Generative artificial intelligence (AI) significantly impacts on social relations and is changing the ways in which people interact in institutional frameworks. There are assessments of negative impacts of AI on teaching, research and also in politics, and this is worrying because it can be used to manipulate citizens on the basis of misinformation. Unfortunately, “our current democratic systems are only able to manage the digital transformation to a limited extent” (Duberry, 2022).

Weak information quality and lack of solid data available to citizens leads to poor performance of productive systems, and affects the democratic framework of nations. It should be remembered that “democracy requires a well-informed citizenry and involves people with diverse perspectives coming together to debate and find common solutions through dialogue and participation” (UNESCO, 2024, p. 16), so it is important to evaluate the responsible use of AI for governance, particularly as there is evidence of the creation of false messages online to affect electoral processes (Luminate, 2023; Marsden et al., 2020; Schia and Gjesvik, 2020).

The presence of false information in elections is not a new phenomenon (Posetti and Matthews, 2018), but after the start of Covid-19 it attracted much attention from international bodies due to the threats it poses to trust in public institutions (Palau-Sampio, 2024). In the United States, “misleading AI-generated content shaped US electoral discourse by amplifying other forms of disinformation and igniting political debates” (Stockwell et al., 2024, p. 3). “In Venezuela, state-controlled media used AI-generated videos [.] to spread pro-government propaganda” (Mansell et al., 2025, p. 19). In 2022, fallacies circulated, appeals were made to fear and “attacks to delegitimise, damage or harm the figure of the candidates for the presidency of the Republic of Colombia, which aim to exacerbate emotions that deepen the breakdown of social consensus” (Echeverría and Rodríguez, 2023, p. 239); in that country, the use of AI tools tends to increase “the propagation of fake news, the manipulation of trends in public opinion and the discrediting of opposition candidates, which may encourage political groups to favour conflict over cooperation” (López et al., 2024, p. 18). This is an era of “high-risk elections” (Alvim et al., 2024).

In the electoral terms, “the irruption of AI has given rise to a new phase in the evolution of electoral campaigns, algorithmic campaigns, which threatens the progressive dehumanisation of electoral campaigns” (Rubio, 2025, p. 75), AI can have an impact on predicting and forecasting voter behaviour and possible outcomes, developing and disseminating content (text and image) for campaigns (Dad and Khan, 2023). “Although the threats of AI are not fully understood, its potential capacity erodes public trust in democratic institutions and processes” (Forum on Information and Democracy, 2024, p. 100), and among these institutions are the mass media (Londoño and García-Perdomo, 2024).

In the Global Risks Report 2024, disinformation and misinformation driven by AI and social media are recognised as the world’s greatest near-term threats, especially in the context of their negative impact on the credibility of elections (World Economic Forum, 2024). “There is an urgent need for governments, social media platforms and other businesses to protect the fundamental pillars of a democratic society and its implementation mechanisms: elections” (UNESCO, 2022, p. 26), among other reasons because “AI-assisted political decision-making may ultimately lead to the establishment of a form of automated democracy and deprive humans of autonomy over political processes” (Ramos, 2022). As noted above, both the United States and the European Union have taken steps to safeguard the integrity of their elections, but “protecting democratic processes in the digital age requires a multifaceted approach involving governments, regulators, the private sector and civil society. The goal is not to stifle freedom of expression, but to promote a fairer information ecosystem” (Muto, 2025). Potential regulations should be evaluated so that they do not restrict the freedoms and recognition of political positions from which people choose.

In general, AI regulation should be complemented by the support of civil society, corporate ethics committees and public bodies for multiple and diverse oversight. This approach gained weight because “following the European elections in May 2019 [.] they see audited co-regulation as a more desirable system of governance because it is more appropriate for this era and context, and for the scale and rapid evolution of the problem” (European Union, 2019). Co-regulation, or regulated self-regulation, “is a model that involves private regulation (either in the form of self-regulation or individual regulation) that is actively supported or encouraged by government, and is usually backed by legislation” (Article 19, 2019).

