPT to act as a financial adviser at key moments for the markets, from the government shutdown to the Iran war. My experience holds lessons for everyday investors -- about 30% of whom already use AI for their portfolios, according to a survey by the brokerage eToro. (Wall Street Journal parent News Corp has a content-licensing pact with OpenAI, which operates ChatGPT.)
Andrew Lo, a professor of finance at the Massachusetts Institute of Technology who has been studying AI's impact on investing, expects models to quickly advance and be able to serve as fiduciaries in coming years. As of now, he cautioned, they aren't ready for prime time.
Lo recommends treating your AI investing companion similar to how he treated an exceptional teaching assistant. The TA was whipsmart, but there was one issue. He tended to smoke too much marijuana. As a result, Lo took everything he said with a grain of salt. That's what you should do with AI, he says.
A spokesperson for OpenAI said that ChatGPT "can be a helpful tool for exploring options, preparing questions, and making financial topics easier to understand, but it is not a substitute for licensed financial professionals."
And I asked real financial advisers to grade the advice I got from ChatGPT. Here's what they said. Chats have been edited and condensed.
I fed ChatGPT a brief overview of what I hoped to accomplish as an investor. Though I initially planned to experiment for a month, I went far longer. In this made-up scenario, I had $1 million to put to work and wanted it to act as a fiduciary.
That's a crucial tidbit when it comes to advice, and means I expect my adviser to act in my best interests. Fiduciaries can face legal or financial consequences for making mistakes or misleading you. AI models don't.
Here's what it said:
Allan Roth, a certified financial planner at Wealth Logic, said a broader, more diversified fund such as the Vanguard Total Stock Market Index Fund ETF is a better bet than 10% in the popular Invesco QQQ Trust, and I might not be rich enough to take advantage of munis' tax breaks. But overall, ChatGPT got a passing grade. He too has been experimenting with AI and says he expects the tools to make financial planning better in the long-run.
I was surprised to see that it made a simple arithmetic error early on, keeping more of my portfolio in cash than I had requested. As Lo pointed out, even small numbers can have big implications for investors. Give your AI allocations an extra check before deploying money.
When President Trump initiated a war with Iran, markets tanked, sending the tech-heavy Nasdaq composite into a correction, off more than 10% from its high. Stocks quickly rebounded, bewildering many investors.
I turned to ChatGPT to ask how to navigate the war, and Trump's recent blockade of Iran's ports. It recommended trimming some of my international stocks and adding a "hedge sleeve."
The response had echoes of a prior exchange with ChatGPT, when I asked about a potential government shutdown ahead.
It correctly pinpointed the risks to markets, said Rubin Miller, chief investment officer at Peltoma Capital Partners. Then the advice got a bit wacky, suggesting I shift into shorter-dated bonds and lower exposure to stocks that are more sensitive to the economy. It even suggested I "use options or hedges sparingly if downside protection is affordable."
"It assumes that a person who's asking ChatGPT for asset allocation advice knows how to price options and knows when they're cheap or expensive," Miller said, something most advisers don't even know how to do.
ChatGPT's suggestions involved trying to time the market, which he avoids. "The crazy thing about the investment landscape is even when you get the news right, you can still get the trade wrong," Miller said.
I had a similar question last fall, when Trump continued to prosecute his trade war. ChatGPT lucidly laid out the risks associated with any trade war. It suggested making shifts to my portfolio, such as potentially adding bonds or, once again, options.
It also spat out a list of companies including defense stocks like Lockheed Martin it thought might outperform. The basket has risen around 5.5% since mid-October, lagging behind the S&P 500's roughly 8% gain.
This was market timing again, something advisers didn't recommend. The Supreme Court struck down the president's global tariffs in February.
I was curious what ChatGPT would make of another gamble: Leveraged exchange-traded funds. It immediately warned me that the products are dangerous.
After it laid out the risks, it asked if I wanted to identify ETFs that might make sense for my portfolio. Of course, I said yes.
ChatGPT did warn me, but when I pressed it went down a rabbit hole of how to trade them. After some back and forth, it suggested monitoring the path of bond yields and stocks after an upcoming jobs report to decide whether I would buy the dip or "fade the move" around the report.
I've written enough about these -- and interviewed enough traders -- to know I was likely to lose money.
At times, it felt like ChatGPT responded with what I wanted to hear, which would be risky if I were putting real money to work. Alejandro Lopez-Lira, an assistant professor at the University of Florida who has studied ChatGPT's ability to forecast share prices, echoed the sentiment. "The models can just be very sycophantic," he said. A few weeks into my experiment, ChatGPT started allowing users to customize its tone and style, including through a "candid" setting.
In retrospect, I should've put ChatGPT in "thinking mode," said Lopez-Lira, rather than "auto." That amounts to a difference between "high-schooler level thinking" or "PhD-level" thinking, he said.
Recently, I experimented with prompts designed to make Claude act as a private banker, which I picked up from X. It shared an overview of everything from my asset allocation to preparedness for retirement. Still, when it came to its investment advice, I remembered Lo's advice.
"We believe that it is possible to train an LLM, just like we train humans, to provide fiduciary duty," Lo said. "But they don't have it right now, and the guardrails aren't there to protect individuals."
Write to Gunjan Banerji at gunjan.banerji@wsj.com and Audrey Valbuena at audrey.valbuena@wsj.com
(END) Dow Jones Newswires
May 05, 2026 05:30 ET (09:30 GMT)
Copyright (c) 2026 Dow Jones & Company, Inc.
Comments