Critics worry that AI-generated summaries of financial news articles may oversimplify or distort important information for investors, but a recent study suggests the summaries act like a trail of breadcrumbs, leading readers deeper into a story’s more complex points and creating a better foundation for understanding them.
Examining a year’s worth of articles in The Wall Street Journal, Harvard Business School Associate Professor Joseph Pacelli found that articles with AI summaries generated stronger and faster market reactions, suggesting investors processed the information more quickly. A follow-up experiment also showed that summaries improved readers’ understanding of both the information included in summaries and material that appeared only in the articles themselves.
The findings offer a counterpoint to concerns that AI introduces errors and bias—and that the technology encourages people to skim content superficially.
“I went into this thinking AI summaries might not be so useful, leading to cheap, quick reads,” admits Pacelli, the Gerald Schuster Associate Professor of Business Administration. “These articles are helping to grab your attention and inform you, which is great for markets.”
Pacelli coauthored the April working paper “Generative AI and Investor Processing of Financial Media” with Tony Cho and Allen Huang, professors at the Hong Kong University of Science and Technology, and Cornell University Professor Kristina Rennekamp.
Financial summaries boost market efficiency
The Wall Street Journal started attaching AI summaries to some stories in 2024, creating natural conditions for an experiment. The researchers identified 1,734 articles providing financial analyses of 158 companies between July 2024 and June 2025. They then examined an article’s effect on a company’s stock in the 30 minutes after the article appeared online and found that companies featured in articles with AI summaries saw a 3.5% increase in trading volume.
The researchers further probed the content of the articles, designing their own generative AI algorithm to investigate the tone of different paragraphs, both those included in the summary and those only in the body of the story. They found:
The market reacted to information highlighted in the AI summaries with stronger immediate trading volume, suggesting that investors used the summaries to make buying decisions.
Articles with summaries also seemed to lead to prices adjusting faster than those without.
“This suggests that traders are becoming informed more quickly, creating more efficient markets,” Pacelli says. “Our effects are larger when articles are more complex or there is a lot of competing news and earnings announcements, showing that these summaries are really helping people get through a busy workload.”
Like an ‘appetizer’ for the full course
To better understand how readers process the summaries, Pacelli and colleagues conducted a lab experiment: They gave 124 graduate school students two identical WSJ articles, one with an AI summary and the other without, then asked participants questions to test their recall of the articles.
The results of the experiment showed:
Readers with summaries remembered more of the overall article. They correctly answered 4.72 content questions out of 7. And they correctly answered 2.49 of 4 questions about information that appeared only in the article’s main body.
Meanwhile, readers of articles without summaries recalled less information. They got 3.95 of 7 questions overall correct, and 2.05 out of 4 questions correct about information not included in the summary.
What’s more, the summaries often spurred investors to dig deeper. Google search results for a company increased for articles with AI summaries, suggesting that people sought additional information from outside sources.
“We’re not seeing that summaries lead to less engagement. If anything, they seem to lead to greater engagement,” Pacelli says. “It’s as if they are an appetizer that makes you hungry to stay for the full meal.”
A note of caution
In general, generative AI summaries seem to work as intended, helping readers manage their cognitive workload as they sift through a daily bombardment of information to focus on what matters. But Pacelli cautions that it’s important to consider the context and the type of content AI is summarizing.
Readers are highly incentivized to seek out financial information for investments in companies where they are putting real money on the line, says Pacelli, but it’s not clear if the same would be true for news stories in which readers don’t have a direct interest, potentially leading to a more superficial analysis.
“It takes a lot more than three bullet points to truly understand inflation,” he says. “If things continue to move in this direction, readers may lose the ability to appreciate the richness of complex issues.”
How to use AI summaries
In terms of what media companies and other businesses should take away from the findings, Pacelli suggests that companies:
Consider using generative AI to summarize information they release to the public. “Everyone is using their own AI to summarize information,” he says. “If your company doesn’t do this, the risk is someone else is going to do it for you, and they may do it in a way that represents their own biases.”
Weigh using AI summaries for complicated material. That’s because AI summaries appear to be most valuable when audiences face information overload or particularly complex material.
Have people review AI summaries. Because AI-generated summaries can raise the risk of misinformation through bias or hallucinations, they require rigorous fact-checking and proofreading, Pacelli notes.
Illustration by Ariana Cohen-Halberstam.
Have feedback for us?
Generative AI and Investor Processing of Financial Media
Cho, Tony, Allen H. Huang, Joseph Pacelli, and Kristina Rennekamp. "Generative AI and Investor Processing of Financial Media." Harvard Business School Working Paper, No. 26-074, April 2026.
