Researchers say they found that China’s TikTok “systematically boosts” content using in favor of Socialist New York City mayoral candidate Zohran Mamdani and against former NYC Mayor Andrew Cuomo (D) using its content algorithm. The team asserts that this goes beyond simple bias into a new area, “algorithmic steering.”
“Our new research suggests TikTok’s algorithm systematically boosts pro-Mamdani and anti-Cuomo content in NYC’s mayoral race,” Yehonatan Dodeles, the CTO of Spring AI — an open-source app framework for artificial intelligence engineering — announced in a Tuesday X post. The research report has been published to Medium by Dodeles.
“This analysis was done by Spring AI,” Dodeles continued, adding that “the company uncovers the hidden mechanics of social media algorithms to defend democracies from weaponized AI.”
“Based on deep reverse-engineering of TikTok’s recommendation system and representative data (real & synthetic users), we built an AI model that measures when videos get more exposure than they should. We call these ‘Excessive Publicity’ videos,” Dodeles explained in a follow-up X post.
The software engineer went on to say that, across all videos, roughly 17 percent “show this kind of non-organic boost,” adding that his team’s model “correctly flagged” 76.4 percent after testing known paid ads, “confirming it works.”
“In short: it spots hidden promotion,” Dodeles said.
“Then we checked if any topics deviate from the norm,” Dodeles continued, noting, “Most don’t.”
“Even trending hashtags stay near the 17% baseline. That means our model distinguishes popularity from promotion,” the chief technology officer said. “But one topic exploded off the chart: Politics — 55% of political videos got excessive promotion. triple the normal rate.”
Inside that set, two names “dominated,” which were “Mamdani” and “Cuomo,” Dodeles said, disclosing that his team then built another AI model to separate “pro” and “anti” videos, “and manually validated its accuracy.”
The software engineer explained that when he and his team normalized within politics — which he noted were already 55 percent promoted — “pro-Mamdani” and “anti-Cuomo” videos were boosted “above” the baseline.
Meanwhile, “pro-Cuomo” content was “suppressed below” it,” Dodeles added, which he described as “algorithmic steering” rather than just simply “bias.”
“These are preliminary findings and further research should be done, but the implications are huge,” the Spring AI CTO said.
“The algorithm that shapes what we see may also shape what we believe,” Dodeles concluded in his X thread, which included a link to his team’s article for readers to review the full findings of the distressing discovery.
China’s TikTok app — widely popular among young Americans — is already known for its constant stream of short and over-stimulating videos that make for a powerfully addictive source of propaganda.
As Breitbart News senior contributor and New York Times best-selling author Peter Schweizer noted, “Colonel Dai Xu, a professor at China’s top military academy,” wrote in a restricted Chinese academic journal that the real battle between the U.S. and China is “information-driven mental warfare.”
Alana Mastrangelo is a reporter for Breitbart News. You can follow her on Facebook and X at @ARmastrangelo, and on Instagram.
