Tutorials Get the App

Home · Learn · Does AI trading work

Does AI Trading Work? Is It Profitable? What the Evidence Says

Updated November 22, 2026 · By Moby

AI trading works as a tool for research, rules and execution, but it does not reliably make money on its own. No AI can predict markets. Most active large-cap funds have trailed a simple benchmark, and fees decide a lot of the rest. Treat any promise of steady profits as a red flag.

Short answer: what "works" means

People pack three different questions into "does it work." One: does the software run and send an order. Two: does it follow the rule you wrote. Three: does it beat holding. Only the first two are reliable. The third is a research claim, and the evidence below is why this page will not make it for you.

A bot that runs is a tool. A bot that follows a written rule is discipline. A bot that is sold as a paycheck is a different product, and it is the one regulators keep warning about. Keep the three meanings separate when you read a sales page. If the page only answers the third, and with a screenshot, you are not looking at evidence.

What the evidence says

The CFTC press release of January 25, 2024, Release 8854-24, checked November 22, 2026, says AI technology can't predict the future or sudden market changes. The same release says claims of high or guaranteed returns are red flags, and it walks through a pool that took more than $1.7 billion in bitcoin from at least 23,000 people while showing fake demo balances. No AI can predict markets.

Professionals miss simple benchmarks too. S&P Dow Jones Indices, in the SPIVA U.S. Year-End 2024 summary checked November 22, 2026, says 65% of all active large-cap U.S. equity funds underperformed the S&P 500 in 2024. Over the 15-year period ending December 2024, the same summary says there were no categories in which a majority of active managers outperformed. That is fund managers with staff and data, not a chat window. It is not a Solana statistic. It is a reason to doubt anyone who skips the benchmark.

Public contests where models trade real money are a separate story, with their own figures. This page does not repeat those figures. The sourced account is on public AI trading contests. A contest window is not a career, and a leaderboard is not a rule you can paste into your account.

Where AI helps in trading

AI is useful at clerical work that traders skip. It can summarize a token page, turn a checklist into questions, apply a rule the same way on Tuesday as on Friday, and do position size arithmetic without rounding in your head. It can watch a log and tell you which rule fired. Those jobs are real. They are also boring, which is why people skip to the profit screenshot.

Copyable versions of the clerical jobs are on AI trading prompts. Use them as drafts. Check every number the model types against a source you opened. A summary that invents a holder count is worse than no summary.

Where AI fails

It fails when asked to call the next price, which is the job the ads sell. It fails when the market regime changes and the old rule keeps firing. It fails when the backtest was fit to one path. It fails when it states a number it did not look up. And it fails when the input was manipulated, including volume that was printed to look like demand. That last failure mode is explained on Solana volume bot. A model that treats volume as truth will treat a fake tape as truth.

Worked example: what a bot must earn just to cover fees

Example, not a recommendation. The fee schedule is the one on this site, used as arithmetic, not as a backtest of a strategy. A 0.75% swap fee (minimum $0.90 per swap), plus onchain fees. At small sizes the $0.90 minimum dominates. At about $120, 0.75% equals $0.90, so larger trades pay 0.75% each way.

Trade sizeFee per tradeRound tripRound trip as a percent of size
$20$0.90$1.809.0%
$50$0.90$1.803.6%
$120$0.90$1.801.5%
$500$3.75$7.501.5%

Frequency is the second half. Five round trips a day at $50 cost $1.80 times 5, which is $9.00 a day. On a $500 wallet that is 1.8% of the wallet per day, about 54% over 30 days if you add the days, before any losing trade. Most bots that "stop working" were never ahead of their costs. They were a fee machine that had a good week. Run your own sizes on the trading fee calculator and the crypto profit calculator.

Does AI stock trading work?

The same split applies. Software can screen and send orders. It does not reliably beat a simple hold. On Moby, agents trade tokenized stocks on Solana, which are tokens, not brokerage shares. Whether you can buy a tokenized stock depends on where you live and on the issuer's rules. The wider map is AI stock trading. A list of suggested names is a different job from sending the order.

