Who Gets the Last Word? Election Disinformation in Immigrant Communities
Pakistan Chronicle | By Tariq Khan
A voice note forwarded in a family WhatsApp group. A video claiming immigration agents are stationed outside a polling site. A chatbot that quietly gives voters the wrong election date. These are now the everyday channels through which false and misleading information reaches immigrant voters in the United States, according to a national news briefing hosted by American Community Media (ACoM) on Friday, 23 October 2026, titled Who Gets the Last Word? Election Disinformation in Immigrant Communities.
About the Briefing
With the November midterms days away, the virtual briefing brought together three specialists in mis- and disinformation research to walk ethnic media reporters through how false narratives about voting and immigration status travel through trusted community networks, how artificial intelligence is reshaping that landscape, and what journalists can do to respond. The session was moderated by Pilar Marrero, Associate Editor at ACoM, and opened by ACoM’s Sandy Close.
Understanding the Problem
The first speaker, Cameron Hickey, CEO of the National Conference on Citizenship, set out the basic vocabulary of the problem: misinformation that misleads regardless of intent, disinformation that is knowingly fabricated, conspiracy theories that attach to a larger boogeyman narrative, content stripped of its original context, and biased framing that manipulates emotion without being outright false. Hickey stressed that although the distinctions matter for researchers, the outcome for voters is the same in every case. People become frightened to vote, frightened to participate, and make decisions based on information that simply is not true.
He pointed to two forces that keep this content spreading: the attention economy, in which platforms profit from keeping users engaged with whatever provokes the strongest reaction, and basic human cognitive bias, our tendency to trust claims that already match what we want to believe. As an example, he described a recent rumour that a bombed Russian data centre had caused internet bot activity to collapse. The data centre strike was real, he said, but the bots in question had nothing to do with election interference, and the same volume of bot activity was recorded the very next day. The claim still spread rapidly because people wanted it to be true.
Slowing the Spread With SIFT
For reporters and community members trying to slow that spread, Hickey recommended a four-step method called SIFT, developed by information literacy researcher Mike Caulfield: Stop before reacting or sharing, Investigate the source, Find better or alternative coverage, and Trace the claim back to its original context. He cautioned that AI tools are not necessarily reliable arbiters of truth either, and in some cases carry their own risks of surfacing untrustworthy sources.
Threats Facing This Election Cycle
On the specific threats facing this election cycle, Hickey flagged two converging trends. Generative AI is being used to manufacture images of immigration enforcement agents intimidating voters at polling sites, images that are fake but are circulating widely on Instagram, YouTube and TikTok. At the same time, real ICE presence near polling locations is feeding genuine fear, and voter rights groups report that outreach to newly naturalised citizens is becoming harder as a result. He also raised concerns about the federal SAVE database, a citizenship verification system that the administration has used to flag tens of thousands of registered voters, including in California, as potential non-citizens, when many are in fact recently naturalised citizens whose records simply had not yet been updated.
AI Chatbots and Voter Information
New polling data NCoC plans to release shortly shows how fast this is moving. Two weeks ago, 14 percent of eligible voters nationally said they had used an AI chatbot to learn about candidates for the 2026 midterms. By last week that figure had risen to 25 percent, and among Hispanic voters specifically it climbed to roughly 30 percent, nearly double the rate recorded among non-Hispanic voters. Hickey urged reporters to publish clear, well-sourced articles answering the exact questions voters are likely to type into a chatbot, since these systems are already pulling from whatever obscure sites happen to rank, not from vetted journalism.
Questions About Voting Machines
Asked by reporter Nestor Fantini about complaints in Florida, Texas and Tennessee that touchscreen voting machines were switching voters’ selections, Hickey said the same claim has resurfaced in some form during each of the last three election cycles, and is more likely a recycled narrative than new evidence of tampering. He advised reporters to apply conventional verification, tracing images back to a first-party source and using reverse image search before treating viral footage as credible.
Chinese Language Disinformation
The second speaker, Jinxia Niu, programme manager for the Chinese Digital Engagement project at Chinese for Affirmative Action (CAA), a civil rights organisation founded in San Francisco in 1969, outlined three recurring themes her team has tracked across more than 600 pieces of Chinese-language disinformation. The first is election fraud narratives pushed by pro-Trump, anti-CCP Chinese-language influencers, who have amplified unverified claims of Chinese interference in the 2020 election and accused mainstream outlets of suppressing the story. Niu noted that the centre of gravity for this content has shifted in the past year from WeChat, which has tightened censorship of political content, to X, now the platform generating the most engagement before claims are translated back into WeChat groups.
The second theme is non-citizen voting disinformation, which Niu said exploits a genuine cultural gap: in China, identification is required for nearly every transaction, so claims that US elections do not require voter ID land as deeply alarming within immigrant communities already anxious about immigration enforcement. She cited a real registration software error in New Jersey that, once it reached Chinese-language WeChat and Xiaohongshu groups, triggered panic among naturalised citizens fearful that an accidental registration error could put their status at risk. A forthcoming CAA report, Divided by Design, examines how this same pattern of fear-driven, anti-immigrant disinformation is fracturing not only Chinese but also Indian, Filipino and Vietnamese communities across the country.
