Do AI trading bots work? An honest answer for crypto traders

The honest answer

AI trading bots work in a narrow but useful sense: they follow rules reliably, around the clock, without fear, greed or fatigue. They watch markets you cannot watch and act the moment a rule is met.

What they do not do is turn a weak strategy into a winning one. A bot is an amplifier. Good rules, tested honestly and run with limits, become easier to follow. Weak rules lose money faster and more consistently than they would by hand.

So the useful question is not "do AI trading bots work?" but "do these rules work after fees, and will the bot stop when they do not?"

What bots do well

• Consistency. A bot never skips a rule because it is tired, busy or scared.
• Speed and coverage. It watches every minute of a market that never closes.
• No emotion. It sells at the stop even when you would hope for a rebound.
• Record keeping. A good bot logs every decision, which helps you learn what your rules actually do.

Why many bots lose money

The reasons are rarely mysterious:

• The rules never had an edge. A rule that sounds clever but did no better than holding the coin in a fair test will not suddenly start working live.
• Backtests without fees. Many rules look good before costs and lose after them, especially ones that trade often.
• Over-trading. Bots that check the market every second and trade on tiny moves pay fees and slippage on every one.
• Leverage. Borrowed money multiplies losses as well as gains, and liquidations can close a position before any rule acts.
• No hard limits. Without a stop-loss and a daily loss limit, one bad day can undo months.
• Strategy hopping. Switching to whatever worked last month means always arriving late.

The AI part: what to believe

Since 2026, AI assistants can place trades through exchange tools built for agents. That makes "an AI that trades for you" easy to build. It does not make it profitable.

Language models are good at turning your words into precise rules and at explaining what happened. They are not reliable price forecasters, and they do not hold limits unless something outside the model enforces them. Be sceptical of any product whose pitch is that its AI predicts the market. Prefer one where the AI helps you write and understand the rules, and a plain rules engine enforces them.

How to spot a scam

Walk away if you see any of these:

• Promised daily, weekly or monthly returns, or any "guarantee" of profit.
• Results with no period, no fees and no worst drop.
• Screenshots of gains in place of downloadable, dated records.
• A request to deposit money with the bot company rather than keeping it on your own exchange.
• A request for an exchange key that can withdraw.
• Pressure to recruit friends or upgrade quickly to "unlock" better results.

How to test a bot honestly

1. Write down the rules. If you cannot state what the bot will do, you cannot judge it.
2. Backtest with fees and slippage over more than one period, and compare with simply holding the coin.
3. Check the worst drop, not just the result. Could you live with it?
4. Run it on paper with live prices until it has traded several times.
5. Go live small, with a daily loss limit. TradeCorgi caps a live agent at 20% of your planned amount in its first week.
6. Judge on many trades. Ten trades tell you very little; fifty start to.

A worked example

Two people run the same dip-buying rule on BTC.

The first finds it on social media, sees a screenshot showing gains, connects a bot without a backtest, and adds leverage. A fast fall hits two days later; the leveraged position is closed by the exchange before the stop-loss can act.

The second describes the rule, sees a 90-day backtest after fees that does slightly better than holding with a smaller worst drop, runs it on paper for a month, then goes live with $100 and a $25 daily loss limit. The same fast fall triggers one stop-loss and pauses the agent for the day.

Same rule. Very different outcomes, decided entirely by testing and limits. This is an illustrative example, not a real result.

What a realistic first month looks like

A careful first month is quiet. A dip-buying rule might trade three or four times; a weekly buyer four times. Some trades gain, some hit their stop, and the agent stays out most of the time. That is normal. Judge the rules on whether they did what you wrote, not on one month's result.

What TradeCorgi does and does not claim

TradeCorgi does not promise returns and does not tell you what to trade. It turns your rules into an agent, tests them with real fees, runs them on paper first, and enforces a budget, stop-loss, daily loss limit and open-position cap before every live order. Every trade comes with a reason. Whether the rules make money is decided by the rules and the market.

A checklist before trusting any bot

• The rules are written down, and you understand them.
• The backtest includes fees, slippage, a period and the worst drop.
• It has been run on paper on live prices.
• The exchange key is trade-only, with withdrawals off.
• Loss limits are enforced before every order.
• You can stop it with one tap.
• Nobody promised you a return.