AI Copy Trading: The Complete Guide

Copy trading is often presented as the simple option — connect an account, mirror a better trader, collect the difference. The mechanism genuinely is simple. What is not simple is everything around it: how your position size is derived from someone else's, what happens when execution lags, how providers are selected and de-selected, and who carries the loss when the two accounts diverge. This guide covers those mechanics in detail.

It assumes you already understand roughly what automated trading is. If you do not, start with the companion guide to AI trading, which covers how the underlying decisions get made. This article is about the delivery layer: how a decision made on one account becomes a position on yours, and everything that can go wrong in between.

1. What AI copy trading is — and is not

Copy trading replicates the trading activity of a source account onto yours automatically. When the source opens a position, an equivalent position opens on your account, scaled to your capital and settings. When it closes, yours closes.

The AI component describes the source, not the copying. The copying mechanism itself is plumbing — it moves orders. What makes it "AI copy trading" is that the decisions being copied originate from a model-driven system rather than a discretionary human trader.

This distinction is worth holding firmly, because the two halves fail in completely different ways and are usually marketed as one thing. A brilliant strategy delivered through poor copying infrastructure produces disappointing results. Flawless infrastructure copying a bad strategy produces losses with great efficiency. When you evaluate a service you are evaluating two products.

What it is not

• It is not a managed fund. Nobody takes custody of your money. Your capital stays in your own brokerage account throughout, which is the single most important structural protection in this arrangement.
• It is not passive income. The trading is automated; the oversight is not. Accounts left genuinely unattended are how people discover a drawdown three months late.
• It is not risk transfer. The provider takes none of your downside. You carry every loss in full, whatever their fee structure.
• It is not identical results. Your returns will differ from the source account's published figures, sometimes materially. Section four explains why.

2. The mechanics: how a trade reaches your account

Understanding this sequence tells you where the failure points are.

Step one: the signal is generated

The source system reaches a decision — instrument, direction, size, stop level, target. On the source account, the position opens.

Step two: the copier detects it

The copy platform notices the new position, either by polling the source account at intervals or by receiving a push notification from it. Polling introduces delay proportional to the interval. Push is faster but requires tighter integration. This is the first place milliseconds are lost, and a provider who cannot tell you which method they use is a provider who has not thought about it.

Step three: size is calculated for your account

The platform converts the source position into a size appropriate for your balance and settings. This calculation is where most of the divergence between accounts originates, and it is covered in detail in the next section.

Step four: the order is sent to your broker

Your broker receives and fills the order at whatever price is available at that instant — which is not the price the source received. The gap between the two is your slippage, and it is a permanent structural cost of copying rather than a bug.

Step five: the position is managed

Stops and targets are placed on your account. When the source closes, a close instruction is sent to yours. Crucially, ask what happens if the connection drops while a position is open: does your stop-loss sit at the broker, where it will execute regardless, or does the platform manage it remotely? Only the first is safe.

3. Allocation models and why they matter more than they look

This is the most consequential setting in copy trading and the one most people accept by default without understanding it.

Fixed lot

Every copied trade opens at a size you specify — always 0.10 lots, for instance. Simple and predictable, but it ignores both your balance and the source's conviction. A 0.10-lot position is a trivial risk on a $50,000 account and a dangerous one on $1,000. It also flattens the source's own sizing decisions, discarding information.

Proportional to balance

Your position is scaled by the ratio of your balance to the source's. If the source risks 2% of its account, you risk 2% of yours. This is generally the correct default: it preserves the source's risk decisions and scales automatically as your account grows or shrinks.

Fixed multiplier

You copy at a set multiple of the source's size — 0.5×, 2× and so on. Useful for deliberately running more or less aggressively than the source, but it requires you to understand what the source's baseline risk actually is. A 2× multiplier on a system already running high risk is how accounts end.

Risk-based

You define a maximum risk per trade — say 1% of equity — and the platform sizes each position so that the stop-loss distance equals that amount. The most sophisticated approach and the safest, because it accounts for the fact that a 20-pip stop and a 200-pip stop represent completely different risks at the same lot size. It requires the source to publish stop levels reliably.

The practical recommendation: proportional or risk-based for most people. Fixed lot is only appropriate if you have calculated what that lot represents as a percentage of your account at typical stop distances — and most people never do that calculation.

4. Latency, slippage and why your results differ

New copiers are frequently surprised that their account does not match the published figures. The divergence is expected, and understanding its four sources lets you judge whether what you are seeing is normal or a problem.

