Abstract
The use of artificial intelligence transforms election campaigns with tools such as microtargeting. Campaigns can personalize their political communication and influence voters’ decisions. This offers new opportunities and raises ethical challenges related to misinformation and the use of personal data, which impact on the quality of democracy.
In recent years, the use of artificial intelligence (AI) has become increasingly common, especially in the field of image generation. Political parties and electoral processes have been notably affected by this phenomenon. The widespread use of social media has been followed by the emergence of these new technologies, which are playing an increasingly sensitive role in the dissemination of political propaganda. This has raised countless debates, as new and more sophisticated approaches raise both ethical and technical questions.
At its core, AI-generated content allows for the creation of images based on user input. This enables campaigns to construct visual narratives representing potential future scenarios (showcasing the positive outcomes of certain policies or the negative consequences of others). However, from the outset, this practice introduces the issue of creating artificial images that do not reflect reality, potentially falling into the realm of fake news or biased narratives. A similar result occurs with AI technologies that can mimic a person’s voice, facilitating the spread of falsified audio recordings.
In recent years, many campaigns have employed AI. On the surface, we see its use in generating videos that depict highly optimistic alternative scenarios in the event of an electoral victory (or, conversely, devastating scenarios in case of a loss), as was done by the campaigns of Sergio Massa and Javier Milei in the 2023 Argentine elections. Similarly, AI-generated images of Donald Trump being arrested following his court conviction (featuring increasingly implausible arrest scenes aimed to generate viral content and memes).
Social media has leveled the playing field in terms of reaching the electorate, particularly for sectors that may have been marginalized by mainstream media. These platforms provide a straightforward and low-cost means of dissemination. However, their rapid reach also facilitates the widespread propagation of fake news with far greater ease, making fact-checking and refutation more complicated.
Although its use is relatively recent, numerous studies have already explored the impact of artificial intelligence on electoral processes, especially in content creation for campaigns. In this paper, we will attempt to shed light on the impact of artificial intelligence on network design and development, as well as its role in organizing social support—an area that has received less attention in research.
AI on Electoral Campaigns
Artificial intelligence refers to
The concept of artificial intelligence dates back to the second half of the 20th century. The famous Turing test posed the challenge of determining whether ordinary people could distinguish between interacting with a human or a chatbot in a written conversation (Turing, 1950). However, global enthusiasm for AI did not reach its current heights until recent years. Since 2023, AI-generated images capable of emulating human photographs have begun to emerge.
In the political arena, the impact of AI has been significant. Electoral campaigns are characterized by the need to persuade voters to support a particular candidate, and in recent years, to also demobilize or push voters to reject an opposing option when gaining their support seems unattainable. Although campaigning today is different from fifty or a hundred years ago, the ultimate goal remains the same.
Electoral campaigns in the 20th century were marked by their mass appeal and focus on general issues. While there are various theories about voter behavior, there is a consensus in the literature about the importance of social class in voters’ decision-making, especially when this meant a determining factor.
Campaigns were largely conducted through mass rallies led by prominent political figures, while party cadres worked within their communities. The main communication channels were radio, television, and print media, all of which shared a “common characteristic: they were indiscriminate. In other words, they were not tailored to specific profiles but aimed at persuading society as a whole” (Cebrián Beltrán, 2024).
In traditional or modern electoral campaigns, communication occurred in a controlled environment, with format limitations that curbed or reduced aggression and disinformation. The electorate generally maintained a moderate stance, which discouraged and even punished extremism. Moreover, there were fewer media outlets, and their impact and reach depended heavily on their credibility. As a result, they were highly cautious in fact-checking and verifying the information they disseminated (Rubio Núñez et al., 2024).
However, the decline of major political parties in the 20th century across Europe and Latin America coincided with a transformation in the electorate, which shifted from being a relatively uniform mass to numerous subgroups organized around much more specific ideals (environmentalists, advocates for women’s political participation, minority rights, regionalist parties, separatist or independence movements, etc.).
