LaoLong

AI trading

What is an AI crypto trading bot? How it works and what AI really does

An AI crypto trading bot uses software and machine learning to trade for you. What the AI really decides, what the evidence says, and how to check claims.

An AI crypto trading bot is software that trades cryptocurrency automatically and uses artificial intelligence in at least one part of the process: picking trades, sizing them, or deciding when to step back. The useful question is not "does it have AI?" but "which decision does the AI make, and how do I verify the result?" That's what this guide covers.

The short version

  • A trading bot automates three jobs: reading data, deciding, and placing orders on an exchange.
  • "AI" can mean a classic machine-learning model, a large language model (LLM) such as GPT or Claude, or a marketing label on simple rules.
  • The evidence for AI as a price forecaster is weak. The US Commodity Futures Trading Commission puts it plainly: "AI technology can't predict the future or sudden market changes" (CFTC advisory, January 2024).
  • The evidence for systematic rules such as trend following is much stronger and goes back decades (sources below).
  • A sensible design, and the one LaoLong uses, puts rules in charge of trading and has AI watch over the risk.

How an AI trading bot works, step by step

Every automated trading system, AI or not, runs the same loop:

  1. Data. Prices, volume, order books, funding rates and open interest from derivatives markets, and sometimes news or social media.
  2. Signal. A rule or model decides whether to go long, go short or do nothing.
  3. Risk. A separate layer sets position size, stop-loss, maximum exposure and when to cut risk.
  4. Execution. Orders go to an exchange through an API key, a credential that lets the software trade on an account.
  5. Bookkeeping. Fills, fees and profit or loss are recorded.

AI can sit in step 2 (prediction), step 3 (risk control) or both. Where it sits matters more than any other design choice.

Three kinds of "AI" in trading bots

Type What the AI does Strengths Weak spots
Rules dressed as AI Nothing. Fixed indicators like moving averages, marketed as "AI" Transparent if the rules are disclosed Misleading label; regulators fine this ("AI washing")
Machine-learning predictor A model trained on past data forecasts returns Can find patterns in large datasets Overfitting: strong backtests that fail live; drifts when markets change
LLM agent A language model reads data and news, then decides trades in plain language Reads context and news fast Inconsistent decisions, no proven edge as a forecaster (see the 2025 test below)
Rules + AI risk overlay Rules trade; AI decides how cautious to be Uses each tool where evidence supports it AI lowers risk but cannot remove it

What happened when LLMs traded real money

In October–November 2025, the research lab Nof1 ran "Alpha Arena" Season 1. Six leading models each got $10,000 to trade crypto perpetual futures on Hyperliquid with no human intervention, from 18 October to 3 November 2025. Alibaba's Qwen 3 Max finished first with about +22%, DeepSeek also finished in profit, and the other four models lost money. GPT-5 ended with roughly $3,733 of its $10,000 (GN Crypto summary of final results; season overview, KuCoin News). Nof1 itself pointed to small samples and a short test period.

Seventeen days proves little about any single model. It does show that the same instructions gave very different results across models, from −63% to +22%, which is the opposite of a reliable forecaster.

What research says about AI reading news

Academic work points to what language models are actually good at. Lopez-Lira and Tang found that GPT-4 predicted the direction of a stock's initial reaction to a headline with daily portfolio hit rates of about 90%, but that is the immediate move, which is hard to trade, and the authors report that returns from such strategies decline as more traders use LLMs (Lopez-Lira & Tang, arXiv 2304.07619, v6 October 2025). In plain terms, AI reads news quickly, and the edge from reading news quickly gets competed away.

That makes a language model a good watchman (is this news dangerous for an open position?) and a poor oracle (where will price be next week?).

What actually has evidence: systematic trend following

The best-documented edge in crypto is not AI but momentum and trend:

A backtest is not a guarantee. Trend strategies lose money in choppy, sideways markets and can go through long drawdowns. But a rule you can describe and test is something you can check, and "the AI knows" is not.

Red flags in AI trading products

Regulators have published the patterns:

  • Guaranteed or fixed returns. The CFTC describes a bot that "guaranteed at least a 10 percent monthly return" and turned out to be a Ponzi scheme (CFTC). The same advisory cites Mirror Trading International, a roughly $1.7 billion bot fraud.
  • Vague AI claims. On 18 March 2024 the SEC fined two investment advisers, Delphia ($225,000) and Global Predictions ($175,000), for claiming AI capabilities they could not substantiate (SEC press release 2024-36).
  • No full trade history. Screenshots of winning trades without the losses, without dates, or without a way to check them.
  • Unclear custody. You can't tell who controls the money or whether the trading key can withdraw it.

Questions to ask any AI trading bot

  1. Which decision does the AI make: entries, sizing, exits or risk mode?
  2. Can I see every closed trade, losses included, with dates and a definition of "win"?
  3. Where is the money held, and can the key that trades also withdraw?
  4. What happens in a crash: is there a stop on every position, and who triggers it?
  5. What does it cost: exchange fees, funding, subscriptions, profit share?

How LaoLong AI Trading Bot does it

LaoLong AI Trading Bot trades crypto perpetual futures on Hyperliquid for you. It splits the work the way the evidence suggests:

  • A rules-based trading engine makes the trades. It trades liquid coins long and short and follows trends, the family of strategies with the longest research record.
  • An AI risk officer decides how carefully to trade. It reads the market and the news around the clock and moves the account between three states, Normal, Caution and Defense. It can filter new entries when conditions turn, hedge the account with BTC, and close a trade that became dangerous before its stop is reached. It does not invent trades, and it lowers risk rather than preventing every loss.
  • Guardrails don't depend on the AI. Trades on the AI's calls use isolated margin with a stop order resting on Hyperliquid itself. The trend book has an emergency stop checked every minute. The trading key can place orders but cannot withdraw funds.

Our published track record is also built to be checked: 214 of our AI's trade calls have closed in profit, and the win rate of the public AI trade calls is 78.1% (as of 17 Nov 2026, 15:34 UTC). Every closed call is listed, losses included, and each public call links to its original post. These records measure price moves from entry to exit without leverage. The methodology page explains exactly what each number measures.

Bottom line

An AI crypto trading bot is only as good as the decision you trust it with. Rules with a research record, an AI that guards the risk, and a full public history of trades is a sound combination. A promise that "the AI predicts the market" is the warning sign regulators keep describing.

Questions and answers

What is an AI crypto trading bot?

It is software that places crypto trades automatically and uses some form of artificial intelligence, such as a machine-learning model or a large language model, in at least one step: choosing trades, sizing them, or deciding when to cut risk.

Can AI predict crypto prices?

Not reliably. The US CFTC states that 'AI technology can't predict the future or sudden market changes.' In a 2025 live test on Hyperliquid, four of six leading language models lost money over 17 days. AI is more useful for reading context and managing risk than for forecasting.

Are AI trading bots safe?

The software can be well built and still lose money, because markets carry risk. Check who holds the funds, whether the trading key can withdraw, whether every past trade is published with losses, and whether the AI claims are explained in concrete terms.

How does LaoLong AI Trading Bot use AI?

A rules-based trading engine makes the trades. An AI risk officer reads the market and the news around the clock and switches the account between Normal, Caution and Defense: it can filter new entries, hedge with BTC and close a trade that became dangerous before its stop.