AI Forex Trading – ChatGPT EA FREE Download

ChatGPT EA is an AI-branded grid trading robot for MetaTrader 4 (MT4) and MetaTrader 5 (MT5) that trades six forex pairs from a single chart: NZDUSD, USDCAD, AUDNZD, AUDCAD, NZDCAD, and GBPCHF. It runs on the M5 timeframe and layers a neural-network trade filter on top of a classic grid engine. The draw is its public MQL5 live signal, which has been running since 2019 and is one of the longest verified track records in this category.

This 2026 review covers what the EA actually does under the hood, what its live signal shows after roughly seven years, and a real-tick MT5 backtest we ran ourselves on the default settings. The download at the bottom is free and now includes both the MT4 and MT5 builds.

Free MT4/MT5 grid robot trading six forex pairs on M5 with an AI entry filter. The live signal is genuinely long (~7 years, 339% growth, 65.7% wins) and every pair is net positive. The catch is the engine: it adds losing positions on a grid with escalating lot sizes and no hard stop on the basket. Our own MT5 backtest matched the live win rate and profit factor, but showed a 25.6% equity drawdown behind a much calmer 7.9% balance line. Long record, real grid risk.

What is ChatGPT EA?

ChatGPT EA is a free MetaTrader 4 and MT5 expert advisor that trades six forex pairs (NZDUSD, USDCAD, AUDNZD, AUDCAD, NZDCAD, GBPCHF) from one M5 chart. It opens a position, then adds orders on a grid if price moves against it, closing the whole basket at a combined take-profit. A neural-network filter decides which signals to take, and news and spread filters block bad conditions.

The name points at large language models, and the EA does expose ChatGPT and Bard forecast options in its inputs. In the default configuration those two toggles ship turned off. The AI that is actually working out of the box is the neural-network entry filter, which we cover below.

Core Strategy and Logic: How the ChatGPT EA Works

Reading the default input set from our own MT5 test, the strategy breaks down like this:

• Grid, not single-shot – The EA opens a first trade in the chosen direction, then adds more orders every 25 points (the TradeDistance setting, adjusted by a SmartDistance filter) if price runs against it. The basket is closed together at a weighted take-profit.
• Escalating lot sizes – Each new grid level uses a larger lot than the last, following the multiplier ladder 1, 2, 4, 8, 14, 24, 41, 69. This is a martingale-style progression (increasing size to lower the basket’s average price), which is the real engine behind both the high win rate and the drawdown risk.
• Capped basket depth – MaximumTrades defaults to 9, so the ladder stops at nine levels per symbol rather than running unlimited. That bounds the worst case, but the top levels carry heavy size.
• No hard stop on the basket – By default there is no fixed stop loss on the grid. Risk is managed by the trade filter, a spread cap, an optional drawdown cutoff, and the news filter rather than a per-trade stop.
• Neural-network entry filter – EnableNeuralNetworkFilter is on by default with a 90% probability-of-profit threshold. This is the AI element that decides whether a fresh grid is allowed to open.
• ChatGPT and Bard forecasts off by default – The UseChatGPT_Forecast and UseBard_Forecast inputs exist but are disabled in the default set, so the branding is more about the machine-learning filter than a live language-model call.

This is the same broad family as the Multi-Currency Grid Robot, which grids across 17 pairs, but that EA uses uniform-size orders while this one escalates. The trend-aligned grid in the Gold Grid Strategy EA is a useful contrast too: it also holds a basket with no hard stop, but keeps lot sizes uniform rather than doubling up.

Please test in a demo account first for at least a week. Also, please Make yourself familiar and understand how this AI Forex Trading Robot works, then only use it in a real account.

