Experts Warn of AI Disinformation Targeting Immigrant Voters

With the Nov. 3 midterm elections approaching, misleading claims about voter eligibility, immigration enforcement, and election fraud are circulating across social media, raising concerns that confusion could discourage eligible naturalized citizens in immigrant communities from voting, according to a report by Neha Gopal for The Contra Costa Pulse.

Speaking at an Oct. 23 American Community Media briefing, researchers warned that AI-generated images, misleading political videos, and inaccurate chatbot responses are complicating efforts to separate fact from fiction.

“People are scared to vote. People are scared to participate,” said Cameron Hickey, CEO of the National Conference on Citizenship, noting that researchers have documented multiple AI-generated images circulating online that falsely depict immigration agents intimidating voters at polling places.

628 Disinformation Posts Documented Across Chinese-Language Platforms

Jinxia Niu, senior digital engagement manager at Chinese for Affirmative Action, said false claims about noncitizen voting and voter identification requirements exploit immigrants’ unfamiliarity with U.S. election rules. While noncitizens cannot vote in federal or California state elections, naturalized U.S. citizens who meet eligibility requirements can.

In a 2024 report, Chinese for Affirmative Action documented 628 pieces of disinformation across Chinese-language social media over roughly 13 months. Niu said false narratives increasingly originate on X before spreading to WeChat and YouTube, losing context as they cross platforms and languages amid the decline of traditional ethnic media outlets.

Multilingual AI Chatbot Errors and Undisclosed AI Political Ads

Researchers also cautioned that AI chatbots do not provide equally accurate answers across languages. Laura Zommer, co-founder and CEO of Factchequeado, cited an Institute for Strategic Dialogue study finding that 45% of Spanish-language chatbot responses to election questions were incorrect or incomplete, compared with 30% of English-language responses. In one instance, Meta’s chatbot twice gave the wrong date for the 2026 midterm elections when asked in Spanish, answering Nov. 4 instead of Nov. 3.

Hickey shared preliminary polling showing that the share of eligible voters using AI chatbots to learn about candidates rose from 14% to 25% between two weekly surveys. Meanwhile, Zommer cited a Wesleyan Media Project report identifying at least 164 political ads using AI—accounting for nearly $80 million in spending—of which 69% did not disclose their use of artificial intelligence.

To help voters verify claims, Hickey recommended the four-step SIFT method—Stop before sharing content that provokes fear or anger, Investigate who published it, Find better coverage from reliable news organizations or official sources, and Trace claims back to their original source—alongside reverse image searches and trusted community messengers.