Is AI Trading Legit? What the Evidence and Regulators Say

"Is AI trading legit?" is really three questions bundled together. Is the technology real? Is it legal? And does it make money for ordinary people? The honest answers are different for each, and mixing them up is exactly how bad products get sold.

This article is educational, not investment advice. Any form of trading, automated or not, can lose money.

Question 1: Is the Technology Legit?

Yes. Computer-driven trading is not fringe. A Coalition Greenwich survey of buy-side equity investors found that roughly 37% of their 2023 U.S. equity volume was executed through algorithms and/or smart order routers, and respondents expected that share to rise. Large trading firms, asset managers, and banks have used algorithms for decades to split up big orders, make markets, and run quantitative strategies. Machine learning and, more recently, large language models are now part of that toolkit.

But notice what that statistic describes: professional firms with teams of engineers, expensive data, direct market access, and compliance departments. It says nothing about whether a subscription app sold to individuals can do the same thing. That is the gap that scammers and aggressive marketers exploit. The existence of real algorithmic trading is used as borrowed credibility for products that have little in common with it.

Question 2: Are Retail "AI Bot" Products Legit?

Some are legitimate software tools: they connect to your brokerage account, execute rules you define, and are honest about what they do. But the category has a serious and well-documented problem, and U.S. regulators have said so repeatedly.

Joint regulator warnings. On January 25, 2024, the SEC's Office of Investor Education and Advocacy, FINRA, and NASAA issued an investor alert about investment frauds involving AI. It describes unregistered platforms promoting AI trading systems with claims like "Our proprietary AI trading system can't lose!" and cautions investors to be wary of claims, even from registered firms, that AI can guarantee returns. The same day, the CFTC issued its own advisory, "AI Won't Turn Trading Bots into Money Machines," which says AI technology cannot predict the future or sudden market changes and lists red flags such as guaranteed returns, referral bonuses, social media influencers, and fake demo accounts.

Real cases. The CFTC's advisory points to Mirror Trading International, in which, the agency says, Cornelius Steynberg stole more than $1.7 billion in bitcoin from at least 23,000 people over about three years. It promised a minimum 10% monthly return, but ran as a Ponzi scheme, with little actual trading. More recently, the SEC alleged in May 2026 that a Texas man raised about $12.3 million from roughly 150 investors by touting AI-based crypto trading bots, with promised returns of 40% to 50% in 30 to 45 days, when the bots did not function as represented. In September 2026 it charged the operators of two platforms, Cryptoaiml and TSAI Pro, with defrauding investors of at least $15 million between them, alleging that one claimed "98% accuracy" for AI trading signals and that no trading took place on either platform. These are allegations in court filings; defendants are presumed innocent unless and until a court finds otherwise.

"AI washing." Not every problem is outright theft. Some firms simply overstate what their AI does. In March 2024 the SEC announced settled charges against two investment advisers, Delphia (USA) and Global Predictions, over false and misleading statements about their use of AI, with $400,000 in combined civil penalties. In 2025 the SEC and the Justice Department also brought actions against technology companies over inflated AI claims. As then-SEC Chair Gary Gensler put it, "Investment advisers should not mislead the public by saying they are using an AI model when they are not."

The scale of the broader problem. The FTC said consumers reported losing $7.9 billion to investment scams in 2025, the biggest category of fraud losses. That figure is not specific to AI, but it frames the environment you are shopping in.

The takeaway: the category is not a scam, but the marketing in it deserves more skepticism than almost any other corner of retail finance. Our checklist for evaluating AI trading bots walks through how to check a specific product.

Question 3: Is AI Trading Profitable?

This is the hardest question, and the answer is: we do not have good evidence that retail AI trading products reliably beat the market after costs, and many reasons to doubt it.

What independent evidence exists is mostly discouraging for active retail trading in general. Barber and Odean's classic study of about 66,000 U.S. households (1991 to 1996) found that the households that traded most earned an annual return of 11.4%, versus 17.9% for the market. A study of Brazilian futures day traders by Chague, De-Losso and Giovannetti found that 97% of those who persisted for more than 300 days lost money. And a study of Taiwanese day traders (Barber, Lee, Liu and Odean) found that less than 1% predictably and reliably earned positive abnormal returns after fees. None of these studies is about bots; they are about human traders. They matter because automation does not remove the two main obstacles: costs and competition. A bot that trades constantly pays the same spreads and commissions, and it competes against professional firms with better data and faster systems.

Research on AI specifically is mixed and early.

