Can AI beat the stock market?

Last updated August 2026

Short answer

There is no evidence that any AI product beats the stock market over a long period, and the honest reason is that almost nothing does. S&P Dow Jones Indices measured 78.78% of active large-cap US equity funds underperforming the S&P 500 in 2025, and 92.89% over 20 years. Those are full-time professionals with expensive data. AI changes what an individual investor can read, check and understand about their own holdings. It does not change the arithmetic of who wins against an index.

The question usually arrives with an assumption attached: that beating the market is a solvable problem, and better technology should solve it. The published record says something less convenient.

What the professional record looks like

S&P Dow Jones Indices publishes the SPIVA Scorecard, which compares active funds against the index they are measured on. The Year-End 2025 US edition found 78.78% of active large-cap funds trailed the S&P 500 over one year.

The longer horizons are worse for the managers: 88.96% over five years, 85.59% over ten, 89.93% over fifteen and 92.89% over twenty. Across all domestic funds against the S&P Composite 1500, the twenty-year figure reaches 95.01%.

Those funds employ analysts, quantitative researchers and data nobody sells to retail. The failure is not a shortage of intelligence applied to the problem.

Why the arithmetic is hostile

An index is close to the average outcome of everyone trading, measured before costs. For one investor to beat it, another has to lose to it by the same amount, and both pay fees to play.

A prediction only pays if it is both correct and not already reflected in the price. Public information tends to be in the price quickly, and an AI product reading public filings is reading what everyone else can read.

None of that makes outperformance impossible. It makes it rare, hard to distinguish from luck over short periods, and unlikely to persist once a strategy is widely known.

Where the claims come from

The regulators have been direct about this. A CFTC customer advisory published on 25 January 2024, titled "AI Won't Turn Trading Bots into Money Machines", states plainly that AI technology cannot predict the future or sudden market changes, and that claims of high or guaranteed returns are red flags of fraud.

The SEC has gone further and charged firms for the claim itself. On 18 March 2024 it settled with two investment advisers, Delphia and Global Predictions, for false and misleading statements about their use of AI, with $400,000 in combined civil penalties.

Backtested numbers deserve particular scepticism. A strategy tuned on history will look excellent on history.

Try it in Walnut

Walnut does not claim to beat the market. It connects to your brokerage and answers questions about what you already hold, which is a smaller promise and a checkable one.

Create a free Walnut account

What AI does change

The gap between an individual investor and an institution has never mainly been predictive skill. It has been time, attention and the ability to read a lot of material carefully.

Reading a 10-K and pulling out what changed, explaining why a holding moved, showing that four of your positions are the same bet on one supply chain, checking whether your allocation still matches what you intended: those are jobs where a model that reads well is genuinely useful.

They also happen to be the jobs where being wrong is recoverable, because you can check the answer against the source.

How to read a performance claim

Ask what it is measured against, over what period, and whether the record is live or backtested. A strategy with three good years and no benchmark is not evidence.

Ask whether costs are included. Gross-of-fee outperformance that disappears after fees is the normal outcome, not an unusual one.

Ask who is registered and where the filings are. Registered advisers file a Form ADV, and it can be read for free.

Sources

Fund underperformance figures are from the SPIVA U.S. Scorecard, Year-End 2025, Report 1a, published by S&P Dow Jones Indices. Regulatory warnings are from the CFTC customer advisory of 25 January 2024 and SEC press release 2024-36. Walnut is informational and is not an investment adviser. This guide is educational and not personalized investment advice.

FAQ

Has any AI fund actually beaten the market?

Individual AI-branded funds have beaten their benchmark in individual years, as active funds always do. What nobody has shown is a durable edge over a full cycle. S&P's persistence work finds that top-quartile performance in one period is a poor predictor of the next, and that applies to a quant strategy as much as a human one.

Why do so few active managers beat the index?

Fees, and the fact that the index is the average of everyone trading, before costs. S&P Dow Jones Indices measured 78.78% of active large-cap US equity funds underperforming the S&P 500 in 2025, rising to 92.89% over 20 years. These are well-resourced professionals with the same data an AI product has.

So what is AI actually good for in investing?

Reading, summarising and checking. It can go through a filing, explain a holding, show what your portfolio is concentrated in, and answer questions about what you own without you building a spreadsheet. None of that requires predicting prices, which is the part that does not work.

Should I be suspicious of an app that claims market-beating AI returns?

Yes. The CFTC warns that claims of high or guaranteed returns are red flags of fraud, and the SEC has charged advisers for overstating their AI capabilities. A performance claim you cannot verify in an audited filing is marketing until proven otherwise.