The body that intends to regulate AI in elections must uphold democratic values, social justice including “transparency, media literacy and the introduction of human-centred forms of co-regulation” (Marsden et al., 2020, p. 18). Previous studies emphasise that, “to successfully navigate the complexities of a world increasingly shaped by AI, it is imperative to develop robust regulatory frameworks, promote AI literacy and foster critical thinking” (Arda and Başarır, 2024, p. 137). Furthermore, “AI systems intended to be used to influence the outcome of an election or referendum [.] should be classified as high-risk AI systems” (European Union, 2024).

As in the past cases, in the face of emerging phenomena, the international human rights framework appears to be the best mechanism for formulating standards for generative AI practices. AI must operate in accordance with the Universal Declaration of Human Rights. In this regard, in March 2024, the UN General Assembly “adopted a resolution [.] to promote “safe and reliable” AI systems that respect human rights and contribute to sustainable development” (United Nations, 2024). “Any scheme of self-regulation, co-regulation or state regulation should promote the adoption of user-directed content policies that are in line with Inter-American human rights standards” (Chorny et al., 2022, p. 71).

In Latin America, as in other countries, the regulatory landscape of AI is discussed, and it is sought that “as a premise, any initiative to contain the development of AI must have as its unwavering objective the protection of internationally recognised fundamental rights” (Giandana, 2024, p. 3), because “it is easier to harmoniously agree on practices around the development and implementation of these systems” (Access Now, 2024, p. 11). Implicitly, “a good legislative environment, transparency and effective controls are the main tools against manipulation techniques” (Hajdúková, 2022, p. 22).

Therefore, it would be expected that, based on lessons learned and the guidance of supranational bodies, manipulations and disinformation would decrease, however, this is not the case in Ecuador. In this Andean country, which in recent years has attracted global attention for its poor governance, incipient practices of using AI for manipulation in electoral periods through fake accounts, troll farms and cyber troops were identified (Barredo-Ibáñez et al., 2021; Bradshaw and Howard, 2017). Interference and disinformation in election campaigns through social media and AI is not new, there is evidence of digital manipulation strategies in X during the 2017 and 2020 presidential campaigns (Ackerman et al., 2020; Puyosa, 2017). In 2019, Quito mayoral candidate Jorge Yunda used AI in his campaigning and is considered to have contributed to his victory (Calderón, 2019).

In 2025 it was reported that “disinformation was widespread in the campaign, various fraud narratives proliferated. The EU Election Observation Mission observed AI-generated and manipulated content on social media, including paid content” (EU Election Observation Mission, 2025, p. 2), similarly, “multiple tactics were used to discredit candidates, including AI-generated black campaign ads” (EU Election Observation Mission, 2025, p. 7). The country’s media also warned about the use of AI through “audio montage, manipulation of video images, alteration of candidates” photographs, among others, were content generated with (AI), which were disseminated on social networks” (Jara, 2025). “In Ecuador’s presidential campaign, the sophisticated use of AI was evident, such as the circulation of incredibly realistic altered videos. The candidates, mainly Luisa González and Daniel Noboa, were targets of attacks and disinformation generated with AI” (Primicias, 2025). The local organisation Usuarios Digitales (2025) reported that, between 18 January and 2 April 2025, official election campaign time (Consejo Nacional Electoral, 2025), publications were systematically circulated on social media accounts where AI was used to support, attack or satirise rival candidates, as shown in Table 1.

There are initiatives to regulate AI in Ecuador that were presented as draft laws, but none of them mention the risks in elections (Asamblea Nacional del Ecuador, 2025). The proposals share guiding principles such as privacy and security of personal data, transparency and technological neutrality. They also incorporate the qualification of facial recognition systems, autonomous vehicles and automated health decision-making systems, among others, as high-risk AI projects.

Based on the above, the objective of this work is to understand the perceptions of Ecuadorians regarding the use of AI to generate disinformation in Ecuador’s, 2025 presidential election. The hypotheses of the research are (1) there is a significant association between compliance with the regulation of the use of AI in electoral campaigns and the perception of the outcome of Ecuador’s presidential election in 2025. (2) There is a significant association between the use of AI in election campaigning and the level of trust in candidates’ advertisements. (3) There is a significant association between the use of AI in election campaigning and trust in election results. The research questions are (1) Do Ecuadorians agree with regulating the use of generative AI in electoral campaigns in Ecuador? (2) Do Ecuadorians perceive that the misinformation created through AI influenced who won the presidency of Ecuador in 2025? (3) Did the use of AI in Ecuador's, 2025 presidential campaign increase trust in the candidates and election results?

