ETFs run entirely by AI: about $30 million, a few months of history, and a lot of caveats

On October 14, 2026, Reuters looked at a small corner of the fund industry: exchange traded funds where an AI model, not a portfolio manager, controls everything from research to stock selection. The count is tiny. Fewer than half a dozen ETFs do this, collectively managing about $30 million of the ETF industry's nearly $17 trillion in assets.

The broader trend is bigger and quieter. Earlier this year, 91% of asset managers surveyed by Mercer said they plan to increase their use of AI-based portfolio management tools this year, and are working out which tasks can safely be handed to AI nContext: n: not disclosed · survey earlier in 2026 · Mercer, via Reuters · margin: not disclosed.

Two models of AI in a fund

• AI-managed: the model makes the investment decisions. These are the funds Reuters counted as relying solely on AI.
• AI-enhanced: humans stay in charge, and AI runs quantitative analysis, screens stocks or monitors risk. Insiders told Reuters this cautious approach is more common.

The early numbers, as reported

Reuters compared each fund's return since launch with the S&P 500 over the same window. The windows differ by fund, so the rows are not comparable with each other.

Who is behind them

FinQ is an Israeli asset manager that launched two funds in February. The High Conviction US Equity AI-Managed ETF comes from Ai Funds, founded by Tal Schwartz; it is built on a proprietary AI learning algorithm that analyzes decades of market data, and had attracted less than $20 million in assets according to ETF.com data cited by Reuters. 'I want to offer a third option: where it's AI that is doing it all for you,' Schwartz told Reuters, adding that 'what will convince them is performance.'

Pictet Asset Management in Geneva takes the opposite stance. David Wright, its head of quantitative investments, said the firm has been training models using AI for seven or eight years and using them to run live capital for about four. Its AI-enhanced fund has attracted $163 million since February, which is several times the combined assets of the AI-only funds. Wright compares AI to a plane's autopilot: it may be able to take off, cruise and land, but 'I wouldn't go on a plane that didn't have a pilot.'

The catch

Reuters and the people it quoted were direct about the limits.

• Short records. The AI-only funds' track records stretch back only a few months. The oldest in the table launched in February 2026.
• Fees. Advisers and analysts said above-average fees may limit investor interest. Reuters did not list the fee levels.
• No bear market yet. 'Proof that this approach works will really come down to who can thoughtfully manage risk and ride out a down market,' Todd Sohn, ETF analyst at Strategas, told Reuters.
• Tail events. Dan Taylor, deputy chief investment officer of systematic at Man Group, said he still worries about 'AI's ability to deal with the most extreme, unlikely events, where all the action happens.'
• Adviser appetite. 'Advisers are comfortable with AI that informs a decision and uneasy with AI that makes it,' said Jeff Judge, a financial adviser at Chesapeake Financial Planners, adding his clients are not clamoring for these funds.

The article ran the same day US stocks had their worst session since July, so the next few weeks will be the first real stress test for several of these strategies.

How to read early AI fund returns

• Check the window. A return since February and a return since August cannot be compared, and neither says anything about a full cycle.
• Find the benchmark gap, then the fee. An edge over the S&P 500 of a few points over a few months can shrink after costs.
• Ask what the AI actually decides. 'AI-powered' covers both a model that picks every stock and a human team that uses AI for screening. The prospectus is where that is spelled out.
• Look for the down-market record. None of the AI-only funds has one yet.

None of this is a reason to own or avoid any of these funds. It is a reason to read the numbers for what they are: a few months of results from a very small group of very new products.