AI can write you a trading system in minutes. That’s the problem.
A system that compiles, places trades and shows a smooth backtest looks like progress. It isn’t, until you know whether it has a real edge, whether it was curve fit to the past, and what it will do to your account in a bad year. AI makes the first part effortless. It does nothing for the second part, and the second part is where the money is made or lost.
I have been a systematic trader for more than 25 years, and I now use AI in my own research and system development. AI has changed how fast I can test ideas. It hasn’t changed what makes a system worth trading.
This guide covers what AI trading actually means, what works, what doesn’t, and how to use AI without handing your account to a black box. It’s also the hub for all my articles on AI and trading, so you’ll find the deeper guides linked as we go.
What Is AI Trading?
AI trading is using artificial intelligence to find, test or place trades. That covers everything from machine-learning models that look for patterns in price data, to large language models like ChatGPT and Claude that analyse information and write code, to trading bots that place orders on their own.
The term hides three very different things, and it pays to know which one someone is selling you:
- AI as the trader. A bot or app makes the decisions and places the trades. You switch it on and hope.
- AI as the signal. A model predicts price moves or reads news and sentiment, and you act on its calls.
- AI as the research partner. AI helps you write, code, backtest and review a trading system whose rules you define and understand.
The first two ask you to trust the AI’s judgement. The third asks the AI to do the work while you keep the judgement. In my experience, only the third reliably helps private traders.
How Does AI Trading Work?
Every form of AI trading follows the same four steps: data goes in, a model processes it, a decision comes out, and the decision is executed, by you or by a bot.
- Data: price history, volume, fundamentals, news, even social media sentiment.
- The model: a machine-learning model trained to spot patterns, or a large language model that reasons over text and writes code.
- The decision: a buy or sell signal, a ranking of stocks, a forecast, or a block of trading code.
- Execution: either you place the trade, or a bot sends the order to your broker through an API.
It’s worth knowing how this differs from algorithmic trading. An algorithmic system follows fixed rules that someone wrote and can read. Many AI systems learn their own rules from data, and those rules can’t be read at all. That difference matters enormously when the system starts losing money and you need to decide whether to keep trading it. I explain it in more detail in how to use AI for stock trading.
Do AI Trading Bots Work?
Some bots run exactly as designed. Very few of them have an edge you can verify, and running is not the same as making money.
AI has made it easy to build a bot that connects to a broker and trades. Plenty of traders now have bots that work in the technical sense: they compile, they trade, they don’t crash. The question that matters is whether the bot beats a simple alternative after costs and on a risk-adjusted basis. Often it doesn’t even beat buying an index fund and holding it.
Then there’s the marketing. The US Commodity Futures Trading Commission has published a customer advisory titled AI Won’t Turn Trading Bots into Money Machines, warning about schemes that promise huge returns or near-perfect win rates from “AI-created” algorithms. If a product promises returns, shows you no rules and can’t give you a backtest you can rerun yourself, walk away.
Before you trust any system an AI hands you, bot or not, read can you trust a trading system generated by AI.
Can You Make Money With AI Trading?
Yes, but the money comes from a tested edge and proper risk management, not from the AI.
AI doesn’t create an edge. It speeds up the work of finding and testing one. If you already understand how to design, backtest and validate a trading system, AI lets you test far more ideas in the same time. If you don’t, it lets you build untested and curve-fitted (overfitted) systems faster, which is a quicker way to lose money.
That’s why I see two very different outcomes:
- Traders who already know the systematic process use AI to go faster, and their results improve.
- Traders who skip the learning process just ask AI for a strategy, trade whatever comes back, and abandon it at the first drawdown because they have no idea whether it still works.
If you’re weighing up whether it’s worth getting into at all, I give my honest assessment in is AI stock trading worth it.
If you want to see what a rules-based approach looks like before you bring AI into it, watch my free 10-Minute Trading Formula training, where I show how to trade the markets in as little as 10 minutes a day with a rules-based approach.
What Are the Risks of AI Trading?
The risks of AI trading are mostly the old risks of trading, made faster and harder to see.
- Overfitting disguised as intelligence. AI can find a pattern in any dataset, and most bots pick the best-performing settings from history. That’s curve fitting with a better story and an overly confident LLM giving you the spurious recommendations. It looks brilliant in the backtest and falls apart in live trading.
- No verifiable edge. If a system can’t show you the rule behind every trade, you can’t check whether its edge is real. Without that, you’re gambling.
- The black box. Every system eventually hits a rough patch. When a black box does, you have nothing to diagnose, so you’re left watching your account fall with no basis for deciding whether to keep going. Black box trading systems have been around for decades, they have never worked because they are just marketing scams and using curve-fit rules. Sticking an AI badge on it does not change that, so don’t fall for the hype.
- Missing risk management. Most AI tools focus on entries and exits. Very few deal properly with position sizing, portfolio heat or drawdowns, and that gap alone can destroy an account.
- Outsourcing your understanding. The more you let AI make decisions you don’t understand, the less able you are to manage your trading when something goes wrong. This is the risk most traders never see coming.
There’s also a quieter risk: AI-written backtests often contain the classic backtesting mistakes, such as survivorship bias, future leaks and unrealistic fills. The AI won’t warn you about them unless you know to ask.
Is AI Trading Legal?
Using AI tools to research and place trades in your own account is legal in the US, Australia and most other markets. What’s regulated is giving financial advice, managing other people’s money or selling trading signals, whether a human or an AI does it. And fraud is fraud: regulators actively pursue “AI trading” schemes that make false promises. If a product claims guaranteed or outsized returns, treat that as a warning sign, not a feature.