The methodology is mixed, quantitative and qualitative, with an explanatory scope. The instruments used are two surveys, semi-structured interviews, before and after the presidential elections, and sentiment analysis of the opinions expressed in the second survey, that is, a methodological triangulation was sought to investigate the causes of the phenomena (Igartua, 2006), in addition to seeking to review different perspectives of a phenomenon by combining instruments and orientations of different methodological designs (Creswell, 2014; Creswell and Plano, 2011).

The first survey was conducted between November 12 and December 9, 2024, prior to the first round of elections, and 201 people took part. The second survey took place between January 21 and February 2, 2025, with the participation of 64 citizens. The surveys respond to non-probabilistic samples due to the availability of the participants, and because it optimises time “according to the specific circumstances surrounding both the researcher and the subjects or groups under investigation” (Sandoval, 2002, p. 124). The distribution mechanism of the surveys was the “snowball” dynamic, applied through Google Forms. The data were processed in SPSS statistical software, version 22.

The participants in the first survey are, by geographical region: 51 personnel from the coast, 6 residing outside the country, 13 from the Amazon and 131 from the highlands. According to sector, 25 live in rural areas and 176 in cities. According to sex, 128 women and 73 men. By age group, 115 people between 18 and 27 years old; 46 between 28 and 37 years old; 30 between 38 and 47 years old; 4 between 48 and 57 years old; and 6 people over 58 years old. The average age is 29 years. According to occupations: 102 works in the private sector; 20 works in the public sector; and 79 are in other occupations (students, retired, housewives, etc.).

The contributors to the second survey are 30 men and 34 women (64 participants). According to occupations: 10 private civil servants, 24 civil servants, 23 entrepreneurs, 7 freelance professionals. According to region of residence: 45 live in highland cities, 15 in coastal cities, 4 in Amazonian cities. All of them were adults. This survey received free responses because three open questions were proposed: (1) Do you think that the use of artificial intelligence in the creation and dissemination of content during electoral campaigns affects your ability to be informed in an impartial manner? (2) Do you think that artificial intelligence can be used to manipulate opinions during elections? (3) What mechanisms do you think should be implemented to regulate the use of artificial intelligence in electoral processes and avoid disinformation? Perhaps a regulation on the use of AI in campaigns proposed by the National Electoral Council (CNE)?

Sentiment analysis is a research technique for classifying the polarity or emotion expressed in large volumes of unstructured text. Its purpose is to convert qualitative data, such as responses to open-ended survey questions, into quantitative information. In this study, 64 responses were analysed manually due to the number of cases in order to understand the nuances of each response. In addition, this modality provides information on words, phrases, or entire documents in broad contexts (Rumaisa et al., 2019).

Both surveys were conducted in strict compliance with the ethical principles established by the Declaration of Helsinki, ensuring the protection of the rights and welfare of respondents. Citizens’ participation was completely voluntary and the data provided were treated anonymously. The questionnaire was constructed based on an adaptation of the survey “Americans think AI will harm elections” from Morning Consults. National Tracking Poll, and from the study by Garriga et al. (2024) which considers checks by anti-disinformation agencies.

On the other hand, semi-structured interviews conducted before the elections were carried out with five professionals: four men and one woman, including experts in communication, law, academic research, politics, and an AI consultant. These interviews were conducted between 26 and 28 February 2025 via the Zoom videoconferencing platform. The interviews after the vote, with nine experts, took place from 22 to 26 July 2025, also via videoconference.

The profiles of the interviewees are: Interviewee 6: lecturer and researcher at Simón Bolívar Andean University, Ecuador campus. Independent electoral advisor. Areas of expertise: electoral law, digital governance, and ethics in technology. Interviewee 7: lecturer at the Central University of Ecuador. Interviewee 8: former electoral coordinator for the National Electoral Council in Cuenca. Interviewee 9: lawyer, former member of the Provincial Assembly of Loja. Interviewee 10: founder of a local political movement in Azuay. Areas of expertise: electoral law and public policy. Interviewee 11: councillor for the Municipality of Playas. Interviewee 12: secondary school teacher. Interviewee 13: lawyer, official of the National Electoral Council in Riobamba. Interviewee 14: manager of Computer Forensics.