Execution delay

Between the source opening and your order filling, time passes. On a good platform this is tens of milliseconds; on a poor one, seconds. In a fast-moving instrument like gold around a data release, a two-second delay can mean a materially worse entry — occasionally the difference between a winning and losing trade.

Spread differences

The source account might trade with a broker offering 0.2-pip spreads on a raw account. If yours quotes 1.5 pips, you pay 1.3 pips more per trade in each direction. Across hundreds of trades that difference alone can consume the strategy's entire edge, and it has nothing to do with the strategy itself.

Rounding

If proportional sizing calculates 0.037 lots and your broker's minimum increment is 0.01, the order rounds. On small accounts this rounding is significant and systematically distorts your risk relative to the source — usually upward.

Swap and commission

Overnight financing rates vary between brokers and account types. A strategy holding positions for days can see meaningfully different net results purely from swap.

The practical consequence: expect your results to track the source's direction but not its magnitude. If you are within a reasonable band of the published figures over a few months, the copying is working. If you are dramatically worse, investigate spread and latency before concluding the strategy has failed.

5. Copy trading models compared

Model	How it works	Main trade-off
MAM / PAMM	Broker-level pooled allocation across sub-accounts	Efficient, but broker-dependent and less transparent per trade
Bridge copier	Third-party software linking source and follower terminals	Broker-agnostic and flexible; adds a dependency and some latency
Cloud copier	Hosted service connecting via broker API	No VPS, works from a browser; you trust the platform's uptime
Signal service	Alerts sent to you; execution is manual or semi-automated	Maximum control, but you reintroduce the human delay and emotion

Most modern retail platforms, including this one, use the cloud model. It removes the VPS requirement and the associated cost and maintenance, at the price of depending on the platform remaining online — which makes their uptime record a fair question to ask directly.

6. How to choose a provider

Beyond the general system-evaluation criteria in the AI trading guide, these checks are specific to copying.

1. Is the source account real and verifiable? Ask whether the published account is live capital or a demo. Demo results carry no slippage and no emotional or financial stakes.
2. What is the average copy latency? A provider who measures this will tell you. One who has never measured it is telling you something too.
3. Which brokers are supported? If you are forced onto a single broker, examine that broker's spreads carefully — the arrangement may be where the real revenue sits.
4. Can you set your own risk parameters? Maximum lot, maximum concurrent positions, daily loss limit, instrument exclusions. A platform that offers no follower-side controls hands you all the risk and none of the steering.
5. What happens on disconnection? The safe answer: stops live at your broker and no new positions open.
6. Can you close positions manually without breaking the connection? You should always be able to intervene on your own account.
7. How is the strategy allowed to change? A source that quietly alters its risk profile is a different product from the one you subscribed to. Ask whether material changes are announced.
8. What is the historical maximum drawdown of the source, and when? As always, the number that matters most.

7. Risk controls you should insist on

The controls that live on your side of the connection matter as much as those inside the source system, because they are the ones you can actually set.

• Maximum position size. A hard ceiling regardless of what the source does. This protects you against a malfunction or a strategy change at the source.
• Daily and weekly loss limits. Automatic disconnection after a defined loss. The single most valuable follower-side control.
• Maximum concurrent positions. Prevents the account being fully committed during a period when the source opens many trades at once.
• Equity stop. Halt everything if account equity falls below a floor you set in advance, while you are calm.
• Instrument filters. The ability to decline instruments you do not want exposure to.
• News-window pause. Optional, but valuable if you have seen how wide spreads become around major releases.

Set these before you start, not after the first bad week. Limits chosen during a drawdown are chosen by a different person than the one who opened the account.

8. Account security and credential handling

Copy trading requires giving a platform access to your trading account. This is normal, and it is also where the most serious avoidable mistakes happen.

Understand the three password types

• Investor password — read-only. Cannot trade, cannot withdraw. Safe to share, but insufficient for copying.
• Trading password — can open and close positions. Cannot withdraw funds. This is what a copy service needs.
• Master password — full account control, potentially including withdrawal and settings changes. Never required for copying.

If a platform asks for withdrawal rights, or for your broker portal login rather than your terminal credentials, stop immediately. There is no legitimate copying architecture that requires either.

Practical precautions

• Use a dedicated trading account rather than one holding your wider savings.
• Enable two-factor authentication on your broker portal, which is separate from your terminal credentials.
• Know how to change your trading password — it is the fastest way to revoke access instantly if you need to.
• Check whether the platform stores credentials encrypted and whether staff can view them.

