MW Most robo-advisers will never profit from Wall Street's AI-generated stock picks

By Nahyeon Bak and Daisoon Kim

Retail robo-advisors excel at tax-loss harvesting and portfolio discipline, but market-beating returns aren't part of the package

For a monthly fee smaller than a dinner out, an artificial-intelligence tool will analyze your invesment portfolio, optimize your taxes and personalize recommendations to your risk tolerance and retirement timeline - with no hidden commissions and no adviser who forgets about you.

To the 41% of Americans (according to Gallup) who historically have been priced out of quality financial advice, that sounds like a breakthrough.

But before you hand your savings to an algorithm, you should consider what kinds of choices you should trust AI to help you make. Is AI a reliable stock picker? If AI were so good at picking stocks, and able to consistently outperform a basic stock-index fund, would Wall Street sell the service to you and everyone else for $29 to $60 a month? And if an AI had cracked the code to stock picking and was willing to offer it to everyone, could it possibly continue to work?

What AI does well

AI does several things in personal finance that human advisers do poorly or expensively. A traditional adviser managing hundreds of clients cannot recalibrate each portfolio in real time as a client's income, spending habits or retirement timeline shifts. AI can and does continuously.

The cost argument is equally compelling. A 1% annual fee charged by human advisers compounds painfully over decades. On a $500,000 portfolio, that is $5,000 per year. AI-driven platforms charge a fraction of that. Vanguard's 2024 research found that automated tax-loss harvesting alone adds from 0.47% to 1.27% in after-tax returns annually.

The behavioral case is strongest of all. AI helps avoid big mistakes that humans tend to make. Dalbar's 2025 study shows that the average U.S. stock investor pocketed a 16.5% return in 2024, compared to the S&P 500's SPX 25% - an 8.5-percentage-point gap driven not by bad investments, but by bad timing choices. Investors sold during downturns and re-entered too late. Automated platforms prevent such costly decisions.

Where the wheels come off

Problems emerge the moment the use of AI moves from using information to customize the use of index funds and into market prediction and stock picking. Markets are not data puzzles to be "solved" by powerful computation. They are driven by human information processing - knowing how to weigh news as it arrives - and by fear, greed, panic, herd mentality and other emotions. AI cannot read the news to discern either the meaning of news or the emotions it will likely elicit.

The 2022 bear market is instructive. AI models trained on post-2008-recovery data were blind to the inflation shock that drove the selloff. The Aug. 5, 2024, collapse of the Nikkei JP:NIK - a 12.4% single-day decline triggered by a modest rate change and amplified by algorithmic responses - illustrates the same fragility at speed.

The IMF's 2024 Global Financial Stability Report identified a deeper structural risk: When most market participants use similar AI models and data, their strategies converge. Today's AI models may follow different trading strategies, but some experts draw a parallel to convergent evolution: Just as unrelated species independently arrive at the same biological solutions under similar pressures, competing AI models may converge toward similar behaviors over time - even without coordination. That convergence can amplify volatility under stress. The Financial Stability Board argued in November 2024 that AI is already amplifying correlated, herd-like market responses.

Cheaper isn't better

When a startup offers an AI tool that claims to beat the market, ask yourself why.

More fundamentally, if AI programs actually were able to read the news and make profitable stock picks, why would the owner of the AI sell it to you for a cheap price rather than trade on his or her own account?

The quantitative finance world is intensely competitive. Hedge funds employ the best AI researchers available, spending hundreds of millions of dollars annually to extract fractions of a percentage point of edge. Renaissance Technologies, the most successful quant fund in history, has kept its Medallion strategy closed to outside investors for decades, because a portfolio's alpha - the excess return above what the market delivers to everyone - is not scalable.

So when a startup offers an AI tool that claims to beat the market, ask yourself why. Superior algorithms are held privately, run with institutional capital and protected like trade secrets. And if a hedge fund with a reliable AI stock picker selflessly decided to give it away to everyone, it would immediately cease to be valuable, because market prices would incorporate its information too quickly to be of use to anyone. The tools available to retail investors are, almost by definition, those that have not demonstrated enough edge to be worth keeping private.

That is not an argument against using AI in your financial life. It is an argument for limiting what you are asking it to do.

AI cannot make you a better stock picker. It can make you a more disciplined, age-appropriate, tax-efficient and consistent investor. Given that the average investor has underperformed the index for 15 consecutive years, that is not a small thing. The goal is to invest in the market appropriately and stop getting in your own way - and on that score, the AI tools are already good enough.

Nahyeon Bak is a Ph.D. economist. Daisoon Kim is a scholar at the Andersen Institute for Economics & Finance.

-Nahyeon Bak -Daisoon Kim

This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

June 06, 2026 14:50 ET (18:50 GMT)

Copyright (c) 2026 Dow Jones & Company, Inc.