The qualitative analysis of the responses from the semi-structured interviews was carried out using thematic coding in accordance with the approach of Braun and Clarke (2006). The process included three phases: first, open coding was performed to identify emerging units of meaning in the transcripts; second, axial coding was performed to bring together the initial codes into thematic categories related to perceptions of artificial intelligence, institutional trust, regulation, and electoral manipulation; and third, selective coding was performed to integrate these categories into a narrative consistent with the research objectives. Two people collaborated by independently coding 30% of the corpus to establish inter-coder reliability, resulting in a Cohen’s Kappa coefficient of 0.82, which implies substantial agreement. Discrepancies were resolved by consensus.

The criterion of theoretical saturation was verified when the last three interviews, after the elections (interviewees 12, 13, and 14), did not contribute any new codes or categories with respect to the dimensions of analysis identified previously. Therefore, based on the findings of Guest et al. (2006), the collection of additional data does not alter the understanding of the phenomenon under study. This ensured systematicity and rigour in the interpretation of the qualitative data.

During the writing of this paper, Qwen3-Max, a generative artificial intelligence tool, was used to organise the interviewees’ testimonies and structure the sentiment analysis. After using the aforementioned tool, the content was reviewed and edited as necessary. The author assumes responsibility for the text.

The results of the first survey are presented in Tables 2, 3.

Pearson’s chi-square test allows us to assess the independence between compliance with the regulation of the use of artificial intelligence (AI) in the election campaign and the perceived influence of AI-created disinformation on the outcome of Ecuador’s presidential election in 2025. The result is a value of (χ²) of 9.298, with 3 degrees of freedom (gl = 3). The bilateral asymptotic significance is 0.26 (p = 0.26). Cramer’s V coefficient is 0.215, with an approximate significance of 0.26 (p = 0.26). The value of the chi-square statistic does not reach statistical significance (p ≤ 0.05); therefore, no statistically significant association was found between the variables.

Pearson’s chi-square test allowed us to determine the independence between the use of artificial intelligence (AI) in the election campaign and the level of trust in the candidates’ advertisements using AI as a support. The chi-square value (χ²) of 27.095 with 6 degrees of freedom (gl = 6). Bilateral asymptotic significance is less than 0.001 (p < 0.001). Exceeding the conventional statistical significance threshold (p ≤ 0.05), indicates the existence of a statistically significant association. The magnitude of the observed association, Cramer’s V coefficient, is 0.26 with an approximate significance of less than 0.001 (p < 0.001), that is, the relationship between the variables is of modest magnitude.

Also, a Pearson’s chi-square test was conducted between the use of artificial intelligence (AI) in the election campaign and the level of confidence in the election result. The chi-square value (χ²) is 18.833 with 6 degrees of freedom (gl = 6). The bilateral asymptotic significance is 0.04 (p = 0.04), indicating a statistically significant association. The Cramer’s V coefficient value is 0.216 with an approximate significance of 0.04 (p = 0.04), therefore, there is a relationship, albeit modest in its strength, between the variables.

In the results of the second survey, what is common is the recognition of the manipulative potential of AI, the need to establish regulations for its use in political campaigns, and the demand for greater media literacy to mitigate the risks of misinformation. Participants consider that it is the responsibility of citizens to verify information and check it against reliable sources before sharing it or assuming it to be true, and they recognise that they should proceed in a responsible and ethical manner on social media.

The responses to the first question of the second survey are organised as follows: out of 64 participants, 31 (48%) estimate that the use of AI in the creation and dissemination of content during electoral campaigns does affect the ability to be informed; 21 (33%) believe it will partially affect; and 12 (19%) believe it will not affect. The use of AI in election campaigning in Ecuador raises concerns because it is used to manipulate information and spread fake news, can generate bias by creating content that is difficult to verify, which increases the risk of misinformation, undermines trust in democratic institutions, leads to voter apathy, reduces voter participation and weakens democracy.