9. Fee structures decoded

Copy trading fees are structured in several ways, and the label matters less than the incentive it creates.

• Fixed subscription. Predictable. The provider earns the same whether you profit or not, which means their incentive is retention rather than performance — acceptable, but understand it.
• Performance fee. Typically 10–30% of profit. Better aligned, provided there is a high-water mark so you are not charged twice for recovering the same losses. Confirm this explicitly; its absence is a serious warning.
• Spread markup. The provider receives part of the spread through a broker arrangement. This is the least visible structure and creates an incentive toward trade frequency, since revenue rises with volume regardless of profitability.
• Per-lot commission. A fixed amount per lot traded. Same frequency incentive as above, but at least it is stated openly.

The question worth asking directly: does the provider earn more if the system trades more? If yes, that is a structural conflict of interest. It does not make the service illegitimate, but it tells you which direction the pressure runs.

A worked example: how a source position becomes yours

Numbers make this clearer than any amount of theory. Take the following case:

• Source account balance: $50,000
• Your account balance: $5,000
• The source opens a gold position of 1.00 lot with a stop 300 points away

Proportional to balance

Your ratio to the source is 5,000 ÷ 50,000 = 0.1. Your size is therefore 1.00 × 0.1 = 0.10 lots.

On gold, the point value at 0.10 lots is roughly one dollar. Your risk is 300 × 1 = $300, or 6% of your account — the identical percentage the source risked ($3,000 of $50,000). The proportional model preserved the risk decision exactly.

Fixed lot at 0.10

Here the numbers happen to match. But suppose the source later opens only 0.20 lots because its confidence is lower: fixed sizing opens 0.10 lots for you again, and you now carry double the relative risk the source intended. This is precisely where fixed sizing discards information.

Risk-based at 1%

Your target risk is 1% × 5,000 = $50. With a 300-point stop you need a point value of 50 ÷ 300 ≈ $0.167, which is roughly 0.017 lots.

And here the rounding problem appears. If your broker's minimum increment is 0.01 lots, the order rounds up to 0.02 — an actual risk of about $60 rather than $50, 20% more than you intended. On a $5,000 account that is tolerable. On a $1,000 account the distortion becomes large enough to genuinely change your risk profile.

The practical lesson: always calculate what your settings mean in currency at the typical stop distance, not in lots. A lot size is a meaningless number until you multiply it by point value and stop distance.

10. What to monitor once it is running

Copy trading is not passive, but it also does not require daily attention. A weekly review of the right five metrics is sufficient, and far more useful than watching every trade.

Drawdown from peak

The most important number. Track your own account's drawdown, not the source's published figure — yours will differ because of the divergence factors in section four. Compare it against the limit you set before starting.

Divergence from source

Compare your return against the source's over the same window. A consistent gap is normal; a widening gap indicates an execution problem worth investigating — usually spread or latency.

Trade frequency

A sudden change in how often the system trades often signals a strategy change or a malfunction. Either warrants a question to the provider.

Average position size

If sizes creep upward without you changing settings, something is wrong. This is an early warning of martingale behaviour that was not disclosed.

Concurrent exposure

How much of your account is committed at the busiest moment in the week. This is your true risk exposure, and it is usually higher than people assume from looking at individual trades.

The live dashboard on this site publishes exactly these figures for a real account — balance, equity, drawdown, open exposure and the reasoning behind each decision, updated every sixty seconds. Whatever provider you use, this is the level of visibility worth expecting.

11. When to stop copying

Deciding this in advance, in writing, is the difference between a considered exit and a panicked one. Reasonable triggers:

• Your predefined drawdown limit is hit. Not "approached and it feels like it might recover" — hit. This is the whole reason you set it.
• The strategy's behaviour changes materially without explanation: position sizes, frequency, or instruments traded.
• Your divergence from the source widens persistently, which means the arrangement is not delivering what it advertises regardless of strategy quality.
• The provider becomes unresponsive or opaque — particularly during a losing period, which is precisely when communication matters.
• Your own circumstances change. Capital you can no longer afford to risk should not be risked, regardless of how the system is performing.

Note what is not on that list: a few losing weeks. Every strategy has them, and exiting into a normal drawdown is the most common way people convert a temporary loss into a permanent one.

