ChatGPT vs. MSCI World After 1 Year: Outperformance With Errors

TVS News AI, editorially supervised, presents ChatGPT vs. MSCI World After 1 Year: Outperformance With Errors.

The hard truth: The excerpt does not prove the outperformance

The Spiegel excerpt is dated 21.10.2026 and explicitly states: “Unfortunately, you can no longer access the article.” It cites two possible reasons: The link is more than 30 days old or the article has already been opened 10 times. The printout provided therefore consists of access notices and pricing information, not the results of the investment test.

The excerpt lacks the MSCI World return, the ChatGPT portfolio return, the exact observation period, the portfolio details and documentation of the errors. Neither “after one year” nor the statement that the result was better than expected is substantiated by a time period, an expectation or a metric. The claimed outperformance can therefore neither be confirmed nor refuted on the basis of the available source. This is not counterevidence against the test; it is, rather, a clear limit on what the excerpt at hand can establish.

The route to the original is clearly set out in the excerpt

Anyone who already has a digital subscription is directed to log in. According to the provider, print subscribers can add digital access free of charge. For new customers, the access page lists €1 for 4 weeks, followed by €6.49 per week. Alternatively, it offers 52 weeks at €4.49 per week, with a saving of 31 percent; after that, the cost is €6.49 per week.

Following the subscription, access includes all S+ articles on SPIEGEL.de and in the app, as well as the weekly digital edition with the e-paper, digital archive and S+ newsletter. The offer can be canceled at any time. For the substantive review, however, one point remains decisive: The visible excerpt itself contains neither the portfolio decisions nor the corresponding return figures.

Four pieces of information underpin a reliable assessment

For evaluating an AI portfolio, the excerpt lacks precisely the information needed to identify success and errors:

• Benchmark: What return did the MSCI World post over the same observation period?
• Portfolio: Which investments were chosen, how were they weighted, and at what points in time were they bought?
• Result: How high were the benchmark and portfolio returns, and how large was the difference between them?
• Errors: Which decisions were classified as wrong, and what effect did they have on the result?

The strongest counterargument is that the full Spiegel article could contain these details, while the excerpt provided shows only the paywall. That is precisely why the preliminary assessment remains so unequivocal. Without the full text, neither the specific outperformance nor its errors can be reliably assessed.

Without the Full Text, No Reliable Investment Decision

For you, this means: The comparison cannot be evaluated using this source. You can say that Spiegel offers a test of ChatGPT against the MSCI World and that the access page documents the article. But you cannot yet infer what return was achieved, whether the goal was met after one year, or what errors the model actually made.

Check the original: The Spiegel link is the direct route to the full text. If the link is more than 30 days old or has been opened 10 times, the page denies access, according to the excerpt; existing subscribers are instead directed to log in.