In Oct/2024 I asked ChatGPT for a portfolio to endure a messy world (original post here). At that time I expected the Israel–Palestine conflict (on top of Russia–Ukraine) to be the main driver of financial uncertainty in the coming months.
Fifteen months later, the conflicts are still there, but the volatility came from a different menu: U.S. surprises, Trump’s tariff war, a U.S. strike on an Iranian nuclear facility, and most recently Maduro’s capture.
This post is my promised check-in: what would have happened if I’d followed ChatGPT’s financial advice? I re-ran the original constraints using a 2024 model snapshot, translated the output into a real ETF portfolio, and measured the results.
The portfolio setup
The original post suggested a broad allocation. To make it testable, I re-ran a prompt in the old GPT-4o model (via OpenAI’s API, specifically gpt-4o-2024-08-06 to avoid lookahead bias) to generate a concrete, investable portfolio, using the exact constraints from that post.
Here’s the generated response:
This gave me a realistic, investable snapshot of “what ChatGPT would have done” at the time.
The results (Oct 4, 2024 → Jan 27, 2026)
A couple of days ago I pulled the adjusted prices (dividends, splits, capital gains) from Yahoo Finance (e.g., GLD history) and calculated the ROI for each sleeve.
| Asset | Weight | 2024-10-04 | 2026-01-27 | ROI |
|---|---|---|---|---|
| SPDR S&P 500 ETF Trust (SPY) | 25% | 564.43 | 695.49 | 23.22% |
| iShares MSCI ACWI ex U.S. ETF (ACWX) | 15% | 54.59 | 71.97 | 31.84% |
| Vanguard FTSE Emerging Markets ETF (VWO) | 10% | 46.56 | 57.64 | 23.80% |
| iShares 7-10 Year Treasury Bond ETF (IEF) | 15% | 92.01 | 95.98 | 4.31% |
| iShares iBoxx $ Investment Grade Corporate Bond ETF (LQD) | 15% | 105.48 | 110.81 | 5.05% |
| SPDR Gold Trust (GLD) | 10% | 245.00 | 476.10 | 94.33% |
| United States Oil Fund (USO) | 5% | 76.35 | 75.66 | -0.90% |
| Vanguard Real Estate ETF (VNQ) | 5% | 90.63 | 90.39 | -0.26% |
Portfolio result: +23.74% over 15 months. Not bad.
Closing thoughts
Net result: the portfolio held up. Most of it came from a surprisingly good gold run, with stocks doing their job and everything else mostly hovering. That’s encouraging, since diversification did its job, and also a warning, because one lucky ingredient can dominate the outcome.
So: would you trust an AI to build your portfolio, or are you already using it (even if just as a second opinion)?