Recommendations for this ChatGPT EA

• Minimum account balance of $1,000 with 1:100 leverage. With 1:30 leverage (EU brokers), $6,000+ is safer for anything above low-risk settings.
• Attach to one M5 chart only. AUDCAD is recommended. If your broker uses a suffix (e.g. AUDCAD.a), update the Symbol parameter.
• Use MT5 for testing. MT5 can backtest all six pairs from one chart with real tick data, which MT4’s single-symbol tester cannot do.
• Run it on a VPS continuously so the news filter and time detection keep working (Reliable and Trusted FOREX VPS – FXVM).
• The EA is NOT sensitive to spread and slippage. But We advise using a good ECN broker (Find the Perfect Broker For You Here)
• Size positions with the Lot Size Calculator before going live, since the grid’s escalating lots can build exposure fast.

FXVM Recommended VPS · From $19.99/mo Get a VPS Deriv Binary Trading Open Account

Recommended Brokers Compare All

XM Editor's Pick

Best for low-deposit beginners

$5 min deposit1:1000 leverageMicro accountsMT4/MT5

Open Account

Plexytrade Allow USA Traders

Best for high-leverage trading

$50 min deposit1:2000 leverageSmall accountsMT4/MT5

Open Account

Prop Trading Firms Compare All

Moneta Funded Editor's Pick

Instant funding · 1-step evaluation

Instant Funding1 Step EvaluationMT5 SupportUSA Traders

FLASH20 20% off challenge fee Join Now

Open Tools –> Options –> Expert Advisors. Check the option “Allow WebRequests for listed URL” Add the following: “https://www.worldtimeserver.com“, “http://timesrv.online“, “http://valeryservice.com“, “http://valerytools.com” and press “OK.” The EA takes news from the specified websites.

Performance Review: Live Signal and Our Own Backtest

This EA is worth reviewing because it gives us two independent things to check against each other: a long public MQL5 live signal, and a clean backtest we can run ourselves. The live signal proves the strategy has traded real money for years. Our backtest is the part the sales page never shows, a controlled test that tells us how the engine behaves and whether the live numbers are believable.

Before the numbers, the limits. A backtest is not a copy of the live account. The live signal runs on a $5,000 deposit with balance-scaled lots on a 1:300 broker, took roughly $8,000 in withdrawals over the years, and almost certainly changed its risk inputs along the way. Its exact settings are not public. So we did not try to clone the live curve. We ran the EA on its default risk with a fixed 0.01 lot from 1 January 2025, on 99% real-tick data, to see the strategy’s behaviour in a clean, fixed-size test. Treat the backtest as a reference for how it trades, not a prediction of the live equity line.

Live Signal Headline Numbers

1. Growth and the Declining Trend

The account turned a $5,000 deposit into $13,998.75 of net profit with no additional deposits, while taking about $8,000 in withdrawals along the way. That is a real, long, verified result. The more useful detail is the yearly column: +43.99% in 2020, +30.1% in 2021, +28.93% in 2022, then +14.33% (2023), +17.35% (2024), and +9.54% (2025). The trend is down. That can mean the operator dialed risk back, or that the edge is thinning as more capital chases the same setups. Either way, a new user should plan around the recent single-digit years, not the 2020 peak. Model your own path with the Compounding Calculator.

2. Win Rate and Profit Factor Line Up With Our Test

Here is where the two sources validate each other. We ran the default settings on Tickmill demo data (MT5, 99% real ticks) over roughly 17.5 months from January 2025, on a $1,000 account with a fixed 0.01 lot across all six pairs.

The win rate (68.1% vs 65.7%), profit factor (1.67 vs 1.51), and average hold time (16.8h vs 19h) all land close. That agreement is the most reassuring thing in this review. It means the live signal is doing what the strategy does in a clean test, not leaning on something hidden. The one gap is direction: our 2025-2026 window skewed long (66%), while the live account over seven years is balanced (51/49). That is a window effect, not a design difference.