Among the arguments of people who consider that there is affectation, it is mentioned that AI “today helps to lie, put you in places you have never visited, shows works you have never done and the worst, they invent news at convenience” (respondent 55). “It is difficult to differentiate content created from AI, it makes my opinion about candidates not entirely clear” (respondent 22). Therefore, AI “impacts the way people access information during an election campaign. It can help citizens get information faster and more accessible. However, it also carries risks, such as the creation of biased content” (respondent 8).

AI “leads to digital platforms personalising content according to the preferences of each user. This is a limitation because, by only being informed of things of interest to us, new proposals and new candidates are not opened up” (respondent 64), so “I could be receiving only one type of message, which does not allow me to know all the options and points of view” (respondent 49). Therefore, “AI can be used by opposing parties to deceive citizens with audios, recordings or messages that candidates have not said” (respondent 57), and “when information is repeated, people tend to believe it is true” (respondent 35), “all false information that comes out of AI is reflected in social networks and lies that are repeated 1,000 times become truth” (respondent 60).

Specifically, AI “articulates algorithms that can be programmed by a political tendency to benefit its candidate” (respondent 4), which implies that it “affects impartiality in election campaigns by limiting diversity of opinion and amplifying misinformation” (respondent 18), and “although it facilitates access to information, its use in election campaigns poses significant challenges to the impartiality and quality of information available to voters” (respondent 61).

People who believe that the use of AI during election campaigns does not affect their ability to be impartially informed point out that “we are aware of technology and there are several ways to inform ourselves” (respondent 7), “there are accurate media that would allow me to be impartially informed” (respondent 24). “AI is an information tool, like social media, where not everything is truthful, and impartiality is going to depend on each person’s judgement, considering the stance of the candidates” (respondent 41), “I am clear on the conditions and circumstances of such campaigns at the time” (respondent 27). “The content we see depends on each person and their context, that is, many young people will be informed through social networks, while others will use traditional media, but it does not depend on AI” (E44). One person even said “I am not so attached to technology, I keep myself informed thanks to news and radio opinions, also, my daughters and granddaughters talk about it, so it does not affect me” (E37).

For the second question, from the second survey, of 64 participants, 55 (86%) believe that AI can be used to manipulate opinions during elections; 5 (8%) think that partial manipulation would occur; and, for 4 (6%) respondents there would be no manipulation. AI “does not allow us to be well informed” (respondent 19), “it is used to manipulate people’s decisions by spreading false information and also by creating videos to reach a large part of the population to create confusion at election time” (respondent 23). “There are people who find it difficult to detect what information is real and what is created by AI, this causes us to decide again who deserves to vote” (respondent 29), “people who are not trained are confused because they do not know what it is about, and in general it impacts the citizenry” (respondent 53).

There is “a lot of dirty campaigning that creates rivalry between candidates and society at general” (respondent 10), “supporters or list members themselves can use it in a negative way against other political parties” (respondent 13), “unfortunately people do not contrast the content they consume, and manipulation existed before there was AI” (E30). “AI can be manipulated at anyone’s convenience” (respondent 39), “content can be created that makes a candidate look better or worse than they really are, and influence how we think. Likewise, spreading rumours that change the way we see candidates, and that’s not fair” (respondent 49).

Unfortunately, “in Ecuador we do not look for fair competition, so AI can be misleading” (respondent 64). “Voters without knowledge believe in false influences” (respondent 35), on this basis it is requested “to also regulate the use of social media” (respondent 6), since “many people consult the Internet for information” (respondent 16). In the extreme, “technology without legal, ethical and moral control has no limitations, it can replace human beings” (respondent 35). Among those who believe that AI cannot be used to manipulate opinions, it is argued that ‘it is simply a tool to generate publicity’ (respondent 44). “Opinion is public but it is also individual. AI can possibly skew data, misinform in a way, but it does not in itself affect public opinion” (respondent 43).

