Are AI Investors the Next Noise Traders?

• AI could create a new generation of noise traders. Retail investors relying on similar models may make correlated decisions, amplifying market mispricing.
• Herding could create opportunities for active managers. Shared AI assumptions may produce price overshoots, particularly in retail-dominated markets.
• Independent judgment remains the professional advantage. Portfolio managers who challenge AI-generated consensus may identify opportunities others overlook.

Alpha is not a fixed pool. Its size depends on two things: how much money is hunting for mispricing, and who is making the mistakes that create it.

Each era of asset management has changed both. Index funds pulled money out of the hunt and added a large pool of money that buys regardless of value. Artificial intelligence has begun to add another: individual investors who let a chatbot do their analysis.

These investors will not be careless. They will apply sound methods in the same way without knowing when those methods stop working. When they are wrong, they will be wrong together; when they are right, they will crowd into the same trade. Markets have always had noise traders: investors who, as Fischer Black put it, trade "on noise as if it were information." Herding crowds are not new either. What is new is the source of the crowding: the same tool, rather than a rumor or a mood. That is why careful stock picking could pay again.

Index funds are now the largest pool of US fund money: 55% of US fund assets at the end of 2025, up from 34% in 2016. Passive strategies of all kinds own roughly a third of the US stock market (PWL Capital, using Morningstar data).

Money that buys regardless of price moves prices more than textbooks assume. Gabaix and Koijen estimate that each dollar flowing into the US stock market, from any source, raises its total value by about $5. The clearest case is the index effect: Stocks added to the S&P 500 earned abnormal returns averaging 7.4% in the 1990s (Greenwood and Sammon).

That gain has since faded to less than 1% over the past decade, even as index-linked assets grew. One reading is that once a price-insensitive flow becomes predictable, others compete the gain away.

AI-guided money adds a different kind of flow. It does weigh value, but many investors use the same tool. Unlike index funds, AI does not pull capital out of active investing. It reduces the independent judgment behind that capital: when many investors hand the same analytical steps to similar models, more active money may compete with the same assumptions and reach the same conclusions.

Every analytical method works within limits. The expert's edge is knowing where those limits are. AI hands the novice the method without the limits.

This is no longer a thought experiment. In a March 2026 Investing.com survey of 938 US investors, 51% said they use AI tools regularly or occasionally, and generating trading ideas was among the uses they cited (34%).

The gap between method and limits appears at every horizon:

• Discounted cash flow (DCF), months to years. The arithmetic is easy. The judgment sits in the growth, margin, capital spending and discount-rate assumptions, and in knowing which of them do not fit this business.
• Multiples, weeks to months. A chatbot will readily compare four companies on price-to-earnings. The expert knows why they deserve different multiples: different cycles, balance sheets, accounting or growth quality.
• News and central bank statements, minutes. A model reads the words of a press release or a Fed statement. The expert reads what changed, what was left out and what the market already expected.

In each case the novice's error is not bad arithmetic. It is a valid method applied where it does not belong.

These examples come from practice, not from a study. The research so far shows that AI-guided decisions move together, not yet why they go wrong.

Those choices also look alike, because the tools look alike. In a 2026 SSRN paper, Steven Edwards had GPT-4o play 100 investor personas, from value to momentum, and asked each one separately to pick 50 US stocks. He then took the 50 stocks the personas named most often and called them the consensus portfolio. Just 25 personas, drawn at random, recovered about 88% of it. The consensus portfolio also had 92% of its stocks in common with the picks from a single neutral prompt, with no persona at all. Edwards concludes that models trained on the same data make correlated picks, whatever the persona.

Similar is not the same as wrong. That consensus portfolio returned 29.5% a year against 15.3% for an S&P 500 fund over the test period, which the author attributes to its tilt toward large technology stocks. The opportunity comes by two routes. When the shared view is wrong, everyone makes the same error at once. When it is right, everyone acts on it at once, and the crowd may carry the price past fair value.

Independent mistakes point in different directions and partly cancel out. Shared mistakes add up and move prices.

Researchers have started to document this risk. A 2026 working paper by Xiaoxue He of Lancaster University, not yet peer reviewed, uses 14 unexpected ChatGPT outages during US trading hours as a natural experiment. When ChatGPT was available, retail trading was more correlated, within stocks and across industry peers, and trades moved prices more: price impact was about 12.5% higher.

The two sides of the trade differed. Retail selling pushed prices too far, then reversed over the next one to five trading days. During outages, that reversal largely disappeared. The author's reading: A shared low-cost signal may crowd out investors' own information. Retail buying, by contrast, became more informative with ChatGPT.

The evidence so far covers short horizons. For longer horizons, where the errors would come from valuations and multiples, there is no direct study yet: that part is an expectation, not a finding.

The opportunity will be uneven. It should be largest where AI-guided money is big relative to professional money.

