The short answer
Yes, computers can pick stocks. The harder, more honest question is whether AI picks them well enough, and consistently enough, to beat a simple, cheap index fund after fees. On that question the published evidence is humbling. The most-watched real-world AI stock fund has trailed the S&P 500 every calendar year since it launched, and the broader history of computer-driven "quant" funds is a mix of a few famous winners and a long tail of also-rans. If you are 50 or older and protecting a nest egg, that distinction matters more than any sales pitch.
What AI and "quant" funds actually do
Most AI-in-investing falls into two buckets. The first is academic and institutional research, where machine-learning models scan decades of data to estimate which stocks may earn higher returns. The second is consumer products: ETFs, apps, and "robo-advisors" that put an AI label on the outside. These are very different things. A model that looks brilliant in a research paper still has to survive real trading costs, taxes, surprises, and human emotion. The phrase you will hear is "past performance does not guarantee future results," and with AI that warning is doubly important, because a model trained on the past can be confidently wrong about a future that does not rhyme with it.
Where the research looks impressive
In academic terms, machine learning genuinely helps. A widely cited 2020 study in The Review of Financial Studies, "Empirical Asset Pricing via Machine Learning" by Shihao Gu, Bryan Kelly, and Dacheng Xiu, found that neural networks and decision trees roughly doubled the performance of older regression methods at forecasting stock risk premiums, and that timing the S&P 500 with their model lifted a key risk-adjusted return measure (the Sharpe ratio) from 0.51 to 0.77 in out-of-sample tests. That is a real result. But it lives in an idealized laboratory: no fund fees, no slippage, no taxes, and a researcher's freedom to test ideas after the fact. Translating that into a product you can buy is where the wheels often come off.
Where AI funds lose in the real world
Consider AIEQ, the Amplify AI Powered Equity ETF, which launched in October 2017 and was marketed as an AI stock-picker built on IBM Watson technology. According to performance data from Morningstar and analysis on Seeking Alpha, AIEQ has underperformed the S&P 500 in essentially every year since inception, while charging about 0.80% a year versus roughly 0.09% for a plain S&P 500 fund. It has also carried higher volatility and, in some years, eye-watering trading turnover. The lesson is not that AI is useless. It is that an AI label plus high fees plus heavy trading is a hard combination to win with.
The bigger picture among professional "quant" funds is similar. As Institutional Investor and Hedgeweek reported on 2024 results, most quant hedge funds fell short of the S&P 500's roughly 23% gain that year, even though several posted solid double-digit returns. A handful of elite shops, like Renaissance Technologies, did well, but they are the rare exceptions that survive precisely because most do not. One analysis noted that only a tiny number of hedge funds beat the index in 2024 at all. Longer term, the Eurekahedge AI Hedge Fund Index badly trailed the S&P 500 over the 2010s, and an academic review found AI-powered mutual funds were statistically indistinguishable from the overall market in 25 of 26 months studied between 2017 and 2019.
The yardstick that humbles everyone: SPIVA
To judge any stock-picker, AI or human, the cleanest scorecard is S&P Dow Jones Indices' SPIVA report, which compares active managers to their benchmarks. The SPIVA U.S. Year-End 2024 Scorecard found that 65% of large-cap U.S. active funds underperformed the S&P 500 over just that one year. Stretch the horizon and it gets brutal: as S&P Dow Jones Indices reported and Institutional Investor summarized, roughly 89% to 90% of large-cap funds underperformed over the 15-year period ending December 2024, and not a single U.S. equity category had a majority of active managers beat their benchmark over 15 years. AI does not get a special exemption from this math; it has to fight the same fees, costs, and competition that defeat most human managers.
Survivorship bias: the funds you never hear about
Marketing loves to show you winners. The catch is that losers quietly disappear. SPIVA data (summarized by S&P Dow Jones Indices and the Bogleheads investing community) shows that over a 15-year window, well over half of domestic equity funds, around 57% to 58%, were merged or liquidated, usually after poor performance. That is "survivorship bias": if you only measure the funds still standing, you flatter the whole group and hide the failures. The same trap applies to a brand-new AI fund with a glossy back-test. A strategy that looks unbeatable on paper may be one of dozens that were tried, with only the lucky survivors shown to you.
Why past performance truly does not predict the future
AI models learn from history. Markets, unfortunately, keep inventing new history: a pandemic, a sudden rate shock, a war, an AI bubble. A model tuned to yesterday's patterns can break exactly when conditions change, which is the worst possible time. Even when an AI strategy works for a while, success can erode it. As more money copies a winning signal, the edge shrinks. Institutional Investor noted that many tech-heavy and quant funds have drifted so close to simply mirroring the S&P 500 that their results sit within a fraction of a percent of the index, raising fees for what amounts to a costlier version of an index fund.
Robo-advisors are not the hype, and that is good
Here is a reassuring twist. The most useful "AI" in everyday investing, the robo-advisor, is the least flashy. As the SEC explains on Investor.gov, a robo-adviser is an automated program that asks about your goals and risk tolerance, then builds and rebalances a portfolio for you. In practice, as Vanguard describes its own service, these tools mostly buy low-cost index ETFs that aim to match the market, not beat it, for a small fee (Vanguard's Digital Advisor runs around 0.15% a year). That is automation doing chores, like rebalancing and tax management, rather than promising magic returns. It is a sensible use of technology precisely because it is not pretending to outsmart the market.
Spotting AI investment hype, and outright fraud
Regulators are watching the hype closely. In January 2024 the SEC's Office of Investor Education, together with FINRA and NASAA, issued a joint Investor Alert warning that scammers are using AI buzzwords to lure victims, and that promises of guaranteed, high, low-risk returns are classic red flags. Weeks later, per SEC Press Release 2024-36, the SEC charged two advisers, Delphia and Global Predictions, with "AI washing", making false or misleading claims about their AI capabilities; they paid civil penalties of $225,000 and $175,000. The takeaway is plain: if a pitch leans on "AI" and "guaranteed," treat it as a warning sign, not a feature.
What this means for your money
None of this means avoiding technology or markets. It means matching your expectations to the evidence. If you want to understand how much risk a stock-heavy or AI-themed strategy could add to your portfolio, you can think it through first with our <a href="/calculators/investment-risk">investment risk calculator</a>, which helps you see how different mixes of stocks and bonds might swing in value. For most people approaching or in retirement, the unglamorous combination of low costs, broad diversification, and patience has beaten the vast majority of clever stock-pickers, AI included.
This article is educational and not personalized financial advice. All investing carries risk and past performance does not guarantee future results. Consider consulting a fiduciary financial advisor about your situation.