In the days leading up to the U.S. presidential election, Latino voters are receiving a high volume of political messages in Spanish, including targeted ads and AI-generated misinformation. AI chatbots have been found to spread incorrect claims about voting rights in Spanish more frequently than in English, which raises concerns over the accuracy of election-related information available to this influential voter group.
Voting rights organizations worry that AI-driven misinformation could disadvantage Spanish-speaking voters by giving them unreliable information about voting procedures and rights.
An analysis by Proof News and Factchequeado found that over half of AI-generated responses in Spanish contained false information, higher than the rate for English responses. The study tested various AI models, including Meta’s Llama 3, which performed poorly in providing accurate election information in Spanish.
For instance, Llama 3 incorrectly stated that “federal only” voters are people in U.S. territories like Puerto Rico, who actually cannot vote in U.S. presidential elections—a critical error for understanding voter eligibility.
Political outreach to Latino voters is strong from both parties, as evidenced by events featuring high-profile figures like Vice President Kamala Harris in Las Vegas and former President Donald Trump in Pennsylvania.
Surge of Spanish-Language Misinformation Targets Latino Voters with AI-Generated Election Errors
Yet, the challenge remains that AI models used in widely accessed platforms like WhatsApp may misinform Spanish-speaking voters about important election issues. Meta and Anthropic have acknowledged the issue and have stated plans to improve the accuracy of their models, but Google has not responded to the report.
Advocates like Lydia Guzman from Chicanos Por La Causa urge voters to rely on multiple credible sources for election information instead of depending solely on AI, given the risk of inaccuracies in current AI-generated responses.
The prevalence of AI-based misinformation could be particularly impactful in states with large Hispanic populations, such as California and Arizona, where Latino voters represent a significant portion of the electorate.
Some voters, like Rommell Lopez from California, have encountered conflicting information from AI-driven sources, highlighting the limitations of technology in providing accurate election-related information. Lopez’s experience points to a broader caution among voters who, while finding technology helpful, recognize the importance of cross-referencing information to make informed decisions.