How to test AI trading without big risk

Five steps, and none of them is "deposit and hope." Write the rule in one sentence, including the exit. Backtest it with fees included. Paper trade or replay a log until the refusals make sense. Go live small, inside limits. Review once a week and stop if you cannot explain the trades. You set your agent's per-trade and daily spending limits and market-cap range; Moby blocks red-rated tokens. Limits are only as tight as you set them. Sells and exits skip the market-cap and safety checks, and tokens with no safety rating are not blocked. Turn on Ask before every trade and your AI asks you first. Moby does not hold or approve trades. One agent per user, with its own wallet. Agents trade Solana tokens only. The agent cannot withdraw, transfer or send funds. A browser backtest is the backtest demo, and the common traps are on backtesting with AI.

Is AI trading worth it?

It is worth it if you want a tool: faster notes, a rule that does not flinch, a log you will actually read. It is not worth it if you need the money, or if you are buying a promise. The fee table is the test. If your plan needs many small round trips, the tool can be faithful and still lose. Treat any promise of steady profits as a red flag, which is the CFTC's point, not a slogan.

A practical way to keep the test honest is to write the benchmark next to the bot. If the alternative was holding the same tokens, the bot has to beat that hold after fees, not beat zero. If the alternative was staying in USDC, the bot has to beat zero after fees, which the $50 example shows is a high bar at five trades a day. Is AI trading legit is the page for scam patterns. This page stays on the evidence and the costs. Crypto trading bot covers rule bots that are not agents, and agentic trading is the agent setup if you still want one after the fee table. The bot comparison is AI trading bot.

A short note on Moby

A 0.75% swap fee (minimum $0.90 per swap), plus onchain fees. Limit, take-profit, stop-loss and trailing orders carry the same flat 0.75% Moby fee as swaps, charged when the order fills. Fees on your agent's trades earn up to 10% cashback in USDC. It is not a discount on the swap fee. AI Trading and copy trading currently need an access pass in the app. Download Moby. Limits and order types are also on AI trading risk management and agent strategies.

What to write down after a week

A week is long enough to see costs and too short to crown a method. Write five lines. How many round trips. The fee dollars, not the percent. The win rate. The result of doing nothing with the same starting balance. Whether you can explain every order without opening the model. If the do nothing line won, the tool still ran. It did not work in the third sense from the top of this page. Keep the log. A feeling that it was a good week is not a number.

Frequently asked questions

Is AI trading profitable?

Sometimes for someone, and not as a rule you can buy. Funds with staff often trail a simple index, and fees take a large share of small trades. No AI can predict markets. Treat a steady profit promise as a red flag, not as a track record.

Do AI trading bots work?

They work as software: they can run a rule and send an order. They do not reliably beat holding. A bot that fires all day on small size can spend its edge on fees before the rule has a chance. Check the round trip cost first.

Does AI stock trading work?

Same answer as crypto. Screening and order entry can work. Beating a hold is a separate claim. On Moby, stock exposure is a tokenized stock on Solana, not a brokerage share. Whether you can buy one depends on where you live and on the issuer's rules.

Is AI trading worth it?

Worth it as a tool if you want notes, consistent rules and a log. Not worth it if you need the money or want a promise. The fee table on this page is the practical test: many small trades can cost more than the idea was worth.

Can you make money with AI trading?

Some people do. Most active approaches do not beat a simple benchmark once costs are counted, and nobody can promise you will be in the first group. Size it as money you can lose, and write the exit before the entry.

Why do AI trading bots stop working?

The market changes, the rule was fit to old data, or the fees were never in the test. A bot that matched one year can lose the next without any bug. If the log shows lots of small round trips, cost is the first place to look.

Do AI trading apps work?

They work as tools when you can see fees, who holds the keys, and what the limits are. An app that only shows a profit screenshot is not a tool you can check. Read the fee, the custody and the off switch before you fund it.

Is a backtest proof that an AI strategy works?

No. A backtest is a rehearsal on old prices. It fails when fees are missing, when the rule saw the future bar, or when it was tuned until the chart looked good. One pretty curve is not evidence. Hold out data and add costs.