What happens when everyone uses AI to trade

It wasn’t that long ago that investors used to make wild hypotheses about the stock market and hoped they would work. Fetching data and learning new strategies were also part of it. 

Now, instead of following hard rules created and interpreted by humans, AI is deciding what risk you will take and managing your portfolio. 

Brokerages, finfluencers, retail traders, and everyday investors are using AI tools, and stacking ChatGPT prompt-packs— there are guides for prompts for intraday and short-term traders — on top of the tips, courses, and Telegram trading bots. 

GitHub has plenty of open-source trading bots that plug an LLM into your brokerage account, allowing a simple chat to drive your trades.

Banks are rolling out AI assistants; hedge funds are weaving large language models into research and trading, and financial advisers are racing to incorporate the technology into everything from portfolio construction to client service. 

WHAT ARE THE RISKS?

Well, the assumption is that better tools should produce better investors. AI is even leveling the playing field between individual and institutional traders. But here’s the thing: things could also go wrong if everyone asks the same chatbots for stock picks. Experts warn that AI could lead to crowded trades and a false sense of agreement in the market.

“The main risk is crowding. If millions of investors rely on similar AI models trained on comparable public information, they may arrive at the same conclusions and pile into the same trades. This can inflate valuations, reduce market diversity, and increase volatility when sentiment changes,” says Hamza Dweik, Head of Trading (MENA) at Saxo Bank, adding that there have been similar dynamics before with passive investing and social media-driven trading, but AI could accelerate the process by spreading the same investment ideas at machine speed. 

So how is that playing out? Sayed A., Chief Business Officer at Graystone Capital, says, “62% of retail investors now use AI for investment decisions, and 24% admit they worry about herding, yet keep using the tools anyway.”

He adds, “LLMs pull from overlapping data, so mass adoption narrows the idea pool instead of broadening it. Add finfluencers repackaging the same chatbot outputs as personal conviction, and ‘independent confirmation’ becomes an illusion.”

If everyone asks the same models for stock picks, markets could experience herding, even more prevalent than we have now, says Ahmad Chaudry, Global Head of Structured Investments and Derivatives at Klay Group.

“Investors may receive similar recommendations, causing large numbers of people to buy or sell the same stocks simultaneously. This could inflate prices, create bubbles, increase volatility, and trigger sharp crashes when sentiment changes. AI can also repeat the same outdated or incorrect information, meaning thousands of investors could make similar mistakes at the same time.”

According to Madhur Kakkar, co-founder of QuantL AI, generative AI can interpret broader information, but if investors ask similar models the same questions, they may receive increasingly similar ideas. “The opportunity lies in combining multiple independent signals with disciplined, guarded execution.”

Dweik agrees, saying, “Markets function best when participants hold different views, and that overdependence on a handful of AI systems could weaken that diversity of opinion.”

WHAT HAPPENS WHEN AI AGENTS COME?

Now comes the next phase of AI: agentic trading, in which AI doesn’t just recommend investments but executes them. While some research suggests they’ll be capable of manipulating markets that could threaten economic security, other research indicates they might act more rationally.

“AI agents go beyond producing recommendations,” says Kakkar. “They can observe, decide, and execute continuously, which offers enormous advantages in speed, precision, and 24-hour monitoring.”

But fully autonomous agents, he adds, may also interact unexpectedly, reinforce common behavior, or accelerate errors before humans can respond. 

According to Chaudry,  AI agents operating at scale could increase market volatility if they react to the same signals, news, or strategies simultaneously. “They might create feedback loops, amplify price movements, or trigger rapid buying and selling that overwhelms human investors.”

While poorly designed or insufficiently supervised systems could increase market swings, Dweik says, “More concerning is the possibility that malicious actors use AI to generate misleading content, manipulate sentiment, or exploit market vulnerabilities at scale.”

Pushing back on the notion of manipulating sentiments, Sayed A. says the real risk is that, even if AI agents behave exactly as designed, too many are designed the same way. “Bank of England’s Sarah Breeden warned in June that agents could ‘amplify volatility in stress.’ Tellingly, the Fed’s own April 2026 guidance excludes agentic AI entirely from its model risk scope. That gap concerns me more than any deliberate scheme.”

DEVELOPMENTS IN A DESIRABLE DIRECTION

Financial markets already have safeguards against manipulation and disorderly trading. Still, Dweik adds that “regulators and firms will need to ensure those frameworks evolve to address increasingly autonomous decision-making systems.”

Experts believe that AI should be used as a decision-support tool, not a replacement for personal judgment. Traders should compare AI-generated ideas with reliable sources, understand the risks, and avoid unthinkingly following popular recommendations. “Platforms should clearly explain how AI produces suggestions and warn users about potential errors. Models are also very sensitive to the quality of the input data. So, the data used to run these models should be validated, cleaned, and also subject to challenge– perhaps, by using multiple data sources,” says Chaudry.

The answer is not to block automation but to govern the freedom it is given, says Kakkar.

Financial institutions deploying AI must maintain clear human oversight, robust testing, and transparent risk controls. Dweik says, “Regulators can require firms to demonstrate that AI-driven trading systems are explainable, auditable, and subject to appropriate safeguards before they are deployed at scale. At the same time, investors should view AI as a tool for research and decision support rather than a substitute for judgment.

Ultimately, three things matter, Sayed A. says. Regulators closing scope gaps fast, like the FSB’s ongoing consultation; institutions building real kill-switch capability, not just paper plans; and investor education keeping pace. “My rule for clients: an AI’s stock idea is a starting point, never a substitute for your own thesis.”

AI is here to stay and evolve. The challenge is for traders to get the best out of it.