The third theme, partisan attacks framing a vote for Republicans around claims that Democrats support open borders and weaker law enforcement, is largely recycled from a nearly identical 2022 flyer, Niu said, with 2026 versions now generated using AI image tools. She added that Chinese-language queries on AI search tools disproportionately surface right-leaning outlets such as The Epoch Times, a systemic gap she said extends well beyond any single chatbot or platform. CAA has identified more than 150 disinformation-linked social media accounts in its dataset. Separately, participant Nancy Tran of Vietnamese Radio TNT noted in the session chat that a parallel dynamic plays out in Vietnamese-language media, where disinformation frequently frames Democrats as communists or socialists to exploit fear among refugees of communism, a pattern several participants said ethnic media outlets are well placed to debunk precisely because they are trusted community voices.
Countering Fear Driven Narratives
Asked how journalists can counter fear-based narratives, Niu advised starting with trusted messengers already embedded in a community, a local leader, a relative, or a familiar outlet, and gradually expanding that circle of trust outward rather than trying to correct misinformation from an unfamiliar source.
AI in Political Ads and Chatbot Answers
The final speaker, Laura Zommer, co-founder and CEO of Factchequeado, a non-partisan organisation working to counter Spanish-language disinformation for the more than 68 million Latinos living in the United States through a network of 151 media and community partners, focused on the accelerating role of AI in this election cycle. Citing Wesleyan Media Project research from early September, she said at least 169 political ads using AI tools had already been identified, representing roughly 80 million dollars in spending, and that 69 percent of those ads did not disclose to viewers that AI had been used. No federal law currently requires such disclosure, she noted, and while 31 states have passed their own rules, enforcement varies widely and leaves many voters unable to tell what they are watching.
Zommer also cited University of California San Diego research showing that 76 percent of Americans are concerned that AI-generated audio or video could affect the 2026 midterms. Perhaps most striking, Factchequeado partner the Institute for Strategic Dialogue tested six major AI chatbots with identical election questions in English and Spanish. In English, three of ten answers were incomplete, outdated or simply incorrect. In Spanish, that failure rate rose to 45 percent. In one documented case, Meta’s chatbot gave an incorrect midterm election date in Spanish, stating 4 November rather than the correct 3 November, on two separate occasions. Spanish-language answers also cited significantly fewer sources overall, she said.
Resources for Ethnic Media
During the question period, this reporter asked panelists where ethnic media outlets could reliably verify claims and source original reporting free of copyright restrictions, beyond established resources such as the Texas Tribune. Zommer noted that all Factchequeado content, including Spanish-English bilingual pieces, is free to republish or translate with credit. Hickey pointed reporters toward VERDAD (verdad.app), a free tool built to help journalists investigate mis- and disinformation circulating on Spanish-language radio, a format several participants noted carries particular weight because listeners often trust the speaker’s voice as inherently credible.
Closing Message
Closing the session, Niu urged ethnic media outlets to subscribe to CAA’s PiYaoBa Substack newsletter and to help share the forthcoming Divided by Design report once it is published, while Zommer confirmed Factchequeado will launch an interactive electoral rumours map next week, inviting journalists and community members nationwide to submit rumours for fact-checking. With the midterms only days away, all three speakers agreed on one point: filling the information gap with accurate, locally relevant reporting, before disinformation does, is the only reliable way for immigrant and ethnic communities to get the last word.
Speaker Bios
 |
Cameron Hickey
CEO, National Conference on Citizenship (NCoC) Cameron Hickey is the CEO of the National Conference on Citizenship. He is an award winning journalist and technologist who founded the Algorithmic Transparency Institute at NCoC to empower journalists, researchers, and civil society advocates to increase accountability for digital platforms. Prior to NCoC, Hickey led the Information Disorder Lab at the Shorenstein Center for Media, Politics and Public Policy at Harvard’s Kennedy School. As a journalist, he covered science and technology for PBS NewsHour and NOVA, where he won a News and Documentary Emmy Award and a Newhouse Mirror Award for his reporting. |
 |
Jinxia Niu
Program Manager, Chinese Digital Engagement, Chinese for Affirmative Action (CAA) Jinxia Niu is the programme manager of the Chinese Digital Engagement project at Chinese for Affirmative Action (CAA), where she manages three WeChat accounts and the Chinese-language fact-checking website PiYaoBa. She has ten years of experience in Chinese media writing and editing, is the author of two technology and society books in Chinese, and was previously a visiting scholar at Stanford’s Peace Innovation Lab. |
 |
Laura Zommer
Co-Founder and CEO, Factchequeado Laura Zommer is the Co-Founder and CEO of Factchequeado. She received the Recognition of Excellence at the 2025 Gabo Award and the Maria Moors Cabot Prize Special Citation 2024. She is an Ashoka Fellow and was a Knight Fellow at the International Center for Journalists (ICFJ) from 2022 until mid-2025. For over a decade she served as Executive Director of Chequeado, the first fact-checking organisation in the Global South, and she created LatamChequea, the first regional fact-checking network, bringing together more than 40 media organisations across 19 countries. She holds degrees in Communication and Law, teaches at the Universidad de Buenos Aires and the Universidad de San Andres, and has received numerous journalism awards, including the Gabriel Garcia Marquez Award for Innovation from the Fundacion Nuevo Periodismo Iberoamericana in 2015. |
