The most effective form of election disinformation is not necessarily a spectacular deepfake or a fabricated story produced by a Russian troll farm. Far more dangerous may be a message forwarded by a relative in the family WhatsApp group, repeated by a pastor to a congregation, broadcast in Spanish by a popular radio host, or shared by a respected community figure on Facebook, X, or WeChat. By the time the lie reaches the voter, the identity of its original source has disappeared. It arrives not as propaganda, but as a trusted warning from a friend.
This is especially dangerous in immigrant and diaspora communities, where a shared language, country of origin, and network of personal relationships often inspire greater trust than American institutions or the national English-language press. At an American Community Media briefing, Cameron Hickey, CEO of the National Conference on Citizenship; Jinxia Niu, a digital engagement expert with Chinese for Affirmative Action; and Laura Zommer, co-founder of Factchequeado, explained how election manipulation has moved into closed community groups—and how artificial intelligence has made it cheaper, faster, and more difficult to recognize.
Ahead of the November 3 midterm elections, the most dangerous lies do not necessarily seek to persuade voters to support a particular candidate. It is enough to confuse them. They provide the wrong election date, invent identification requirements, claim that mail ballots will not be counted, or suggest that ICE agents will be waiting for naturalized citizens at polling places. The purpose is not always to make people believe the entire story. It is enough to make them afraid, anxious, and uncertain until they finally decide to stay home.
This can be especially effective in mixed-status families. The voter may be a U.S. citizen while a spouse, parent, or child lacks secure immigration status. If that voter hears that an immigration raid is expected near a polling place, appearing in person may no longer seem worth the risk. No actual ICE operation is required. A convincing video, a fabricated image resembling a police photograph, and a few urgent words—“forward this to everyone”—may be enough.
A fake video circulated during this year’s campaign purporting to show ICE agents intimidating voters near a polling place. Other videos claimed that electronic voting machines were switching ballots cast for one candidate to another. These stories work because they do not invent entirely unfamiliar fears. They transform genuine anxiety into false evidence. Immigrants are afraid of ICE, while some Americans have distrusted the electoral system since 2020. Disinformation exploits those fears. It does not create a new reality; it rearranges existing anxieties.
One of the central falsehoods is once again the claim that noncitizens are voting on a massive scale. According to the story, Democrats admit immigrants into the country, register them, and use their ballots to determine election outcomes. There is no evidence for this. Noncitizen voting in federal elections is illegal and, according to election reviews, extraordinarily rare—closer to virtually nonexistent. No misconduct on a scale capable of affecting a national election has ever been demonstrated. Claims that the 2020 election was “stolen” failed in more than 60 court proceedings and were rejected by officials in Trump’s own administration. Yet politicians, radio hosts, and influencers are circulating the same story again in 2026, particularly through Spanish-language radio and social media in Florida.
The lie works in two directions. It persuades Republican voters that the system is inherently fraudulent while frightening American citizens from immigrant backgrounds. Green-card holders and people with temporary visas are also being told that an incorrectly completed form, an automatic motor vehicle registration, or even seeking election information could jeopardize their immigration cases. By blending actual rules with political threats, disinformation creates a fog in which even eligible voters may choose to retreat.
Jinxia Niu said the path of disinformation has changed in the Chinese-language information ecosystem. WeChat once served as its primary hub. Today, many stories originate on X, are translated into Chinese, and then circulate through WeChat groups, Telegram channels, and YouTube videos. Research by Chinese for Affirmative Action found that 48 percent of the misleading Chinese-language content it identified appeared on X, while WeChat’s share had fallen to 25 percent. Over one year, the organization documented more than 600 pieces of deceptive content that collectively received 11.6 million views.
That does not mean all Chinese-language propaganda originates in China. Much of it consists of translations of American far-right messaging. Chinese American influencers adapt conspiracy theories created in the United States to the community’s historical traumas: Democrats are portrayed as communists, social reforms as a Maoist takeover, and Trump as the leader who will save America from communism. The propaganda undergoes a cultural transformation. It is not merely translated; it is attached to memories that are deeply personal for first-generation immigrants.
Spanish-speaking voters face a different disadvantage. Research presented by Laura Zommer and Factchequeado examined responses from six leading artificial intelligence chatbots. Twenty-nine percent of their English-language answers to basic election questions were inaccurate, incomplete, or outdated. In Spanish, the failure rate was another 16 percentage points higher. Meta’s system twice gave November 4 as Election Day instead of the correct date, November 3. Four chatbots failed to provide answers that were both accurate and complete to more than 40 percent of election questions asked in Spanish.
This is no longer merely a problem of lies spreading on social media. Voters themselves are asking questions of systems they regard as objective sources of information, only to receive confidently worded false answers. Artificial intelligence may not be intentionally spreading disinformation, but the consequence for the voter is the same. Someone who goes to vote on the wrong day loses the vote.
An October survey by Hickey’s organization found that the share of people using chatbots to research candidates rose from 14 percent to 25 percent in just two weeks. The share using AI for at least one election-related task increased from 28 percent to 40 percent. Adoption was especially rapid among Black and Latino voters. Yet barely one in five respondents expressed strong confidence that the systems provided impartial and comprehensive information. People are increasingly relying on a tool they do not fully trust.
There is no single, convenient answer to the question of who the largest purveyors of disinformation are. Russia remains the most active foreign actor. U.S. intelligence and law enforcement agencies have previously linked Russian operations to fabricated videos, fake news websites, and stories alleging election fraud. Iran has also employed cyber operations and false media content, while China attempts to shape perceptions of American society according to its own geopolitical interests.
Foreign interference, however, does not absolve American political actors. The largest audiences are often reached not by Russian trolls but by American politicians, party activists, radio hosts, and influencers. Foreign intelligence services frequently do not invent the lies; they amplify falsehoods already circulating in American politics. The most successful manipulation occurs when it becomes impossible to tell where partisan propaganda ends and a foreign operation begins.
Warning people to look for poorly synchronized lips, unnatural blinking, or an artificial voice is therefore not enough. The technology is advancing too quickly for deepfakes always to be exposed through visible flaws. It is more important to determine who first published the material, whether a longer original recording exists, whether credible news organizations have reported it, and whether the appropriate local election office has confirmed the information. A screenshot or forwarded message is not a source.
Hickey recommends the SIFT method: Stop, Investigate the source, Find better coverage, and Trace the claim back to its original context. Corrections must also be returned to the same places where the falsehood spread. An English-language rebuttal on an official website will not reach an older Chinese immigrant who receives information through WeChat, or a Latino family that encountered the false warning in a WhatsApp group.
That is why community newspapers, multilingual fact-checkers, and trusted local figures are indispensable. In the struggle against election disinformation, the decisive voice is not necessarily the one that speaks first, but the one people ultimately believe. One of the invisible battles of the November election will determine who gets the last word in the family group chat.
If that last word belongs to a lie, there is no need to falsify a single ballot. It is enough to keep the voter from ever reaching one.
Comments
Comments are provided by Disqus.