Many predicted a revolution in the world of disinformation during the 2024 U.S. election due to the emergence of generative artificial intelligence (AI). But it hasn’t come to pass – at least not yet.
It is true that AI has increased the amount of false and misleading political and electoral content, and that this information has gone viral quickly. However, as of right now, we have not detected any major changes in the way disinformers operate: they continue primarily to leverage the major social media platforms to spread false content. These platforms, meanwhile, are more invested in developing profitable AI tools rather than addressing the disinformation, and the reliability and trust deficits it leads to.
Does this mean we can let our guards down and relax? Definitely not. The fact that a boom of disinformation created with AI has not exploded yet doesn’t mean that it can’t happen in the days leading up to election day on November 5.
This campaign cycle has been unique in a variety of ways, not least of which was the change of Democratic candidate from Joe Biden to Kamala Harris, and the assassination attempts on the Republican candidate, Donald Trump. The race is also neck and neck according to the polls, and could be the closest in recent decades.
Our Factchequeado team has detected AI being used to recreate videos with different audio or images inserted, as well as AI-generated images. We have not seen as many deepfake videos (ones manipulated with realistic multimedia) or manipulated audio clips of the two presidential candidates as we originally feared. Perhaps, as Claire Wardle, a disinformation expert and the co-founder and co-director of the Information Futures Lab at Brown University, often assesses, deepfakes were not necessary during this election cycle because less expensive “cheap fakes” – images that have been doctored or taken out of context, or emotional and misleading headlines – are still reaching people effectively.
Harris, the female candidate in the race, seems to have been the target of a higher volume of disinformation. On social media, thanks to AI, she was shown falsely dressed in a communist leader’s outfit, in a bikini with the American flag, embracing Trump, and more.
Meanwhile, AI has been used to create false images of Trump barbecuing in the Bronx, hugging a group of Black people, and as especially heavy-set. We have also detected false images of Secret Service members smiling after the first assassination attempt on Trump in July.
The vice presidential candidates, Tim Walz and JD Vance, haven’t been spared from AI-generated disinformation. Walz was shown next to Harris in front of a Communist Party poster and falsely dressed as a trans cowboy. Vance, on the other hand, has been depicted more favorably, with a more bearded and angular face than he actually has, in keeping with contemporary beauty standards for men.
Those of us who monitor Spanish-language disinformation are most worried about the possible spread of fake audio messages without context on WhatsApp, which Hispanic and Latino people use to discuss politics more than other groups.
Provided this context of no AI disinformation revolution – at least, for now – what we’ve seen this election cycle is the repetition of 10 types of electoral disinformation that Chequeado and LatamChequea, with the support of UNESCO, identified in 2018 as a constant in elections across Latin America. These same disinformation narratives have also been present in North American elections in 2018, 2020 and 2022.
Here are these 10 types of disinformation:
- Type 1: Irregularities in the electoral process falsely used as proof of fraud
- Type 2: Unsubstantiated claims that authorities are committing fraud
- Type 3: Disinformation that votes were cast in the name of deceased persons
- Type 4: Misinformation regarding who has the right to vote and who doesn’t
- Type 5: Manipulation to prevent people from voting, or invalidate votes
- Type 6: Disinformation about the documentation required to vote
- Type 7: Disinformation about the votes of citizens abroad
- Type 8: Chaos on Election Day
- Type 9: False polls
- Type 10: False statements from candidates
Undoubtedly, the star of this 2024 election cycle has been disinformation narrative #4 – that people who are ineligible to vote have voted or will vote. This narrative lays the groundwork for the disinformation alleging election fraud. (See examples of disinformation related to election fraud here, here, here and here.)
Trump, Vance and other Republican candidates have regularly repeated that “illegal immigrants” were coming to the U.S. to register and vote. These claims have gained traction in various formats and have been included in hundreds of pieces of content – some with paid ad boosting on Facebook and Google to enhance their visibility.
In some cases, further disinformation was added to the narrative, claiming that the Biden administration paid for the travel of these immigrants. In other instances, disinformation also claimed that laws were changed to give undocumented immigrants legal status so they could cast their votes later.
This is all part of the larger disinformation narrative around electoral fraud that Trump promoted in 2020. More than 60 federal courts across the country ruled against Trump and his allies, finding that the “The Big Lie” assertions were unsubstantiated.
In order for you to better inform and protect yourself from manipulation and deception in the final stretch of this frenetic election cycle, our team at Factchequeado created two innovative tools based on what we and other fact-checkers have learned in our work fighting disinformation for over a decade.
Electopedia
Electopedia is a website that provides detailed information in both Spanish and English about the U.S. elections. This project is supported by the Logan Family Foundation and The William and Flora Hewlett Foundation, through the ICFJ’s Disarming Disinformation initiative.
The Electopedia site offers critical resources, including:
- Articles and videos in both Spanish and English on election-related topics and the voting process
- Educational videos that answer common questions about how to vote, the requirements for voting and the importance of casting your ballot
- Direct links to the secretary of state’s website for each state, with details on voter registration deadlines, and online, mail and in-person voting
- Polling site locators and mail ballot drop box location information
- Interactive graphics explaining the voting power of Latino communities
Electopedia aims to inoculate communities against disinformation so that they avoid falling victim to false or misleading information. Electopedia also seeks to encourage members of Latino communities in the U.S. to vote.
Creating Electopedia is part of a pre-bunking (preemptively debunking) strategy that takes into account the 10 types of electoral disinformation described above. The project also seeks to fill information gaps, clarify technicalities related to voting, and explain complex voting processes in a simple way that helps Latino and Hispanic community members avoid falling into mis- or disinformation traps or being misled.
Electobot
Electobot is WhatsApp’s first Spanish-language conversational chatbot that answers questions about the electoral process in the U.S., providing verified information about the candidates and the statements they put out.
The tool helps combat Spanish-language mis- and disinformation by generating clear, accurate responses in real time.
The quick, accessible nature of the chatbot helps ensure that Spanish-speaking Latinos can make informed decisions when voting. When people share their location with the chatbot, it also provides accurate information about where they can vote.
This tool was developed by Media Plus, the marketing and IT arm of the Cuban media outlet El Toque, one of Factchequeado’s 115+ partners.
If you’re a journalist, editor or owner of a media organization serving Latino/Hispanic communities in the U.S., you can learn more about Factchequeado and partnership opportunities here.
Main image by Sora Shimazaki on Pexels.
This article was originally published by IJNet in Spanish. It was translated to English by journalist Natalie Van Hoozer.
