This fall, Meedan facilitated a coalition of more than 30 media, fact-checking, and civil society organizations covering elections in the U.S. In addition to participating in a shared Slack channel and joining regular video calls, several partners used Check — Meedan’s open-source software platform — to operate tiplines for connecting with their audiences. Using WhatsApp and Messenger, users could send in queries about rumors they’d seen online or heard in conversation. Throughout the election cycle, these queries helped to inform research and reporting topics for our diverse set of partners.

All told, the collaborative efforts of the coalition gave us a unique lens through which to observe election-related rumors and falsehoods over time. Here are our main observations, informed by coalition partners and tiplines, on notable trends from the election cycle.

In the era of generative AI, time-tested tactics were what really drove doubt

This year, tech critics warned that AI would run roughshod over election information ecosystems worldwide. But while deepfakes and AI-generated media certainly made the rounds — and we have some observations about that in another section — social media influencers proved that once the seeds of uncertainty have already been sown, simpler tactics remain effective for ginning up suspicion about election processes.

In the run-up to Election Day, the overwhelming majority of election integrity-related tips we received contained decontextualized photos, misleading captions, or mischaracterized explanations of government documents. 

In Texas, the secretary of state issued an advisory explaining that naturalized citizens could use limited-term driver’s licenses or ID cards as valid identification when voting. This was wildly mischaracterized as proof that noncitizens could vote in Texas, an assertion that was amplified by InfoWars conspiracy theorist Alex Jones. 

In Pennsylvania, a screenshot of a page on the state’s open data website displayed the number of people who had requested mail-in ballots twice. This image was used to support viral claims of voter fraud — even though the screenshot specifically showed that officials had declined to fulfill duplicate requests.

By Election Day, there was a surge of real-time misinformation about voting machine malfunctions, ballot tampering, and poll worker malfeasance, particularly in Georgia, Pennsylvania, Michigan, and Kentucky. Again, numerous tips highlighted normal procedures or standard documents but then mischaracterized these items as evidence of wrongdoing. While some posts seemed intentionally crafted to cast doubt, others seemed to be triggered by genuine anxiety and an overall sense of worry that the election would be rigged. A self-identified Trump voter in Pennsylvania posted a photo on X of a Democratic committee sample ballot he claimed was handed to him in line, highlighting the names of Democratic candidates running for office in that district — a legal and commonplace practice outside polling places. “Doesn’t seem to be a choice,” he wrote. “WTF!!!” He later added to the thread, explaining that the card in question was a sample ballot and that he was concerned that it might be evidence of electioneering.

AI-generated fantasy memes took center stage

While generative AI didn’t upend our information landscape to the extent that some experts had forecast at the start of 2024, it did figure into some of the more enduring rumors of the election cycle. Numerous viral AI-generated images seemed intended to cast leaders in a positive, heroic light to appeal to voters’ emotions, more than to suggest that something happened that didn’t.

One of the more popular images in this vein was inspired by former President Donald Trump’s baseless allegation during a presidential debate that Haitian immigrants in Springfield, Ohio, were killing and eating their neighbors’ pets. Springfield officials, and even Ohio Gov. Mike DeWine, emphatically repudiated the rumor. Nevertheless, this false assertion had real-life consequences for people across Springfield when schools and government buildings received bomb threats following the debate.

The rumor took on a life of its own on social media, giving rise to a series of eye-catching, if obviously fake, AI-generated memes. In one, Trump can be seen cuddling a kitten and a duck while submerged in a lake. Another fabricated image showed the candidate running from a pair of shirtless Black men with a kitten tucked under each arm. Similar memes surfaced following Hurricane Helene, including images that depicted Trump wading through floodwaters in what appeared to be an effort to help residents stranded by the storm.

In a recent interview with The Washington Post for an article about AI’s impact on the partisan divide, Purdue University political scientist Kaylyn Jackson Schiff described such memes as examples of voters developing “positive attitudes or an understanding of current events” through their interactions with deepfakes. Voters do this, she said, “even if they don’t think that the image itself is actually real.”

The myth of the noncitizen voter

One of the most prominent, consistently repeated falsehoods shared on our tiplines was the Trump campaign’s assertion that Democrats were ushering immigrants into the country specifically so they could vote in the November election. Early in the campaign, Trump wrote on Truth Social that “crazed” Democrats were allowing migrants into the country “so they can vote, vote vote.” He repeated this message in speeches and press interviews throughout the campaign cycle.

