GMGN Copy Trading: How It Works and Why Most People Lose Money

The pitch for copy trading is the cleanest idea in the whole terminal. Find a wallet that prints, mirror it, collect. The plumbing works. GMGN builds it, documents it well, and it fires on time more often than not.

The results are the problem. Here is the warning GMGN puts at the top of its own copy trade documentation, before any instructions:

a lot of users, although their copy trades profit, but the gas priority fee is too high, resulting in the end of the still lose money

That sentence is doing a lot of work. The company selling you the feature is telling you that the trades can win and the account can still shrink. This guide covers the mechanics properly, then the arithmetic that produces that outcome and how to stack it in your favor.

What Copy Trading on GMGN Actually Is

You point a task at one Solana wallet. When that wallet buys, your wallet buys. When it sells, your wallet sells a proportional slice of what the task bought for you, or hits your own take-profit and stop-loss levels instead, depending on how you configured it.

The constraints, straight from GMGN's copy trade documentation:

• Solana only. Not BSC, not Base, not Ethereum, despite what most third-party guides say.
• Telegram login required. A browser extension wallet session cannot run copy tasks, along with every other automation feature. Our how to trade on GMGN guide covers the login split.
• Ten tasks maximum.
• Cost: "One copytrade transaction = buying/selling amount + Gas priority fee + 1% GMGN handling fee, no other factors."

That last line is worth pinning. There is no copy trading surcharge. The 1% is the same 1% a manual trade pays, which we break down against the rest of the sector in GMGN fees explained.

Setting Up a Task

The web terminal at https://gmgn.ai and the Telegram bot create the same task. On the web you start from a wallet page, the Copy Trade rank, or the Wallet Radar, which scans up to 10 tokens at once and surfaces the addresses that bought them earliest, bought most, or profited most, with a one-click Copy Trade button on each result.

Buy methods

Fixed buy spends the same amount every time. A whale putting in 40 SOL and a test buy of 0.2 SOL both trigger the same order from you. Maximum follow-buy tracks the target's size proportionally until it hits your ceiling, then caps out.

Fixed buy is the safer default for most people, because it removes the scenario where a target's single oversized entry drains your balance into one illiquid token.

Sell methods

Four options, and they behave differently in ways the interface does not make obvious:

• Automatic follow-sell. When the target sells 10% of its position, the task sells 10% of yours. It only touches tokens the copy task bought, so your manual bags and other tasks' positions are untouched.
• No follow-sell. The task buys and then stops. You exit by hand.
• Take-profit and stop-loss. Monitored against the average price of the position the copy task built. Important detail: once TP/SL is on, the task stops following the target's sells. You have swapped the target's exit judgement for your own percentages.
• Batched TP/SL. Multiple rules, each selling a percentage of the copy position. Each rule fires once unless you add to the position, and adding recalculates against the new average price.

The Filters Are the Product

The advanced settings are where a copy task becomes something other than a random number generator. Every one of these is applied to the target's buy, not yours:

Filter	What it does
Market cap min / max	Only copies when the token's market cap sits inside your band at the time the target buys
Liquidity min / max	Same idea against pool size. The minimum is your main defense against unexitable positions
Token age min / max	Measured from first creation of the token, not from when a particular pool opened
Minimum pool burnt ratio	Skips tokens whose liquidity burn percentage is below your threshold
Minimum copy amount	Ignores the target's small probe buys. Set to 0.5 SOL and a 0.3 SOL test from the target is skipped
Maximum copy amount	Ignores oversized buys. GMGN frames this as protection against a target pumping a token it is about to rug
Single coin position increase times	Caps how many times you re-buy the same token while holding it. Set to 0 and you buy once per position
Launchpad platform filter	Restricts copying to tokens issued on the platforms you select. Affects buys only, never sells
Blacklist	Up to 20 tokens that will not be bought or sold by the task

The minimum copy amount filter is the one that pays for itself. Wallets on the smart money leaderboards routinely fire tiny probe buys into dozens of tokens to see what sticks. Mirroring those means paying a full round trip of fixed costs on positions the target never intended to hold.

Lightning Mode, and Why It Is a Trade

Lightning Mode makes your copy about two seconds faster by acting on unconfirmed on-chain messages rather than waiting for confirmation. GMGN documents the cost of that honestly:

• The chain can roll back, so your trade can execute when the target's never did.
• In extreme cases the transaction can execute more than once.
• The mode can observe a sell before the buy, which breaks auto-sell on that position.

Two seconds is genuinely a lot when you are copying a wallet that flips in under a minute. It is worth nothing at all when you are copying a wallet that holds for hours, and the rollback risk stays either way.

Why the Target Wins and You Do Not

Now the part that decides whether any of the above matters.

You are always the later buyer

A copy task cannot beat the wallet it copies. Detection, filter evaluation, transaction construction and confirmation all sit between the target's fill and yours, and on a token minutes old the price is moving inside that window. You buy higher than the target on the way up. When the target sells, the same lag puts you behind it on the way out, into a book that the target's own sell just made thinner.

That is structural. No setting removes it. Lightning Mode shortens it by around two seconds and adds a different risk in exchange. The consequence is that a target's marginal winners become your marginal losers, and you keep only the trades that ran far enough for the lag not to matter.