3. Drawdown: The Smooth Line Hides the Real Risk

The live balance line looks calm, and the reported max drawdown is a controlled 17.9%. But our own test exposes the gap that matters for any grid. On the same run, the balance drawdown was only 7.95%, while the equity drawdown reached 25.64%. In plain terms, the closed-trade curve stayed smooth while open grid baskets floated more than three times deeper before they resolved green. That floating loss is the actual risk you carry, and it is invisible if you only watch the balance figure.

The backtest also flags the grid’s dependency directly. Its Z-Score was -2.84 at 99.55% confidence, which means trades are not independent, they cluster. That is the fingerprint of a recovery basket, and it is why a bad run hits several positions at once rather than one at a time. Check your own tolerance with the Drawdown Calculator.

Our fixed-lot backtest balance curve below is the clean reference the numbers above are built from.

4. Trade Timing and Frequency

The backtest logged 863 trades (1,726 deals) over the window, about 11 a week, close to the live signal’s 15. Entries cluster on Tuesday and Wednesday and spread across the trading day with peaks in the London and New York hours. The monthly view is telling: entries were heavy through the first half of 2025 and then thinned out noticeably, which lines up with the neural-network filter allowing fewer new grids as conditions changed, and with the live account’s cooling returns over the same period.

5. Pair Breakdown

Across the live account, all six pairs are net positive in dollars. GBPCHF leads with about $3,400 of profit, followed by NZDUSD ($2,800), AUDCAD ($2,200), NZDCAD ($2,100), USDCAD ($2,000), and AUDNZD ($1,500). A quirk worth knowing: NZDCAD and AUDNZD are actually negative in raw pips yet positive in dollars, because the grid’s escalating lots make the larger winning closes outweigh many small pip losses. That is the strategy working as designed, and also a reminder of where the size builds up.

ChatGPT EA

Free AI grid EA for MT4 and MT5 trading six forex pairs from one M5 chart. A ~7-year MQL5 live signal shows 339% growth, 65.7% win rate, and 17.9% max drawdown, verified against our own real-tick backtest.

Live Track Record

★★★★★

Strategy & AI Filter

★★★☆☆

Risk Management

★★★★☆

Pros

• ~7-year verified MQL5 live signal, all six pairs net positive
• Live win rate (65.7%) and profit factor (1.51) confirmed by our backtest (68.1% / 1.67)
• One M5 chart trades all six pairs; news and spread filters built in
• One-chart setup trading 6 pairs simultaneously

Cons

• Grid with escalating lot multipliers (1, 2, 4, 8, 14, 24, 41, 69) and no hard stop on the basket
• Equity drawdown (25.6% in our test) runs far deeper than the smooth balance line suggests
• Live returns are declining year over year

Summary

4.0

★★★★☆

Very few grid EAs carry a seven-year public live record, and this one does, with every pair net positive and drawdown held under 18%. The neural-network filter is a real component, not just a label. The reason it does not score higher is the recovery mechanic: escalating lot sizes on a grid with no hard basket stop, plus a live return trend that has cooled from 44% in 2020 to under 10% in 2025.

MT4 MT5 Is this working?

FREE DOWNLOAD 16,600+ downloads

Last tested by @Silent in Jul 2026 for ForexCracked Editorial.

Forum(Mirror)

How It Compares

The takeaway: ChatGPT EA has the longest verified record of the group by a wide margin, but its profit factor sits with the other multi-pair grids, well below the concentrated gold grid. Length of proof is its edge, not per-trade efficiency.

Frequently Asked Questions

Conclusion

ChatGPT EA earns its place mainly on longevity: a seven-year public live signal with every pair net positive is rare, and our own backtest confirms the win rate and profit factor rather than contradicting them. However, this is a grid that recovers losses with escalating lot sizes and no hard stop on the basket, and the real risk shows up in equity drawdown (25.6% in our test) that the smooth balance line hides. With returns cooling year over year, it suits a trader who understands grid mechanics, funds the account properly, runs it on a VPS, and treats the recent single-digit years as the realistic expectation.