How generative AI is reshaping the quantity and quality of expert analysis
Listen to this article:
We have entered an era of information abundance and trust scarcity. Thanks to AI, it’s easier than ever to generate clean, coherent, professional-sounding content, but some tend to assume that it isn’t worth their attention, confidence, or actual decision-making calculus This problem sits at the center of the new working paper “Generative AI Use by Capital Market Information Intermediaries: Evidence from Seeking Alpha,” co-written by Harvard Business School AI Institute associate Yuan Zou. They found a real-world setting where they could measure, with precision, what happens when AI-generated content floods an information ecosystem, and then gets abruptly removed from it. The results are more nuanced than either AI optimists or skeptics tend to allow.
Key Insight: A Natural Experiment during the Bot-Writing Boom
“We begin by examining the prevalence of AI use by [Seeking Alpha] contributors over time.” [1]
Seeking Alpha (SA) is a massive crowdsourced investor community platform, with roughly 20 million active users. SA saw a rapid rise in AI-assisted writing after ChatGPT’s launch in 2022, from about 1.6% to its peak at 13.5% of articles, and then a steep decline after the company equated AI use with plagiarism and enforced a no-AI rule in 2023. The sudden rise and fall provided the perfect laboratory to test how generative AI actually changes the behavior of information intermediaries and the reaction of the markets they serve. Using advanced AI detection tools like Originality.ai and GPTZero, the researchers tracked a dramatic lifecycle of AI adoption across approximately 76,000 articles. Authors who adopted AI tools published 33.1% more articles than non-adopters, a trend that reversed sharply once SA restricted AI use.
Key Insight: The Dilution of Insight
“We find that AI articles are less informative than human articles, eliciting smaller capital market responses and lower receptivity by SA users and editors.” [2]
If AI makes it easier to write, does it make the writing any better? Compared to human-written articles, AI-assisted articles triggered smaller trading volume responses, weaker stock price movements, fewer reader comments on the platform, and were less likely to be selected as Editor’s Picks. To sharpen the causal claim, the researchers examined trading activity in the 10-minute windows immediately before and after each article’s publication. Human articles drove a 15.6% increase in intraday trade volume, but only 8.0% for AI articles. Investors seemed to sense a lack of depth in AI articles, but the researchers add an important qualification. AI-assisted articles written by contributors with at least 18 months of prior platform experience generated significantly stronger market reactions than those from newcomers who joined SA in the months immediately after ChatGPT launched. This suggests that knowledge and expertise wrapped around a tool can be an important differentiator.
Key Insight: Lower Quality, Better Ecosystem
“[W]hile AI articles are currently perceived as less informative than human articles, their low production cost enables greater firm coverage.” [3]
Here’s where the AI content problem resolves. Even though individual AI articles were perceived as worse, the aggregate effect of AI adoption was a meaningful improvement in the information environment. AI-enabled productivity gains caused analysts to cover more firms, particularly smaller, previously neglected companies that hadn’t attracted research attention. During the peak of AI use, the share of firms receiving analyst coverage rose from roughly 40% to 47%. This expanded coverage had a tangible, positive impact on the market’s health, increasing overall liquidity by 5% to 6% and accelerating the speed at which news was priced into stocks. In other words, having AI generate an average report could be better for the market than having no information at all.
Why This Matters
For business leaders and executives, this study is a masterclass in the strategic trade-offs of the AI era. It suggests that AI might not add value by replacing high-level human expertise, but can do so by expanding the boundaries of what is economically feasible to monitor and analyze. The decision to deploy AI doesn’t need to be about just doing the same things faster, but about doing things that were previously too expensive to do at all. A winning AI strategy is a hybrid approach: protect the human-led core of your firm while leveraging AI to cast a wider net, ensuring that no corner of your industry remains in the dark.
Bonus
This research shows that AI can function as a “spotlight” that guides what information we pay attention to and how we evaluate it. For a different look at how the dimensions of novelty and feasibility can interact with that dynamic, check out How AI Can Spot Your Next Billion-Dollar Idea.
References
[1] Bradshaw, Mark T, Chenyang Ma, Benjamin Yost, and Yuan Zou, “Generative AI Use by Capital Market Information Intermediaries: Evidence from Seeking Alpha,” Journal of Accounting Research (JAR), forthcoming (March 24, 2026): 2. https://dx.doi.org/10.2139/ssrn.5226562
[2] Bradshaw et al., “Generative AI Use by Capital Market Information Intermediaries,” 36.
[3] Bradshaw et al., “Generative AI Use by Capital Market Information Intermediaries,” 5.
Meet the Authors
Mark T Bradshaw is the Joseph F. Cotter Professor and Chair of the Accounting Department at the Boston College Carroll School of Management.
Chenyang Ma is a PhD student in accounting at the Duke University Fuqua School of Business.
Benjamin Yost is Associate Professor and Clark Family Fellow for the Boston College Carroll School of Management.
Yuan Zou is an assistant professor in the Accounting and Management unit at Harvard Business School and an affiliate at the HBS AI Institute.
Watch a video version of the Insight Article here.