We Gave 5 AI Models $1,000 Each to Trade Stocks — Here's Who Won
A creator handed Claude, ChatGPT, Perplexity, Grok, and Gemini $1,000 each in real Charles Schwab brokerage accounts and let each model pick its own trades for roughly six months. The spread between the winner and the loser was enormous — and it says a lot about how differently these models handle risk.
The Headline Numbers
Bottom line: combined, the five accounts grew from $5,000 to roughly $7,800 while the S&P 500 rose only about 5% over the same stretch. But that comparison is skewed by heavy use of leveraged ETFs — see the disclaimer below.
The Setup
The experiment, documented in this YouTube video, was simple: open five real Charles Schwab brokerage accounts, fund each with $1,000, and let a different AI model — Claude, ChatGPT, Perplexity, Grok, and Gemini — make all the buy and sell decisions for its own account over roughly six months. No human stock-picking, no hedging the AI's calls after the fact. Real money, real trades, real market volatility.
That framing makes this a useful, if informal, window into how each model actually reasons under uncertainty — not how it answers a hypothetical "what stock should I buy" prompt, but what it does when its own track record is on the line.
Final Results
| Model | Start | End | Return |
|---|---|---|---|
| Claude | $1,000 | ~$2,400 | +140% |
| ChatGPT | $1,000 | ~$1,739 | +70% |
| Perplexity | $1,000 | ~$1,731 | +70% |
| Grok | $1,000 | ~$1,451 | +45% |
| Gemini | $1,000 | ~$445 | -55% |
What Each Model's Strategy Reveals
Claude — Conviction, concentrated in the right themes
Claude's winning portfolio leaned on leveraged, thematic ETF bets — SOXL (3x semiconductors) and TQQQ (3x NASDAQ) — plus BOTZ, a robotics-focused ETF, and a well-timed Intel buy. That's a high-conviction, sector-concentrated approach that paid off because those themes ran hot during the test window. It's consistent with Claude's reputation for more deliberate, reasoning-heavy output rather than hedged, diversified answers.
ChatGPT and Perplexity — Similar, solidly above-market
ChatGPT and Perplexity landed within a few dollars of each other, both around +70%. Both comfortably beat the index without the extreme concentration Claude used or the blowup risk Gemini took on — a middle-of-the-road risk profile that still captured a strong AI/semiconductor market tailwind.
Grok — Positive, but the most modest gain
Grok finished up about 45%, positive but trailing the other three winners. Still a solid real-money return, just less aggressive positioning than Claude's theme concentration or as fortunate as ChatGPT/Perplexity's picks.
Gemini — The cautionary tale
Gemini crashed to roughly $445 after concentrating in a leveraged 2x Solana ETF that moved sharply against it. It's the clearest example in the experiment of a model taking on outsized, single-position leverage risk — the same instinct that helped Claude win (leveraged, thematic bets) backfired badly when the underlying asset didn't cooperate.
The behavioral pattern worth noting
Across the run, some models made increasingly risky "catch-up" trades after falling behind — chasing bigger, more leveraged positions to close the gap rather than settling into a more conservative posture. That's a familiar human trading bias (loss aversion turning into revenge trading), and seeing it show up in model behavior is one of the more interesting soft findings here, even though it wasn't formally measured.
Not financial advice. This is one creator's single-run, six-month experiment with $1,000 per account — not a controlled or repeatable benchmark. Results depend heavily on the specific market conditions during that window and each model's use of leveraged ETFs, which amplify both gains and losses. Past performance in this anecdote says nothing about future results, for these models or any others.
The Takeaway
This experiment isn't proof that any AI model can reliably beat the market — six months, five accounts, and $1,000 apiece is far too small a sample for that. But it is a genuinely useful, low-stakes window into risk appetite: Claude and Gemini both reached for leveraged, concentrated bets, and one paid off spectacularly while the other cratered. ChatGPT and Perplexity took a steadier middle path and still beat the index comfortably.
If you're evaluating which model to trust for research, analysis, or decision support more broadly, that risk-behavior signal is worth more than the dollar figures themselves — see how Claude, ChatGPT, Gemini, Grok, and Perplexity stack up on the tasks you actually care about in our head-to-head comparisons below.
Frequently Asked Questions
Which AI model performed best at real-money stock trading?
Claude finished on top, turning a $1,000 Charles Schwab account into roughly $2,400 (about +140%) over roughly six months, according to a single creator-run experiment. Its portfolio leaned on leveraged thematic ETFs like SOXL (3x semiconductors) and a well-timed Intel position.
Which AI model performed worst?
Gemini finished last, dropping from $1,000 to roughly $445 (about -55%) after concentrating in a leveraged 2x Solana ETF that moved sharply against it.
Did the AI models beat the stock market?
Combined, the five $1,000 accounts grew to roughly $7,800 (from a $5,000 total) over the test period, while the S&P 500 rose only about 5% over the same window. But four of five accounts used leveraged ETFs, which amplify both gains and losses, so the comparison isn't apples-to-apples with a simple index fund.
Is this a reliable benchmark for AI investing ability?
No. This was one creator's single-run experiment with small dollar amounts, a roughly six-month window, and no controls for market conditions during that specific period. It's an interesting anecdote about each model's risk appetite and reasoning style, not a rigorous, repeatable benchmark. Nothing here is financial advice.
Compare These AI Models Head-to-Head
See how Claude, ChatGPT, Gemini, and Grok stack up for the tasks you actually use them for.
Affiliate disclosure: Some links on this page are affiliate links. If you sign up through them, AISO Tools may earn a commission at no extra cost to you. This never affects our rankings or reviews.
📬 Get the best new AI tools delivered weekly
One concise email with fresh launches, trending picks, and featured standouts.
Join thousands of professionals who discover the best AI tools every week. No spam — unsubscribe anytime.