A study released in November 2026 reveals that approximately 50% of investment recommendations from ChatGPT and Gemini are concentrated in just 10 stocks. The research highlights that these AI models prioritize media coverage volume over financial fundamentals when selecting assets for portfolios.
Data from the National Bureau of Economic Research, covering models like ChatGPT 5.0/5.2 and Gemini 2.5 Flash throughout 2025 and 2026, shows a strong bias toward large-cap technology and semiconductor companies. Portfolios generated by these models held an average of 41% in semiconductor stocks, compared to roughly 20% in the S&P 500.
NVIDIA emerged as a primary focus, with ChatGPT allocating between 18% and 20% of portfolios to the chipmaker. Gemini displayed a different strategy, maintaining a median of only 4.5 stock holdings with a focus on momentum stocks and large firms. Both models frequently recommended the same assets, including Alphabet, Adobe, and Broadcom.
The research indicates that the stocks favored by AI models received nearly 10 times the media coverage of average firms. Analysts warn that this trend creates a risk of portfolio over-concentration, as investors relying on these tools may inadvertently build highly skewed positions in specific sectors or individual companies.