Surrounding Election Day, our tiplines received multiple queries about a video of a young Black man who claimed to have come from Haiti six months prior to the election, obtained citizenship, and voted for Kamala Harris in two Georgia counties. He held up a set of drivers licenses as if to corroborate the claim. The video got plenty of attention online, but the FBI soon traced it back to Russian “influence actors.”

Similar falsehoods could be found in multiple languages. On our Spanish-language tiplines, questions surfaced about allegations concerning CBP One, the mobile app that U.S. Customs and Border Protection uses to administrate customs and asylum application appointments at the U.S.-Mexico border. Trump and Elon Musk insinuated in social media posts that CBP One was being used by the Biden administration to grant asylum en masse so migrants could vote on Nov. 5. Other rumors suggested that the app would stop offering appointments or shut down altogether during elections. All such claims were unfounded.

Fabricated headlines surged

Falsified headlines and fake stories attributed to mainstream media outlets were another common occurrence on social media during this election cycle. Doctored screenshots of vote count updates from CNN were a regular occurrence on Election Day. A fake headline in The Atlantic, which is typically seen as a left-leaning publication, read “To Save Democracy Harris May Need To Steal An Election.” And a fabricated Time magazine cover featured Donald Trump as “Man of the Ear,” a dual reference to the injury Trump sustained following the attempt on his life at a Pennsylvania rally and to Time’s annual Person of the Year issue. Although its creators may have been aiming for satire, the image made the rounds throughout global media ecosystems, prompting questions about its veracity from audiences in Indonesia, India, and in multiple Spanish-speaking countries.

Designed to resemble authentic graphics from reputable organizations — usually using photo-editing software or crude cut-and-paste techniques — these fabrications leveraged visual familiarity to spread quickly and influence audiences.

Right-wing solidarity, from Brazil to India

Shortly after Election Day, narratives about Trump’s decisive victory brought about a mix of rumors and fantasy AI fabrications that underscored some of the parallels between Trump and other prominent leaders on the right. 

From Brazil, rumors emerged that Trump had said President Luiz Inácio Lula da Silva was not invited to his January inauguration because “a thief's place is in jail and not in the presidency.” This is a reference to Lula’s conviction and imprisonment on corruption charges following his second term in office.

Our partners in India received tips featuring a fabricated photo of Trump and Elon Musk waving to a crowd of supporters while wearing saffron-colored silk formal wear, a reference to the signature color of Prime Minister Narendra Modi’s Bharatiya Janata Party. We also received submissions from tipline users in both India and the U.S. about claims that, during Trump’s victory speech, his supporters had shouted “Modi! Modi!” This was swiftly debunked — the crowd shouted “Bobby! Bobby!” after Trump gave a rhetorical nod to Robert F. Kennedy Jr., who threw his support behind Trump after ending a third-party run for the presidency.

Key learnings from our coalition 

For Meedan, the U.S. 2024 Election Exchange represented a unique opportunity to experiment with the collaborative potential of a multilingual and globally active network of partners. We knew from the start that online rumors travel between languages and that they can quickly jump from one social media platform to another as they take shape and ultimately influence key narratives and modes of thinking in other parts of the world. The pervasive nature of the noncitizen voter myth — and of related narratives that racialized and politicized immigrants — showed our coalition just how powerful and widespread ideas like these could become, especially during a contentious election season. 

We also saw signs that such narratives will continue to ebb and flow globally regardless of the election cycles of any one individual country. In the future, we may test out coalitions that track specific prejudicial ideas across different markets, without tying that work to an election cycle.

This project also enabled us to test-drive a relatively new feature of Check, our open-source collaboration software, that allowed partner organizations to pool queries sourced from their respective tiplines. The “Shared Feeds” feature helped us to form a more complete picture of the issues that were trending throughout the period in question. 

And we found that real-time interactions on our group Slack channel and during our regularly scheduled video calls proved that, for some parts of our work, nothing could replace a real, live conversation. These shared virtual spaces enabled us to exchange ideas, puzzle through tough questions, and find out what partners across a diverse range of media spaces were seeing and hearing each day. These interactions enabled us to get to know each other and to become more familiar with each group’s working styles — a critical if not always named component of coalition work.

Looking ahead to 2025, we aim to build upon this work as we strive to increase our impact. We look forward to exploring these considerations as a team, and as a community, in the months to come.