This new reality compelled political groups to make greater efforts to understand their electorate and to use new technologies, both to better comprehend them and to represent them.
The arrival of social media, of course, has proved a revolutionary incentive, perhaps on par with or even surpassing the impact of television. While previous methods of communication have not been abandoned, social media has led to an unprecedented mass dissemination of information, allowing political messages to reach nearly every sector of society in much shorter timeframes. This has posed a challenge for majority sectors and even authoritarian regimes seeking to maintain control over the flow of information.
With current technologies, even in authoritarian contexts, an opposition candidate can design a coherent and competitive campaign that provides access to a wide range of resources (spots, jingles, interviews, and the dissemination of propaganda, among others)—something unimaginable in the authoritarian contexts of the 20th century.
This is evident in democracies. We have witnessed new ways of campaigning that significantly reduce the need for physical presence and large-scale rallies. For example, in Chile’s 2021 presidential elections, Franco Parisi (from the Partido de la Gente) ran as a candidate despite residing in the United States. Amid the pandemic, which encouraged the use of various remote communication tools, Parisi campaigned from abroad and garnered 900,000 votes, equivalent to 12.8% of the total ballots cast (Servel, 2021).
Microtargeting: Understanding the Voter
With these new technologies, it is now possible to do something that previous ones either did not allow or limited: create personalized political messages directed at very specific segments of voters, known as microtargeting. This strategy relies on audience segmentation, using demographic, behavioral, and preference data to design messages that resonate with the characteristics and needs of small groups of individuals, often referred to as microsegments or clusters. AI provides sophisticated and innovative mechanisms to facilitate this task.
The process works as follows: individuals provide their personal data when using apps, visiting websites, making online purchases, etc. The aggregation of all this information, along with its analysis and processing to correlate data and identify patterns and trends, is known as big data. This data serves as the foundation for algorithms, which are essentially “a sequence of commands that instruct a computer to convert an input into an output. For example, a list of individuals sorted by age. The computer takes the ages from the list (input) and produces the newly sorted list (output)” (FRA, 2018).
In the electoral realm, which is the focus of this article, the algorithm applied to big data aims to profile voters. In other words, the data is analyzed to understand voters’ political stances, and predict and influence their electoral behavior. This process culminates in the formation of microsegments or clusters—groups of individuals who share similar profiles.
Once the different groups have been characterized, specific messages are crafted for each one. The content of these messages will depend on the objectives at hand. For example, if a group of potential voters is identified as being demotivated or disengaged, targeted messages can be designed to involve them in the electoral process, raise awareness of the political situation, and convince them that their vote is crucial.
Conversely, if a segment is identified as likely to vote for the opposition, specific messages can be crafted to demobilize them. This might involve highlighting cases of corruption within the opposing party or statements from their leaders that conflict with the convictions of that particular segment.
Many of these messages do not appeal to voters’ analytical capabilities but instead target their emotions. As such, it is increasingly common to see controversial topics (such as immigration, abortion, or religion) dominate political discussions to provoke strong emotional reactions and simplify the political debate. This trend has promoted the use of disinformation as a tool to solidify the electorate, as well as the use of generative AI to create and manipulate audiovisual content to incite outrage.
This technology is so fast-paced that it allows campaigns to assess its impact in real-time and make adjustments accordingly. While modern campaigns once relied on traditional opinion studies like polls or focus groups to gauge the impact of a message, proposal, or slogan, microtargeting and social media enable preliminary analyses of campaign strategies’ effectiveness within minutes.
This new strategy enables political parties and candidates to address concerns of specific segments that may have previously gone unnoticed, despite their significant electoral potential, thereby increasing the efficiency of electoral campaigns by allowing real-time analysis.
Microtargeting, however, can also be misused to present voters with misleading or biased information, to spread fake news and disinformation, and to create echo chambers that limit debate. This undermines trust in institutions and weakens the foundations of democracy.
