RESEARCH
APRIL 2026
↓ PDF
LIVE · 1,000+ TRADERS
RESEARCH
Most traders fail alone — losing money, failing evaluations, or never reaching payout. This evidence review examines what public data shows, and where AI support is most likely to change outcomes.
A trader must pass the evaluation, avoid drawdown breaches, maintain consistency rules, and stay profitable long enough to receive payouts. Public data shows this funnel is extremely narrow.
| Funnel Stage | Why Traders Fail Here | Where AI Support Can Help |
|---|---|---|
| Evaluation | Over-sizing, impulsive re-entry, weak stop discipline, inconsistent days | Real-time rule tracking, position-size calculators, pre-trade checklist |
| Funded Survival | Best-day concentration, drawdown spikes, slow adaptation after losses | Daily risk alerts, variance monitoring, behavioral prompts |
| Payout Conversion | Profit concentration, overtrading after a good day, consistency violations | Session planning, post-trade review, consistency monitoring |
| Long-term Retention | Strategy drift, fatigue, weak journaling, failure to learn from mistakes | Pattern detection, journal summaries, recurring coaching loops |
Across leveraged retail products, academic studies, and public prop-firm disclosures, the dominant pattern is consistent: most traders do not survive long enough, consistently enough, or profitably enough.
| Evidence Source | Population | Headline Finding | Why It Matters |
|---|---|---|---|
| ESMA Retail CFD Warning | Leveraged retail CFD accounts, EU | 74–89% lose money | Baseline difficulty before applying prop-firm rules |
| Brazil Equity Futures Study Chague et al., 2020 |
Individuals trading ≥300 days | 97% lost money; only 0.4% earned more than a bank teller | Persistence alone does not rescue outcomes |
| Taiwan Stock Market Study Barber et al., 2004 |
Individual day traders | Day traders as a group lost money; activity was >20% of volume | Heavy participation ≠ profitability |
| Taiwan Futures Market Study Kuo et al., 2020 |
Day traders in futures market | Most individual day traders lose money | Futures access does not remove the profitability challenge |
No large public dataset yet cleanly isolates prop traders by AI usage. However, adjacent evidence from financial decision-making research is now meaningful and directionally consistent.
Customers who received human-AI collaborative investment advice were more likely to align their final decisions with advice received. Measured uplift +15.5 percentage points overall, +21.3 pp for riskier investments, and an average +44.92% increase in final payoffs across the sample.
In a study of 5,179 customer support agents, access to an AI assistant increased productivity by 14% on average and by 34% for novice and lower-skilled workers — consistent with the strongest expected gains for newer traders.
Drawing on AIProp's proprietary dataset from 1,000+ active prop traders on the platform. Figures reflect survey responses and platform-derived user signals.
| Dimension | Self-Directed | AI-Supported |
|---|---|---|
| Research process | Manual, slower, prone to narrative bias | Faster synthesis, standardised plan quality |
| Rule compliance | Depends on memory and willpower under stress | Real-time alerts and objective hard limits |
| Learning speed | Slow — journaling is inconsistent or incomplete | AI summarises errors, detects recurring leaks |
| Decision under uncertainty | Exposed to fear, greed, recency bias, tilt | Human still decides; AI can reduce noise |
| First-month loss rate | 65% lost money | 15% lost money |
| Account volatility | Higher observed | Lower observed among AI/EA users |
| Survey interest in AI/EAs | — | 78% of surveyed traders expressed interest |
The strongest case for AI support is not "better predictions." It is tighter process control, faster feedback loops, and reduced behavioral mistakes — the frictions that actually destroy prop-trader success.
Get the full PDF of Prop Traders: Self-Directed vs AI-Supported — including the complete evidence review, public funnel analysis, AI collaboration findings, and recommended dataset framework.