How much of a problem do fake reviews, generated by AI, pose to online consumers?
That’s what Audrey Tollett ’26, a computer science major from Houston, spent her summer finding out. In partnership with data science professor Tianxi Dong, Ph.D., Tollett worked on a project titled "The Impact of AI-Generated Content on Sales: Evidence from Steam Reviews." This research focused on online video game retailers, specifically on detecting non-human influence on the digital market.
For Tollett, who is a gamer (you can catch her playing Counterstrike 2 and Slime Rancher) and president of Trinity’s Gaming Club, this was the perfect intersection between her academic interests and personal passions.
“This (work) is important because consumer trust is a big deal,” she says. “If we can't trust that the reviews we are reading are actually human-generated, we’d have to rely on in-person word-of-mouth again.”
Tollett’s project delved into ChatGPT and other generative AI models that have impacted the online retail market. “Now, we can't easily be sure if the reviews on a product are human-generated,” she says. “So, I investigated how proportions of AI-generated reviews have changed over time, and how those figures correlate with the sales rankings of these games.”
And it was a perfect opportunity for her to use both her computer science major and data science minor. She programmed in Python through an open-source web application called Jupyter Notebook. “A lot of what I had to do was scrape through data. I made a little ‘crawler’ that goes onto a website and gets a big, horrible file of unclean data, and then I went through and cleaned, formatted, organized, and got it sorted so I could do analytics.”
Tollett tagged a series of online reviews using this program, then performed a panel analysis—a method similar to linear regression—to see how the proportion of large language model-generated reviews correlates with sales rank of online games.
Tollett was one of more than 200 Tigers doing research at Trinity this summer, thanks to a wealth of institutional resources and opportunities centered around hands-on learning at the University.
Her project was funded through a Semmes Grant (through Trinity’s D.R. Semmes School of Science). “It was awesome,” Tollett says, "because they gave us free housing, and the housing staff, the RAs, and the custodial staff, they were all holding it down. I also need to shout out the library staff because I was there all the time, making them help me check stuff out.”
Tollett also loved working with Dong, who’s been a steadying presence during a tumultuous time in the computer science industry. “She's really sweet, she's very patient with me,” Tollett says. “We met about once a week, and she was very receptive to emails. She's been a great mentor to me.”
With family members who went to bigger schools, Tollett says she’s grateful for Trinity’s small size. “Students at bigger schools don't really get to talk to their professor, since they’re talking to the TA most of the time. But here at Trinity, specifically in computer science, most of the professors have an open-door policy, and you can pretty much just walk into their office with your random problems.”
And, perhaps most importantly, Tollett also loves the vibe of her fellow computer science students.
“This is a fun group to be around. I participate in a lot of the clubs, the Association for Computing Machinery (ACM), which is the official computer science tutoring group, and then there's also TU Women in Computing,” she says. “I'm an officer in both of those groups, and it's a really nice community we've got there. The Trinity community—we really uplift each other. When our grads get jobs, they’re always like, ‘Now I can write you a letter of recommendation!’ So it's very tight-knit.”
As a rising senior, Tollett’s project has prompted an additional discovery: what she wants to do with her diploma. “Doing research is helping me develop my understanding of what the industry looks like. And it's helping me solidify my future goals,” she says.