The Cycle of Artificial Intelligence in Political Participation
Source: Cebrián Beltrán (2024).
Impact on Electoral Outcomes
Measuring the impact of various AI-driven tools on electoral outcomes is still a challenging endeavor. Establishing a rigorous methodology for this purpose is, at least for the time being, impossible. However, mechanisms exist to analyze the communication environment and record the discussions taking place within specific societies. In this regard, it is possible to assess the success of AI-powered tools in establishing issues on the public agenda, as well as the positions citizens take concerning them.
It can also be asserted that certain topics or key ideas in some elections have significantly influenced the results. Furthermore, the use of technologies such as microtargeting has maximized the impact of these issues on the electorate.
Below are some emblematic cases where technology, particularly the use of big data to design specific campaigns, has affected electoral outcomes.
2008: The Obama Case
Long before AI became a topic of discussion, a paradigm shift marked a turning point in electoral campaigns: microtargeting. In the 2008 U.S. presidential elections, the campaign team of then-candidate Barack Obama succeeded in profiling every voter in the country, focusing on two key points: whether they were likely to vote and if they would vote for Obama. Based on this information, strategies were developed to influence their decisions, ultimately leading the candidate to the White House. This was one of the first large-scale experiences in voter profiling, albeit not with the level of sophistication we recognize today.
2016: Cambridge Analytica
Donald Trump’s election was overshadowed by the Cambridge Analytica and Facebook scandal, which would become a case study in the use of personal data to craft ultra-segmented political messages.
Aleksandr Kogan, a professor at the University of Cambridge, designed a personality test for Facebook in 2013, through which he obtained data from 50 million people. The test was completed by 265,000 users of the social media, who, to participate, had to grant access to their friends’ information without their consent.
Using the data from these individuals, Cambridge Analytica created psychological profiles and designed specific messages aimed at influencing their political preferences, even disseminating fake news (BBC World, 2018).
2023 in Argentina: Images and Videos
In the 2023 presidential elections in Argentina, the teams of candidates Sergio Massa and Javier Milei extensively utilized generative artificial intelligence to create promotional images and videos, as well as attacks on their opponents. The official candidate’s team employed AI to produce posters and videos depicting Massa as a strong and charismatic leader, drawing inspiration from Soviet styles and pop culture. In response, Milei released images portraying a lion liberating Argentina and depicting Massa as a communist leader. The use of generative AI by Massa’s campaign to illustrate a potential dystopian future in the event of Javier Milei’s victory sparked controversy (Nicas and Cholakian Herrera, 2023).
2024 in India: Chatbots
In India, during the campaign for the general elections of 2024, a controversy erupted regarding deepfakes on social media when a user asked Google’s AI tool, Gemini, about the alleged fascist nature of Prime Minister Narendra Modi and the Bharatiya Janata Party (BJP). The response indicated that Modi’s government was “accused of implementing policies that some experts have characterized as fascist” (Dillon, 2024). Indian Minister of State for Electronics and IT, Rajeev Chandrasekhar, criticized this response, stating it violated the country’s laws. This incident underscores the growing concern in India regarding disinformation and the use of AI in electoral contexts. Google promptly reacted, asserting that it was “working” to “improve the reliability” of the tool (Mukherjee, 2024).
Examples in Denmark and the United Kingdom
The Synthetic Party is a political party in Denmark led by an AI named Leader Lars, a chatbot accessible via Discord. Its goal is to engage citizens who typically do not vote and place technology at the center of political debate, promoting coexistence between AI and humans, as well as the regulation of AI accountability. The party, which defines itself as synthetic, develops its platform based on proposals from minor Danish parties dating back to 1970. Although led by an AI, the project is driven by the artist group Computer Lars and the technology center MindFuture, which seek to ensure the party’s longevity and global expansion. They also propose the creation of a new Sustainable Development Goal (SDG) focused on the relationship between humans and robots (Vicente, 2023).
