An AI label tells you little about a trading product's reliability. A bot may use fixed rules, a statistical model, a language model or a combination. Ask what decisions it actually makes, what data it uses and what authority it has over orders.

Separate Analysis From Execution

A tool that summarises news is different from one that chooses position size or sends live orders. A trained model may remain fixed between releases; “machine learning” does not imply continuous adaptation. Automated software can also malfunction or receive bad data even when its intended rules are clear.

The CFTC's AI trading-bot advisory warns against treating AI as a source of guaranteed profits. A sophisticated description or impressive dashboard is not evidence of future returns.

Ask for the Complete Performance Basis

Find out whether results are live, simulated or backtested; whether they include all accounts and losing periods; and how fees, spread, slippage and financing were handled. Inspect drawdown and exposure as well as win rate.

A very high win rate can coexist with rare large losses. A smooth backtest can result from favourable fill assumptions or rules selected on the same data used for evaluation. Neither a particular win-rate cutoff nor a polished explanation proves fraud or legitimacy on its own.

Require a version history. If the vendor changes the model, an older record may not describe the product now offered.

Limit What the Tool Can Do

Before connecting an account, establish who holds funds and what the requested credentials permit. Prefer the least access needed for the function, and understand how to revoke it. A local stop button is not enough if existing orders remain active at the broker.

In simulation, check rejected orders, partial fills, duplicate messages, stale prices, a lost connection and restart recovery. Record the actual account state after each event. Successful submission does not establish a completed fill.

Count the Full Cost

Subscription fees are only one input. Data, hosting, execution, financing and support can add costs. Verify the current terms of the specific product rather than relying on a generic monthly-price range.

You do not need to become a discretionary chart trader before studying automation. You do need to understand the instrument, order workflow and loss controls. Begin with a bounded, observable function and keep live execution a separate decision. A bot is software with trading authority, not passive income.

Risk Disclaimer: Trading involves substantial risk of loss. Past performance is not indicative of future results. See our full risk disclaimer.