Registered investment advisors (RIAs) describe artificial intelligence (AI) in their fiduciary brochures more often as a risk than as a capability, according to a peer-reviewed study of 388 Form ADV Part 2A brochures published Oct. 14, 2026, in Frontiers in Artificial Intelligence.
Some 23.7% of brochures disclose AI use (22.7% survey-weighted), while 27.6% contain AI-related risk language.
The study also carries a warning for anyone quoting a single number. Depending on how AI disclosure is defined, which advisors are counted and how results are weighted, the same set of brochures supports estimates anywhere from 9.9% to 31.7%. With AI-related risk now on the Securities and Exchange Commission's (SEC) 2027 exam list, that is a compliance problem too. For advisors, the authors' practical answer is a firmwide AI inventory that says plainly where each tool is used.
Co-authors Samir Chincholikar and Robin Chawla, independent researchers in New York City who work in the finance industry, drew a stratified random sample of 400 SEC-registered advisors from a population of 16,223 firms active at year-end 2024, split by advisor type, private-fund versus wealth and retail, and assets under management (AUM) quartile. Usable brochures were available for 388.
A large language model (GPT-4o) sorted each brochure under five prespecified labels: AI in the investment process, AI in operations or client service, AI as a disclosed risk, explicit non-use and named AI vendors. The authors treated the model as a measurement instrument to be tested, not as ground truth, checking it against a blinded recoding by Anthropic's Claude and human review.
"Having just the word artificial intelligence in the brochure does not mean that the firm is using artificial intelligence," Chincholikar said. "It can be for a specific purpose, like artificial intelligence as a risk for investments."
Against human coding, the composite AI-use measure scored precision of 0.76 and sensitivity of 0.81 – meaning roughly three in four brochures it flagged were confirmed by human coders, and it caught about four in five that humans identified. A literal keyword search missed or mislabeled roughly a quarter of cases.
Within individual brochures, 77 carry both risk language and disclosed use, 30 carry risk language only, 15 disclose use only and 266 mention neither. The 3.9-percentage-point gap between risk language and use is statistically significant (McNemar p = 0.036, meaning it is unlikely to be a sampling fluke), and widens to 5.2 points when vendor-only references are excluded.
"AI was disclosed as a risk more times than as a use, which is kind of interesting," Chincholikar said. The authors note the pattern fits a document built to describe material risks, not to promote.
Disclosure rises steeply with AUM, from 10.4% of brochures in the smallest quartile to 43.8% in the largest. Among private-fund advisors it climbs from 12% to 56%; among wealth and retail advisors, from 8.7% to 30.4%. Private-fund advisors disclose more even after controlling for size.
Chincholikar pointed to compliance capacity. Large advisors "have like very large compliance teams," he said, with "more capacity to basically understand their own processes." Chawla cautioned that the data cannot show smaller firms are hiding anything. "We don't have any statistical proof saying... small investment advisors are not using AI but not reporting it," he said.
The authors cite Charles Schwab's 2026 RIA and AI Research Study, which found 63% of 533 Schwab-custodied RIAs surveyed in October 2025 using AI in some capacity. They call the roughly 39-point gap with brochure disclosure "a benchmark divergence, not a nondisclosure rate"; much internal or experimental use never needs to appear in a brochure.
Websites tell a similar cautionary tale. Among 283 firms observed in both places, 23.0% of brochures but only 5.7% of websites disclosed AI use. Yet brochures carry 7.3 times as much text, and the gap disappears once document length is accounted for, so the authors draw no conclusion about candor.
In March 2024, the SEC charged Delphia and Global Predictions with overstating their AI capabilities, including in Form ADV brochures, according to the SEC. The authors turned language from those orders into a screen. It flagged five of the 388 brochures (1.3%): two were false matches, three contained real AI references qualified by context or risk language, and none was judged an unsubstantiated promotional claim.
"It was case of an overselling," Chincholikar said of the 2024 cases. Now, "the question really is if we are telling people exactly where we are using AI." The SEC's fiscal 2027 examination priorities, released Oct. 23, 2026, add AI-related cyber risk – and, following a September 2026 executive order, refer to the technology as "super intelligence" throughout.
Raymond James lures a $125M Merrill Lynch advisor in Charlotte while a Stifel pair breaks away to go independent in Southern Oregon
Pave Finance, DPL Financial Partners and Gottfried & Somberg also add senior leaders as RIAs and wealthtech firms deepen their benches
What the feud between “twin” index funds teaches thematic investors
Fake advisors, fake platforms, fake crypto - all on a six-month timer
A new DOL rule restores a clearer two-factor test, giving independent advisors greater certainty over their business status.
Advisors are broadening the conversation beyond savings to help clients prepare for the realities of longer lives
Growing wealth, evolving priorities, and untapped opportunities are reshaping how women approach long-term investing.