Another example emerged in the United Kingdom, in the lead-up to the 2024 elections, featuring the candidate of Smarter UK, AI Steve, an AI avatar representing legal candidate Steve Endacott in the Brighton Pavilion constituency. The aim was to rekindle interest among apathetic sectors of the population in politics by allowing voters to directly influence decisions. If elected, Endacott would physically represent AI Steve in the British Parliament, acting according to the majority vote of the electorate. The project faced several ethical questions regarding its efficacy and raised various legal concerns. Nevertheless, AI Steve came in last in the constituency (a stronghold of the Green Party) with only 179 votes (Smith, 2024).
Conclusions
The changes in the way electoral campaigns are conducted over the past two decades have been staggering. This dynamic, combined with the decline of traditional parties and the emergence of disruptive, charismatic, and populist figures, has resulted in a 180-degree turn in political communication.
The advent of social media, along with the sophistication of methods for collecting, analyzing, and cross-referencing data on a massive scale, profiling the audience (or electorate), and crafting specific messages has ushered us into a new game—one that cannot be understood through old categories.
However, this new reality also brings forth new challenges, perhaps even greater than those of the past. In recent decades, we have experienced democratic fatigue, characterized by political disaffection, a crisis of representation, and a decline in citizens’ adherence to democratic principles. The emergence of big data and AI presents a challenge not only for parties and the electorate but also for institutions, the integrity of the communication space, and ultimately, the democratic system itself.
On the other hand, new technologies have also allowed for a deeper understanding of the electorate, complicating and problematizing it. This presents an opportunity for parties to consider the needs and concerns of voters, ensuring that these factors have a genuine impact on their programs and proposals.
For now, it appears we are still uncovering the effects of artificial intelligence on electoral campaigns, and it would be premature to draw definitive conclusions. Its use poses significant challenges and raises fundamental ethical debates, such as whether individuals are genuinely free to access plural, broad, and critical information, or whether they are increasingly confined within informational bubbles specifically designed to shape their choices.
References
BBC Mundo. (2018, March 21). 5 claves para entender el escándalo de Cambridge Analytica que hizo que Facebook perdiera US$37.000 millones en un día.
Cebrián Beltrán, S. (2024). De la talla única al traje a medida: el microtargeting político para influir en las elecciones. Paper at XXI Congreso de la Asociación de Constitucionalistas de España, round table “Garantías constitucionales de elecciones libres”, Valladolid.
Dillon, A. (2024, February 26). India confronts Google over Gemini AI tool’s ‘fascist Modi’ responses. The Guardian.
FRA. (2018). #BigData: Discrimination in data-supported decision making.
Issenberg, S. (2012, December 19). How Obama’s Team Used Big Data to Rally Voters. MIT Technology Review.
Márquez Díaz, J. (2020). Inteligencia artificial y Big Data como soluciones frente a la COVID-19. Revista de Bioética y Derecho, 50.
Mukherjee, M. (2024, March 19). AI deepfakes, bad laws – and a big fat Indian election. Reuters Institute.
Nicas, J., & Cholakian Herrera, L. (2023, November 15). Las campañas electorales de Argentina recurren a la IA. The New York Times.
Rubio Núñez, R., Franco Alvim, F., & Andrade Monteiro, V. (2024). Inteligencia artificial y campañas electorales algorítmicas. Madrid: CEPC.
Servel. (2021). Elección presidencial 2021.
Smith, C. (2024, July 2). Britain’s first AI politician claims he will bring trust back to politics – so I put him to the test. The Conversation.
Turing, A. M. (1950). Computing Machinery and Intelligence. Mind, LIX(236), 433-460.
Vicente, M. J. (2023). Inteligencia artificial y política. Los casos de “Synthetic Party” y Tama. In A. Dafonte Gómez & M. I. Míguez González (coords.), El fenómeno de la desinformación: reflexiones, casos y propuestas (pp. 603-617). ISBN 978-84-1170-538-7.