• Markets and stocks with heavy individual trading. These flows matter most where individuals set the marginal price, not where pension and index funds hold most of the float.
• Consensus favorites. Names that every model recommends draw correlated buying. They are not automatic sells: In the persona study, the AI consensus did well. They are where to check for crowding and for the assumption everyone shares.
• Turning points. Shared assumptions are usually extrapolations of the recent past. When a cycle turns, the growth and margin defaults break together, and the selling overshoots together.

The footprint to look for is the overshoot itself: a move that runs further than the news justifies, on heavy individual flow, then reverses within days. It shows only in hindsight, so it marks where to look, not when to trade. At longer horizons, look for valuations that lean on the same borrowed comparisons and assumptions.

Saudi Arabia already looks like the market this thesis describes. Non-institutional investors, as Tadawul classifies them, held only 13% of the main market's free float at the end of 2025 (Argaam, Tadawul data), yet they placed 55% of all purchases in July 2025 and 42% in December (July, December).

Saudi individuals herded long before chatbots. A 2015 study found pervasive herding in the Saudi market that shared fundamentals were unlikely to explain, consistent with individual investors acting as noise traders (Rahman, Chowdhury and Sadique).

Counting funds and counting money answer different questions. Over ten years, 62% of Saudi-focused active funds lagged the index, yet weighted by assets, they beat it by a wide margin over three, five and ten years (Figure 1). The sample is small, about 20 funds, and the index's concentration and Shariah screens may explain part of the record.

Figure 1. Over 10 years, 62% of Saudi equity funds lagged, yet the average riyal beat the index

Even so, Saudi Arabia is a natural test. Most of this record predates consumer AI tools. If the thesis is right, the gap between professional money and the index should widen as individuals adopt them. If AI makes individuals better investors, it should narrow.

The thesis rests on AI making investors' decisions more correlated. That may not hold.

The first risk is that AI makes individuals better investors, not just more alike. In lab experiments, Federal Reserve economists found AI agents herd less than humans and lean more on private information (Hansen and Lee). In the Lancaster study, ChatGPT already made retail buying more informative. If that effect dominates, AI removes noise instead of correlating it.

Use is also not delegation. Among the uses survey respondents cited, research led (62%) and portfolio decisions trailed (22%). And the tools will improve: if models learn when their own methods do not apply, the gap between novice and expert narrows.

Finally, the opportunity is contested and costly to hold. Correlated flows can push prices further before they reverse. Market makers already model individual order flow and would compete for the same overshoots.

The test is simple. If AI adoption keeps rising while individual trades become less correlated and their reversals fade, the thesis is wrong.

The edge shifts from processing information to judging it, and to holding a view when everyone's tools agree.

1. Use AI for the work, not the call. Let it gather data, build the model and summarize filings. Keep the assumptions, the judgment about when a method does not apply, and the decision to buy or sell, human. A professional who lets the same tool make the call simply joins the crowd.
2. Ask what the models would say. Run the popular tools on your names. Where their answer is unanimous and the price already reflects it, find the assumption they all share and test it.
3. Match the capital to the horizon. Short overshoots reward fast, disciplined trading. Mispriced valuations reward patient money: longer lock-ups, clients who understand the process, and position sizes that survive being early.

Index funds created money that buys regardless of value. AI is creating buyers and sellers who weigh price against value through one shared lens. In both cases, the reward goes to the investor who still thinks independently.

• SPIVA MENA Scorecard, Year-End 2025, S&P Dow Jones Indices
• Noise, Fischer Black, Journal of Finance, 1986
• The Passive vs. Active Fund Monitor, Year-End 2025, PWL Capital (Morningstar data)
• In Search of the Origins of Financial Fluctuations: The Inelastic Markets Hypothesis, Xavier Gabaix and Ralph Koijen, NBER, 2021
• The Disappearing Index Effect, Robin Greenwood and Marco Sammon, Journal of Finance, 2025
• How Retail Investors Are Using AI in 2026, Investing.com, survey of 938 US investors, March 2026
• Do LLM "Crowds" Produce Investment Signals? An Empirical Test, Steven Edwards, SSRN, 2026 (summary)
• When ChatGPT Stops Talking: GenAI, Retail Trading Coordination, and Market Liquidity, Xiaoxue He, Lancaster University, working paper, SSRN, September 2026
• Financial Stability Implications of Generative AI: Taming the Animal Spirits, Anne Lundgaard Hansen and Seung Jung Lee, 2025
• Herding where retail investors dominate trading: The case of Saudi Arabia, M. Arifur Rahman, Shah Saeed Hassan Chowdhury and M. Shibley Sadique, Quarterly Review of Economics and Finance, 2015
• QFIs buy SAR 1.9B worth of TASI stocks in July, Argaam, August 2025 (Tadawul traded value by investor type)
• QFIs, Saudi individuals net buyers of SAR 2.8B TASI stocks in Dec. 2025, Argaam (Tadawul traded value by investor type)
• TASI ownership by investor type, December 2025, Argaam (Tadawul data)