How Do Systematic Traders Use AI?
I treat AI as the lab technician, and the trader as the scientist.
The trader decides what to test, writes down the rules, sets the criteria a system must meet and makes every final decision. The AI does the heavy lifting: it writes the code, runs the backtests, applies the criteria and keeps a full written record of every test.
In practice, that looks like this:
- I define the system in plain English first: the setup, entry, exit, stop and position sizing. If I can’t write it down, it isn’t ready to test.
- AI turns the rules into code and runs them in my backtesting software. I use RealTest.
- The test plan is fixed before the first test, including which data is in-sample and which is held back as out-of-sample, so the AI can’t quietly tune the system to the data it’s judged on.
- Every rule has to earn its place. If removing a rule doesn’t make the system worse, the rule goes.
- Everything is written down, so I can see exactly what was tested, what passed and why.
- The AI never decides alone. Accepting, rejecting or trading a system is always my call.
This is the opposite of asking a chatbot for “a profitable strategy”. The AI makes a disciplined process faster. It doesn’t replace the discipline.
It’s also why I keep AI’s access tight. When I run backtests with Claude Code, it works in one dedicated folder with limited access, and I review what it asks to do. Speed is no excuse for giving any tool the keys to your computer or your brokerage account.
Inside the Trader Success System, members learn this exact workflow with EST’s AI Trading Toolkit: running my backtesting process with Claude Code and RealTest, from first idea to a fully validated system.
For more ways to use AI across your trading, from screening stocks to reviewing your systems, see my guide on how to use AI for stock trading.
How Do You Check an AI Trading Tool Before You Trust It?
Run any AI trading tool, bot or AI-written system through these six checks before you risk a dollar on it:
- Can you see every rule? Ask: “What exactly triggers every buy and sell?” If the answer is no, walk away.
- Can you rerun the backtest yourself? Don’t accept a backtest you can’t reproduce. Check it for survivorship bias, future leaks and costs.
- Does it hold up across different markets? Test it through bull, bear, sideways and volatile periods. If small changes to the settings make the results collapse, it’s fragile.
- Does it beat doing nothing? Compare it with buying and holding the index, after costs and on a risk-adjusted basis.
- Does it manage risk? Position sizing, maximum drawdown and what happens in a crash should all be defined.
- Can you explain it in plain English? If you can’t explain why it should work, you won’t keep trading it through a drawdown.
Then trade it small, or on paper, for a few months before you commit real capital. That’s where the gap between a brilliant backtest and real trading shows up.
AI Trading Guides
These guides go deeper into each part of AI trading:
- Is AI stock trading worth it? An honest look at what AI can and can’t do for a private trader.
- How to use AI for stock trading: practical ways a systematic trader can use AI, and what it can’t do.
- Can you trust a trading system generated by AI? How to tell an AI-generated algorithm with an edge from one that only looks good.
AI trading builds on the fundamentals. If you’re new to systematic trading, start with trading systems, automated trading systems and backtesting.
Frequently Asked Questions
Can I use ChatGPT or Claude for trading?
Yes, as a research assistant. Large language models are good at writing trading code, explaining concepts, summarising research and reviewing your rules. They shouldn’t be trusted to pick trades on their own, because they can’t verify that an idea has an edge. Use them to build and test systems, and keep the decisions yourself.
Can AI predict the stock market?
Not reliably. Markets are noisy, and patterns that worked in the past often stop working. Any edge an AI finds still has to be tested on data it wasn’t built on, and most apparent edges disappear once costs and realistic fills are included.
What is the best AI for trading?
The best AI for trading is the one that helps you build and test systems you understand. For a systematic trader, that usually means a capable language model working in the coding interface (Claude Code or Codex) alongside proper backtesting software (I recommend RealTest), combined with comprehensively designed custom skill files which teach the AI model exactly how you want the backtesting process / analysis / decisions to be done. This is very different to blindly throwing money at an “AI trading bot” that promises higher returns, accurate signals and low drawdowns.
Is AI trading good for beginners?
Only after you have truly learned the basics. A beginner with AI tools can code systems faster than they can judge them. This is dangerous and many traders will blow up their accounts as a result. Learn how to design, backtest and manage the risk of a trading system first, then use AI to speed up the process.
How much does AI trading cost?
The tools themselves can be cheap or free. The real cost is the money you can lose trading an untested system. Proper data and backtesting software usually matter more than which AI tool you use.
Can I really make money with AI trading?
You can make money with trading systems that have a tested edge and sound risk management. AI can help you find and test them faster, but only if you give it the correct process to follow. It can’t give you a viable edge unless you have explicitly taught it how to do that – that is what we teach in the Trader Success System.
Use AI to Go Faster, Not to Skip the Process
AI is genuinely powerful, and it’s changing how quickly a trader can research and test ideas. But it’s a tool for traders, not a trader. The people who succeed with it already understand systematic trading. The people who struggle are handing their thinking to a tool they can’t evaluate.
If you want to learn the process properly and then speed it up with AI, the Trader Success System takes you from your goals to a portfolio of backtested trading systems. That includes EST’s AI Trading Toolkit, which is our proprietary backtesting and analysis process codified into AI agent skills so you learn to run your backtests with Claude Code. The EST AI Trading Toolkit even teaches you exactly what it is doing at every step, or can be run in fully autonomous mode (you still make the final calls). This is how I managed to backtest over 500 trading system ideas in under 3 months.
Remember – You are only one trading system away!
