Online discussions surrounding key Senate races in Texas and Kansas ahead of the midterm elections are significantly influenced by automated social media accounts, according to recent research from Gudea, an analytics firm specializing in digital conversations. The analysis of nearly one million posts reveals that much of the discourse is not organic but driven by bot activity.
The study examined 918,427 posts about Texas and 33,059 about Kansas, predominantly on the platform X (formerly Twitter), with some activity across 26 other social media sites. Researchers found that only about 25% of conversations related to the Kansas Senate race appeared to come from accounts focused on the state, likely representing genuine local users. The bulk of the discussion originated from accounts that engaged with both the Texas and Kansas races, many of which were newly created and exhibited characteristics typical of automated bots. Interestingly, these dual-topic accounts posted on Texas at a rate four times higher than accounts dedicated solely to Texas, an indication of coordinated amplification rather than independent commentary.
In the Texas race, which features Democrat James Talarico and Republican Ken Paxton, more than 85% of posts were reshared content, suggesting that a small group of accounts was amplifying messages disproportionately. Keith Presley, CEO of Gudea, emphasized the impact this has on genuine voters: “Real people are walking into a conversation where the table was already set. The topics and framing were in place before most users showed up, a good amount of it put there by automated accounts, and an ordinary person scrolling through has no way to tell the difference.” He cautioned that manufactured volume is often mistaken for public consensus.
The investigation also uncovered a coordinated network of hundreds of new accounts created within a single week that disseminated a false claim against Mr. Talarico. This allegation surfaced on October 20 and was reshared 12 times within 23 seconds. The built-in AI assistant on X, Grok, flagged the claim as uncorroborated and traced its origin to an account established in Romania the previous month. In the following hours, 265 newly minted accounts joined the discussion, sharing the claim nearly a thousand times collectively. Most of these accounts posted only this content before going silent, a pattern aligned with known bot operations.
Gudea’s findings are consistent with other research attributing similar tactics to Storm-1516, an influence operation linked to Russian efforts to destabilize confidence in the U.S. elections through covert online campaigns. Presley noted that the infrastructure identified is not specific to Texas or Kansas and can be deployed in various competitive races nationwide, indicating a broader strategy of manipulation in political discourse.
These developments underscore increasing challenges in verifying the authenticity of social media content related to political campaigns, highlighting the sophisticated ways in which automated networks can shape public perception and voter sentiment ahead of critical elections.