The fee stack, charged per mirrored trade

Here is the arithmetic nobody runs before funding a task. GMGN recommends a total on-chain fee of 0.006 SOL or higher for auto buy, auto sell and limit orders, which is what a copy task is (source). That is a fixed cost per transaction, not a percentage, and a round trip is two transactions.

Copy buy size	GMGN 1% both ways	On-chain fee, 0.006 SOL per leg	Round-trip cost	Break-even move needed
0.1 SOL	0.002 SOL	0.012 SOL	0.014 SOL	about 14%
0.5 SOL	0.010 SOL	0.012 SOL	0.022 SOL	about 4.4%
1 SOL	0.020 SOL	0.012 SOL	0.032 SOL	about 3.2%
2 SOL	0.040 SOL	0.012 SOL	0.052 SOL	about 2.6%

Figures assume a flat exit, so the 1% sell fee is calculated on the same value as the buy. Slippage is on top and is not small on tokens young enough to be worth copying, where GMGN itself recommends 50% or more.

Two things fall out of that table. First, small copy sizes are brutally inefficient: at 0.1 SOL you need a 14% move just to get back to flat, and most memecoin round trips do not clear 14%. Second, the cost is per trade, so frequency multiplies it. Twenty mirrored round trips a day at 0.006 SOL a leg is 0.24 SOL a day in on-chain fees alone, roughly 7 SOL a month, before GMGN's 1% and before a single losing trade.

GMGN's suggested fix in the same document is to run auto slippage, Anti-MEV off, and a priority fee of 0.002 to 0.006 SOL if speed is not critical. Turning Anti-MEV off is a real trade, since it needs a priority fee of at least 0.002 SOL to work at all and it is what stands between you and sandwich bots. Cheap and exposed, or protected and expensive: copy trading makes you pick, on every single mirrored trade.

Trades that never fire, and tasks that quietly stop

Copy tasks miss. GMGN documents the failure ladder plainly:

• A failed buy or sell is retried 1 to 3 times depending on the error type, and then abandoned. GMGN's instruction at that point is to place the trade manually, which on a token moving fast means the opportunity is gone.
• After three consecutive failures the task auto-pauses and sends a message. It will not restart itself. You have to restart it by hand.
• A copy that never fired at all is usually a filter, not a bug. GMGN's first troubleshooting step is to check whether advanced filters are on.
• Lightning Mode's out-of-order sell detection can leave a position with no working auto-sell.

The web copy detail page shows successful copies, failed copies and their reasons, and it is worth reading after every session rather than trusting the summary. Check task state at the same time. A paused task looks identical to a task that simply had nothing to copy. Codes and fixes for the underlying transaction failures are in GMGN transaction failed.

What the PnL number is not telling you

Read the displayed profit skeptically. GMGN's wallet detail page defines its fields as total profit and loss, realized and unrealized profit, without stating anywhere in the documentation that the 1% handling fee, priority fees or tips are netted out of them. The only number that settles the question is your SOL balance. Compare the wallet balance at the start and end of a week against what the copy dashboard claims you made, and trust the balance.

Open the GMGN Telegram Bot

Sizing So the Math Can Work

Nothing above says copy trading cannot work. It says the fixed costs have to be small relative to the position, and the frequency has to be low enough that they do not compound. Practically:

Reading a Target Wallet Before You Copy It

The DISCOVER leaderboards rank wallets by 1 day, 7 day and 30 day PnL, 7 day win rate, 7 day transaction count and 7 day average cost. The wallet detail page adds average holding duration, unrealized profit, cost distribution, and a set of phishing checks that flag wallets which bought blacklisted or honeypot tokens, hold tokens they never bought, sold more than they bought, or trade in under 10 seconds, which is the fingerprint of a bot rather than a trader.

What to weight:

• Average holding duration. The single most useful field for copy trading, because it tells you how much your latency will cost. Sub-minute holds mean you are competing on speed and losing.
• Transaction count against PnL. Big PnL from few trades survives being copied. Big PnL from hundreds of trades is usually an edge in execution speed, and that edge does not transfer.
• Win rate against average cost. A 30% win rate on large sizes is a strategy that needs the position sizing to match, and fixed buy at 0.2 SOL does not reproduce it.
• The phishing and bot flags. A wallet flagged for 10 second round trips is not something a copy task can follow.

What This Is Realistically For

Copy trading on GMGN is a decent tool for exposure to a strategy you have researched, at a size where the per-trade overhead is a rounding error, with filters that skip the noise. It is a poor tool for turning a small balance into a big one by attaching yourself to a leaderboard wallet, because at small sizes the fixed costs are larger than the average edge, and the terminal charges them whether the trade wins or not.

The custody question sits underneath all of this too. A copy task runs from a GMGN-generated wallet whose private key you cannot export, so the balance funding it is a hot trading account, not savings. That tradeoff is unpacked in is GMGN safe and who holds your keys, and the deposit and withdrawal mechanics are in connect Telegram and fund your GMGN wallet.

If you want the discovery side without the automated execution, the trenches and sniper guide covers manual entries from the same data. For how GMGN's copy trading and fee structure compare to the alternatives, see GMGN vs Photon, and what is GMGN has the background on how the platform got here.

Nothing here is financial advice. Copying a profitable wallet is not a strategy, it is a bet that someone else's strategy survives your latency and your fee bill.