12. Mistakes specific to copy trading

1. Copying multiple providers on one account. They cannot see each other. Combined exposure can exceed what either would permit alone, and margin can be exhausted by a combination neither anticipated.
2. Choosing by return alone. The highest-returning provider over six months is disproportionately likely to be the one taking the most risk. Sort by return relative to drawdown.
3. Using a fixed lot without doing the arithmetic. Work out what your chosen lot represents as a percentage of your account at the source's typical stop distance. People are frequently shocked by the answer.
4. Ignoring the broker. Half of your divergence from published results comes from spread and execution. The broker is part of the strategy whether you treat it that way or not.
5. Assuming stops are guaranteed. In a gap, your stop executes at the next available price, which can be considerably worse. Size for that possibility.
6. Not testing the disconnect procedure. Practise stopping the copying before you need it urgently.
7. Treating published results as your expected results. They are the source's results, achieved on the source's broker, at the source's spread.

Copy trading on prop-firm accounts

A large share of retail copy trading now happens on funded accounts from proprietary trading firms rather than on personal capital. The mechanics are identical, but the rules layered on top change the calculation substantially — and this is where a great many funded accounts are lost.

Daily loss limits are the binding constraint

Most prop firms impose a maximum daily loss, often 4–5% of the account, breach of which fails the account immediately and permanently. A strategy that is entirely survivable on personal capital — where a 6% down day is unpleasant but recoverable — is fatal here.

Critically, many firms measure the daily loss on equity including floating positions, not on closed balance. A position that dips 6% intraday before recovering to profit can still breach the rule. Copy settings that ignore this will eventually fail an account that was, on any normal measure, profitable.

Size down, not up

The instinct on a funded account is to maximise return, since the capital is not yours. The correct instinct is the opposite: reduce size relative to what you would run personally, because the failure condition is far tighter. A drawdown that merely hurts on a personal account ends a funded one.

Check the rules the source cannot see

The source system has no knowledge of your prop firm's constraints. It does not know about your daily limit, your maximum position size, prohibited instruments, or news-trading restrictions. Some firms prohibit holding positions over the weekend or through high-impact releases entirely — and a copier will happily breach all of these on your behalf.

This is precisely why follower-side risk controls matter more here than anywhere else. Your daily loss limit, maximum concurrent positions and instrument filters are not optional refinements on a funded account; they are the only thing standing between the source's strategy and your firm's rule book.

Consistency rules

Some firms require that no single day contributes an outsized share of total profit. An automated system that produces one exceptional day can trip this rule even while being profitable overall. Read the specific rule set before connecting anything, because these clauses are rarely prominent.

13. Frequently asked questions

Is copy trading safe?

The structure is relatively safe: your funds remain in your own account and cannot be withdrawn by the provider. The trading carries full market risk and you bear all of it. Those are separate questions and conflating them is how people misjudge the arrangement.

How much money do I need to start?

Enough that proportional sizing produces positions above your broker's minimum increment without rounding distorting your risk. Below roughly $500–1,000 on most brokers, rounding effects become significant enough to meaningfully change your risk profile relative to the source.

Can I lose more than I deposit?

On a retail account with negative-balance protection, generally no. Without that protection, an extreme gap on a leveraged position can theoretically leave a negative balance. Confirm your broker's policy — it is a one-line answer and worth knowing.

Can I still trade manually on the same account?

Technically usually yes; practically it is a poor idea. Your manual positions consume the same margin, and you will no longer be able to attribute results to either approach.

Why are my results worse than the advertised performance?

Most commonly spread, then latency, then rounding, then swap. Compare your broker's spread on the traded instruments against the source's. That single comparison explains the majority of cases.

What happens if the provider stops operating?

Your account remains yours and your open positions retain their broker-side stops. New copying ceases. This is a good reason to prefer platforms that place stops at the broker rather than managing them remotely.

Does the provider see my balance?

Typically yes — it is required for proportional sizing. They cannot move funds. If you are uncomfortable with visibility, a dedicated account holding only your trading capital resolves it.

Is AI copy trading better than copying a human?

Different, not strictly better. A system offers consistency, continuous coverage and auditable logic. A skilled discretionary trader may adapt to genuinely novel conditions in ways a model trained on history cannot. The relevant question is not which category is superior but whether the specific source has a verifiable record and controlled risk.

How long before I can judge whether it is working?

Longer than feels comfortable. Three months is a minimum for any signal at all, and six or more is better. Statistically, short samples are dominated by noise, and decisions made on them are close to random.

Where to go next

To understand how the decisions being copied are actually produced, read the companion pillar on AI trading. For the practical steps of connecting an account, see the MetaTrader 4 and MetaTrader 5 (MT4 and MT5) connection guide. If you are new to the vocabulary, start with the beginner's guide, and for a condensed view of the trade-offs, read the benefits and limits.