• A widely cited paper by Lopez-Lira and Tang found that ChatGPT-style models could predict stock reactions to news headlines better than traditional sentiment methods, but also reported that strategy returns decline as LLM adoption increases, which is what you would expect as more people use the same tools.
• A 2025-2026 paper by Gao, Jiang and Yan on "lookahead bias" found that a non-trivial share of the apparent predictive power of LLM forecasts reflects memorization of training-time material rather than genuine reasoning, so impressive backtests of an LLM on past news can overstate what it can do going forward.
• In live conditions, results have been poor. Bloomberg reported in May 2026 on a series of contests run by a startup called Nof1, in which eight leading AI models each traded $10,000 in U.S. tech stocks for two weeks across four contests. The portfolio overall lost about a third of its capital, and models finished in profit in only 6 of 32 results. The contests were small and short, and Nof1's founder said models need substantial scaffolding and data around them to have a chance. A separate academic benchmark, LiveTradeBench, evaluating 21 LLMs over 50 days of live trading, concluded that high scores on standard AI benchmarks do not imply superior trading outcomes.

Backtests are the trap. Bailey, Borwein, López de Prado and Zhu showed that high simulated performance is easy to achieve by testing enough strategy variations, which is how many vendors produce impressive-looking historical charts. A bot's past results, shown without verified live performance, are an advertisement, not evidence.

Could some AI-driven strategies be profitable? Of course; some professional quant firms are. But as one hedge fund data scientist quoted by Bloomberg noted, if profitable AI trading emerges you probably will not hear about it, because there is no incentive to share. That is a good reason to doubt that a mass-market product available to anyone has found an edge it is willing to sell to you cheaply.

Question 4: Is AI Trading Legal?

In the U.S., generally yes. It is legal to automate your own trades through a brokerage account, whether through a broker's tools, an API, or an AI assistant, subject to the broker's terms and the usual rules. What is not legal:

• Fraud. Lying about returns, strategy, registration, or what the "AI" does violates securities and commodities law, as the cases above show.
• Market manipulation. Automation does not excuse spoofing (placing orders you intend to cancel to mislead the market), wash trading, or other manipulative practices; FINRA's guidance on algorithmic trading specifically points to manipulative strategies such as spoofing. A bot you run is your responsibility.
• Unregistered advice for pay. Under the Investment Advisers Act, a person or firm that, for compensation, is in the business of advising others about securities generally must register as an investment adviser unless exempt. Whether a specific software product crosses that line depends on the facts, which is why checking registration matters.
• Trading on illegal inside information, whoever or whatever pulls the trigger.

Being legal is also not the same as being safe. Many legal products are still bad bets.

The Sober Bottom Line

• Algorithmic trading: legitimate and mainstream, but dominated by professionals.
• Retail AI bot products: a mixed bag, with a documented fraud problem and regulators repeatedly warning against guarantees. Treat every performance claim as unproven until it is verified, live, and audited.
• Profitability: unproven for retail products. The best available evidence on retail trading, with or without bots, is that costs and competition make outperformance hard.
• Legality: automating your own trades is legal. Fraud, manipulation, and unregistered advice are not.

If you want to use AI as a tool rather than a promise, treat it as a research assistant, keep custody with a regulated broker, limit what you risk, and verify everything against primary sources. For example, our insider tracker and congressional trading tracker show filings straight from required public disclosures. And if you are curious about how a regulated broker is structuring AI agents, read our explainer on Robinhood's agentic trading.

Risk, in Plain Language

You can lose some or all of the money you put into any trading strategy, and automation can make losses happen faster. Backtested or "paper" results do not guarantee live results. A tool that claims to be risk-free or guaranteed is making a claim no legitimate firm can make. Only risk money you can afford to lose, and consider speaking with a licensed professional before making significant decisions.

We track insider and congressional trading filings so you can follow disclosed, verifiable activity rather than a black-box promise. Free. Drop your email below.

Sources: SEC/NASAA/FINRA Investor Alert: AI and Investment Fraud, FINRA: AI and Investment Fraud, CFTC Customer Advisory: AI Won't Turn Trading Bots into Money Machines, SEC press release 2024-36 (Delphia, Global Predictions), SEC Litigation Release 26558 (Fuller), SEC press release 2026-95 (Cryptoaiml, TSAI Pro), SEC litigation release on Nate, Inc. founder, FTC testimony on fraud, March 2026, Barber and Odean, Trading Is Hazardous to Your Wealth, Chague et al., Day Trading for a Living?, Barber et al., The Cross-Section of Speculator Skill, Lopez-Lira and Tang, Can ChatGPT Forecast Stock Price Movements?, Gao, Jiang and Yan, Detecting Lookahead Bias in LLM Forecasts, LiveTradeBench, Bloomberg: AI Bots Auditioning for Wall Street Trading Are Mostly Losing, Bailey et al., Pseudo-Mathematics and Financial Charlatanism, Coalition Greenwich via The TRADE, FINRA Regulatory Notice 15-09