In relation to question 3, what mechanisms do you think should be implemented to regulate the use of artificial intelligence in electoral processes and prevent misinformation? Respondents indicated that regulation is necessary to prevent abuses, but it should not restrict innovation and freedom of expression, and there is an urgent need to find a balance between regulation and people’s rights. It was indicated that:

It was revealed that it is important to work with social media platforms and other technology companies to combat misinformation and promote the responsible use of AI, and to articulate actions with international entities to ensure the transparency of the electoral process. “It should be limited to the amount of AI that campaigns use and inform citizens so that they are aware and do not fall prey to malicious information” (respondent 22). “We cannot let AI consume us and render us useless, so a regulation and filters for its use would be a good option” (respondent 37). They also call for the continued invitation of independent bodies to oversee the use of AI in election campaigns. These committees would be made up of experts in technology and ethics, and would be in charge of monitoring digital content and pointing out possible irregularities. A minority position among participants is to ban AI in election campaigns to prevent manipulation.

Most agree with regulating the use of AI in electoral processes, “it would be excellent to have a regulation that prohibits the use of social networks and imposes fines on political parties, because they are the ones that allow AI to do disinformation” (respondent 55). “Like all technologies or electronic devices, AI should have regulations and conditions of use. It does not matter if it is for presidential elections, the important thing is that the rules are respected and avoid fake news” (respondent 64). “Implementing mechanisms to regulate the use of AI in electoral processes is crucial to avoid misinformation, for example, a regulation on transparency in electoral advertising” (respondent 61). “Competent authorities should create a law prohibiting AI in elections because most political parties use it in a bad way, to misinform citizens” (respondent 33). Mechanisms such as “codes of conduct, spaces for dialogue and evaluation” (respondent 5), a regulation, monitoring of content, obliging political parties and candidates to disclose its use in their campaigns, implementing systems to check the veracity of information, imposing labels or identifiers on content so that citizens know when they are interacting with this type of information are suggested. To enforce compliance with the regulation, the creation of legal sanctions is proposed.

For citizens, the National Electoral Council (CNE/ Consejo Nacional Electoral) is the institution that should lead the regulation of AI in elections; it is the entity with the authority to guarantee ethical use. “The CNE should be in charge of regulating the use of this new technology to avoid confusion and false information during elections” (respondent 23). “The CNE should have established rules from the beginning of the campaign to know if something was elaborated with AI” (respondent 41), although “the right to access ICTs is already established [Art. 16 of the Constitution], a regulation is needed, but not from the CNE but from another instance that allows verifying what is created with AI and the contribution to citizenship” (respondent 43). But, “first, we must encourage “healthy politics,” where lies and fallacies do not damage a politician’s image, and government institutions must improve their systems against bot’s and people who seek to misuse information” (respondent 44).

These results coincide with the analysis of emotions in the opinions of the 64 respondents on the use of AI in the creation and dissemination of content during election campaigns. There is significant concern about the negative impact it may have on the impartiality of information, the manipulation of opinions and freedom of expression. Emotional demographic patterns are identified and shown in Table 4.

The predominant emotions are:

The general attitude towards the use of AI, according to respondents, is negative for 55% of participants (35 people), ambivalent for 25% (26 people), and proactive for 20% (13 people). On the other hand, the most recurring words (in order of frequency) are: AI / artificial intelligence (64); false / false (45); manipulate (40); confusion / confuses (35); regulation / regulation (30); disinformation (30); truth/truthful (25); fear/fearful (25); transparency (20); freedom (20); education/literacy (18); social media (15); CNE (10).

The prevailing sentiment is negative, with shades of hope. Negative emotions predominate, especially concern and mistrust regarding AI’s potential to manipulate information and create disinformation. Neutrality and optimism are less frequent, and there is consensus on the need for regulation. Most perceive AI as a threat to democracy, impartiality of information, and freedom of expression. There is concern about disinformation, emotional manipulation and lack of regulation. However, there is also public demand for regulation, transparency and education, reflecting a critical awareness and a desire to participate in protecting the electoral process.

The experts interviewed before the elections agree with the citizens surveyed. AI in election campaigning “has three edges, one as a technical tool, one as a disinformation-generating tool and one as political discourse” (interviewee 1), but “we are early in the regulation of AI. We have some catching up to do and there is some way to go. Legislation could be passed that would require campaigns, when using AI-created images, to have water licenses or a type of message that specifies that they were created by AI” (interviewee 5). Regulating “is very difficult because it does not depend on the legislation itself the use of platforms (Facebook, X, YouTube), it is a content that is difficult to regulate, maybe with regional bodies to help” (interviewee 3). “How do you regulate users who have no nationality, who are spread around the world? This generates a crisis in state systems, today there are new owners, which are the companies (platforms, social networks)” (interviewee 4). However, there are international experiences that show progress, for example, “in the United Kingdom, political parties are required to publish their financing and advertisements on social networks” (interviewee 3), in Brazil “when information was proven to be false or disinformation, the electoral body notified the platform, and they had a time limit to block or remove the content” (interviewee 1).

