UTA professor examines how AI outages affect financial market behavior

For students, a ChatGPT outage might mean losing a study aid they’ve come to rely on. For financial traders, it can mean losing an essential tool that has changed the way they work.

Dr. David Rakowski, professor of finance at the University of Texas at Arlington College of Business, recently published research that examines how trading activity increases during artificial intelligence outages, and how these outages affect investors' trading decisions. 

With so much information now readily available, professionals have turned to AI as a tool to organize information to make business decisions. AI helps sort, organize and summarize huge amounts of information from news, social media, YouTube and other sources that help them make more informed decisions.

“My understanding is that its not really the case that investors are using AI to pick stocks for them and make investment choices,” Rakowski said. “Our suspicion is that investors are using AI to sort, organize and summarize the massive amount of information that they’re getting now.”

Alongside co-author Mahanaz Paydazarnaghi, the researchers examined 207 documented OpenAI services outages between April 2021 and August 2025 and their effects on trading.  Rakowski said they initially thought that when AI goes down there would be less trade, but their findings indicated the opposite. 

“We expected that when AI goes down, you'd have less trading. We thought if investors are using AI and it stops working, investors are just going to sit still and do nothing,” Rakowski said. “They actually traded more, but it's so much messier now because they don't have the information they need. They're trading more, but it's having less of an impact.”

Dr. David Rakowski headshot.

Rakowski said studying AI failures allowed him and his co-author to examine how people change their behavior when AI goes down and how those changes affect markets.

The researchers ultimately found that financial markets behave differently during outages, with more trading occurring, but less efficient movement in prices.

“The two main ways that the market behaves differently is that there's actually more trading, but the second thing is that the trading is a lot messier or less efficient,” Rakowski said. “Normally when people trade prices move, but when AI is down, people trade more, but prices move less.”

The research focused on ChatGPT because it offered more data to study because of its status as one of the first major large language model (LLM) to hit the mainstream. Although there are other AI tools available when ChatGPT fails, Rakowski explained that switching between LLMs is not as seamless as one might think.

“If you have an account with one and it goes down. You don’t immediately want to set up an account with another one, because you won’t have your saved search history or whatever it is that you need,” he said. “Another problem is when one LLM goes down, the other ones go down too. They’re not totally independent of each other.”

Rakowski said that although there’s positives as the industry becomes more reliant on AI, there are also downsides. 

“It’s both good and bad. You're able to do more work more efficiently. But when it goes down, you're going to be basically back where you started,” Rakowski said. “But as the industry becomes more dependent on AI, you also need to have some type of resilience out there, so that when AI does go down, you have either a plan or alternative technology, that you can still get the job done.”

Although this specific article focuses on OpenAI outages, Rakowski’s broader research isn’t actually all about AI. Instead, he studies the sources investors use to receive and process information, and what happens when these sources are unavailable.

“My long-time work has been about information, how investors get information, how they use information and how shocks to the information change the way investors behave.” Rakowski said. “So, in that sense, ChatGPT and AI are just one new type of information tool that's come out.”

The idea of information disruptions affecting investors is not new. Rakowski said similar research has looked at what happened to financial markets when newspapers went on strike, cutting investors off from a traditional source of information.

“It's an evolution of how people use information to make choices, and with that, every time there's new technology, it's mostly the same, but a little bit different.” he said. 

Looking ahead, Rakowski said his research will continue to examine emerging technologies and their effects on the relationship between information and investing. He also plans to explore how these technologies affect different types of investors, the platforms and tools they use, and the growing influence of bots online.