The mutual fund industry has become increasingly concentrated. The five largest fund families managed about 35% of industry assets in 2005, compared with 57% in 2024. One explanation is economic efficiency: Successful managers attract more capital, and larger organizations benefit from economies of scale. Another possibility is that large fund families enjoy an advantage because they are easier for investors to find and evaluate. Brand recognition, advertising, established distribution networks, and prominence in online searches can make investors more likely to consider well-known families.
Generative AI could change this advantage. Conventional internet search typically presents users with a ranked set of webpages and leaves them to decide which links to investigate and how to compare the information they find. Generative AI can instead take an investor’s stated objective—for example, finding a low-cost fund with particular risk and performance characteristics—and identify, summarize, and compare relevant alternatives across fund families. This technology may therefore make it easier for investors to consider funds from families they did not previously know. In a new paper, we study whether this change in search technology affects competition in the U.S. mutual fund industry. We use the introduction of SearchGPT in July 2024 as the beginning of a major shift toward AI-mediated online search and examine what happens to investor website activity and capital allocation across fund families.
Our first finding is that investor search becomes less concentrated than its previous trajectory would predict. Before SearchGPT, website traffic was becoming increasingly concentrated among a small number of large fund families. After SearchGPT, that trend flattens. By the end of our sample, the share of website traffic going to the five most-visited families is approximately 13.5% below the level predicted by continuation of the pre-SearchGPT trend. Other measures of concentration—including the HHI, Gini coefficient, and entropy—show a similar pattern.
The change is concentrated among large incumbent fund families. Following SearchGPT, the historical advantage of large families in attracting website traffic declines. A one-standard-deviation increase in pre-AI family size is associated with about a 20% relative decline in website traffic, while comparing families at the 75th and 25th percentiles of size implies approximately a 28% relative decline for the larger family.
Several additional findings point toward AI-mediated search as a factor rather than simply an unrelated change occurring at the same time. The decline is greater for large families whose website traffic had historically been more responsive to direct visits, conventional organic search, traditional referrals, and marketing. Moreover, when GPT services experience substantial outages, the decline in large families’ traffic advantage becomes substantially weaker. Greater AI-referral activity is also associated with relatively more subsequent website activity for smaller rather than larger families.
Importantly, the change extends beyond website visits to actual investment decisions. After SearchGPT, larger fund families lose relative shares of new fund sales and, more gradually, total net assets. A one-standard-deviation increase in family size is associated with an approximately 8.3% relative decline in new-sales share and a 4.2% relative decline in total-net-asset share. The progressively smaller effects—from website search, to new sales, to total assets—are consistent with a natural sequence: investor search can change quickly, new investment follows, while the existing stock of assets adjusts much more slowly (See Figure 1 below).
Panel A: Top-five share of new sales Panel B: Top-five share of total net assets
Figure 1: Concentration of fund sales and assets across fund families
This figure plots monthly measures of the concentration of fund sales and assets across fund families. Panels A and B report the shares of fund sales and total net assets accounted for by the five fund families with the largest corresponding shares in each month, respectively. The plotted observations are adjusted for the number of fund families, the market return, and the risk-free rate. The dashed fitted line represents the pre-SearchGPT trend, and the solid fitted line represents the post-SearchGPT trend. The vertical solid segment in July 2024 illustrates the estimated immediate level change associated with the introduction of SearchGPT, while the vertical dashed line marks the beginning of the post-SearchGPT period. The fitted lines are obtained from interrupted time-series regressions.
What This Means for Mutual Fund Managers
- Our findings suggest that generative AI may be changing how fund families compete for investor attention. Historically, scale brought an important informational advantage: A familiar brand, prominent web presence, and established distribution channels made a large family more likely to be considered by investors. Generative AI may weaken this advantage by allowing investors to begin with what they want rather than whom they already know.
- This does not imply that large fund families will lose their advantages or that generative AI will necessarily reduce industry concentration permanently. Instead, the evidence suggests that AI-mediated search can make product attributes more important relative to incumbent visibility in determining which funds receive investor consideration. For large fund families, maintaining visibility may increasingly require being well represented in AI-generated comparisons rather than relying primarily on brand recognition and conventional search. For smaller families, generative AI may create an opportunity to reach investors who previously might never have encountered their products.
- The broader message is that generative AI is not simply another source of website traffic. It may change the gateway through which investors discover investment products—and, in doing so, change how mutual fund families compete for investor attention and capital.
What It Means for Policymakers
- For policymakers, the findings suggest that the common-ownership debate should begin one step earlier than is typically recognized. Before asking what happens when a few asset managers own large stakes across corporate America, it is useful to ask why so much investor capital flows to those asset managers in the first place. Our evidence suggests that search frictions are part of that process—and that generative AI, by making smaller fund families easier to discover, may weaken one force contributing to the concentration of investor capital and, ultimately, corporate ownership.
Mengqiao Du is an assistant professor at the National University of Singapore, and Xiumin Martin is a professor at Washington University in Saint Louis’ Olin School of Business. This post is based on their recent paper, “Generative AI and Mutual Fund Industry Concentration,” available here.
Recent Comments