In Ecuador, in “the first round debate, the most important character was AI since all the candidates mentioned it, as an idea, as if to give a message of innovation, which is gradually being prostituted” (interviewee 1), and although the national electoral body “with the support of international cooperation, has shown what is happening, that AI is being misused, that candidates are misusing AI to attack or denigrate their opponents” (interviewee 2), it is not being regulated because “there is a legal loophole in the Democracy Code” (interviewee 3), “generative AI brings new actors and scenarios” (interviewee 4). To control social media and the use of AI “would require a legal framework that establishes what the limits of these publications would be, what content and who could do it so that there is an authority to regulate” (interviewee 2).

Part of “the solution is to generate a culture of verification. We need citizens who are verifiers, who learn that the basis is education, who learn to contrast information” (interviewee 1). There is “a need for training processes in digital literacy, training in respect for the information that is consumed. The problem is not AI, but how people, human intelligence, use this tool” (interviewee 2).

Experts interviewed after the elections point out that Ecuador’s current regulatory framework is not prepared to face the challenges of AI-generated disinformation in electoral contexts. Existing laws do not address new forms of information manipulation, so they had limited effectiveness in mitigating disinformation in the 2025 electoral process. The legislation was not up to the technological challenge. Some interviewees (6, 9, 10, 12) highlighted the absence of rules requiring transparency in the use of AI in electoral content, and the lack of specific regulations governing the use of emerging technologies.

It was stated that any new regulations must be flexible in order to evolve at the same pace as technology (interviewees 6, 7, 9, 10, 12, 13). Several experts (6, 9, 10, 11, 12, 13) propose mandatory labelling of AI content as a principle of transparency, as well as the inclusion of clear penalties for those who spread disinformation and legal liability for automated content (6, 13).

In addition, weaknesses were identified in the CNE’s ability to monitor and sanction AI-generated disinformation (interviewees 6, 7, 9, 10, 13), partly due to the lack of a specialised technical body. The urgency of legislative reform was highlighted. The CNE authorities did not have effective legal mechanisms to request the removal of false content and sanction those responsible. Their response to the rapid spread of disinformation was slow.

The interviewees agree on the importance of civic education and digital literacy as strategies to empower voters. Regulation alone is not considered sufficient. It is suggested that digital literacy should be a state policy integrated into the education system. Voter empowerment is seen as an effective way to reduce the impact of manipulation and strengthen informed participation (interviewees 6, 8, 9, 11, 14).

Based on respondents’ answers (Tables 2, 3) and expert testimonies, Ecuadorians, (1) do agree with regulating the use of generative AI in electoral campaigns, but the mechanisms and model still need to be discussed, and suggest that the Consejo Nacional Electoral is the body called upon to propose a regulation on the use of AI for future campaigns; (2) they perceive that disinformation created through generative AI influenced who won Ecuador’s presidential election in 2025; and, they consider that the use of AI in Ecuador’s presidential campaign in 2025 did not increase trust in the candidates and the outcome of the election of the President of the Republic.

A considerable number of respondents highlighted that the spread of AI-created fake news affects the ability to make informed decisions. Many people believe fake news, do not know or cannot verify it, so misinformation grows and becomes collective. It was also mentioned that the viralisation of manipulated content can affect voters’ emotions and decisions.

Without a proper legal framework, AI will continue to be used to influence voters in unethical ways, but the solution is not to restrict it, but to regulate its application in the dissemination of political information. The urgency was expressed for the National Electoral Council, as well as other government institutions, to implement media literacy programmes so that citizens can identify false content and verify sources. One of the biggest challenges in political campaigns is to distinguish between real and artificial content.

For the citizens surveyed, as for the experts interviewed, CNE is called upon to lead the debate and approval of a norm that regulates the use of AI tools to avoid disinformation in future electoral campaigns in Ecuador. Implicitly, this revels acceptance of self-regulation that should consider the participation of government, civil society, academia and the private sector, and collaboration with technological platforms to improve the detection and elimination of fake news and manipulated content. Following international experiences, self-regulation shows the relevance of national consensus to hold platforms hosting fake messages accountable. The standard to be approved could lead to a regulation that includes transparency measures, limits on the personalisation of political messages, sanctions for the misuse of AI and the clear identification of content generated with this technology. The regulation of AI in politics is an urgent challenge to ensure transparency and equity in access to information.

The testimonies of the experts interviewed after the elections support the above. They identify the absence of specific regulations requiring transparency in the use of AI in the creation of electoral content, the lack of clear protocols for verifying automated content, and the institutional weakness of the CNE in detecting and sanctioning disinformation in a timely manner. During the 2025 elections, the effectiveness of the current mechanisms was, in practice, nil; documented cases of deepfakes, synthetic audio, and massive disinformation campaigns were found to have circulated without control by the electoral authorities.

Respondents also agree that AI does not limit freedom of expression, but does so indirectly by distorting the flow of truthful information, which means disrupting the quality of public debate. AI creates a false sense of plurality, yet users are managed by algorithms. The use of fake accounts to affect conversations and reinforce ideological biases was highlighted. Likewise, the use of fake images, videos and audios in the election campaign was also noted. The data in Table 3 is consistent with the Digital Users report (Table 1), respondents’ perceptions that AI is mostly used to produce false messages on social media through videos, photos and advertisements actually occurred.

Although a minority, some responses suggest that AI does not manipulate on its own, but that its use depends on the intentions of those who program it. Some respondents mention that AI, properly applied, could democratise access to information. From this perspective, impartiality does not depend on the technology, but on the judgement of citizens when evaluating information. Among the alternatives to reduce the disinformation generated by AI in the electoral campaign, also in line with the literature suggests, educational processes of media competences and the promotion of a culture of verification are proposed.

As a result of sentiment analysis, it is clear that Ecuadorian citizens view the use of AI in election campaigns with scepticism and caution, fearing its potential to manipulate and misinform. Beyond the technical aspects, the emotional perception of AI in politics is marked by a fear of losing control over truth and democracy. However, citizens are active, critical, and willing to collaborate in building a fairer and more transparent electoral environment. Despite their fears, they demand regulation, transparency and digital education, as reflected in the frequency of words such as regulation, CNE, truth and education. AI is seen as a double-edged sword: useful, but dangerous if not regulated.

As a result of the quantitative analysis, it was identified that, (1) there is a statistically significant association between the use of AI in the election campaign and the level of trust in the candidates’ advertisements, and (2) there is a significant association between the use of AI in the election campaign and trust in the election results, but both relationships are weak and further analysis is needed to identify the specific nature of these associations. In another context, there was no evidence of an association between compliance with AI regulation in electoral campaigns and perceptions of the outcome of the presidential election. Ecuadorians, although they believe that AI misinformation will affect the presidential election results, are not yet overwhelmingly convinced of the urgency of the regulation.

This work serves as a basis for future research into how AI-driven disinformation can affect the democratic process in Ecuador. The fulfilment of government plans, accountability reports and the treatment of draft laws that aim to regulate AI could be analysed, advocating for an approach that protects human rights and addresses the challenges posed by emerging technologies. It also remains to be seen how disinformation impacts the growing integration of AI into Ecuadorian and regional politics.

These results add to recent research on electoral processes in Latin America, which shows that AI-generated disinformation can have an impact. This technology was used to manipulate public opinion, influence voters and undermine the quality of democratic processes. Misinformation in the elections in Argentina, Chile, Colombia, and Mexico between 2018 and 2023 focused on discrediting candidates and questioning the legitimacy of the processes (Hernández and Guzmán, 2025). Evidence was found that AI accelerated the production and dissemination of misinformation, especially through social media platforms (Ferrara, 2025; Pashentsev and Bazarkina, 2023). Faced with this risk, countries such as Colombia are exploring the application of AI in political campaigns to propose its regulation and thus reduce misinformation (López et al., 2024).