Misleading Discourse on Instagram : A Multimodal Study of Latin American Presidential Candidates in the Face of COVID-19

Instagram as a multimodal information network has helped politicians to position both their brand and their campaign. We analyzed whether the images and texts published during the pandemic contained misinformation. We studied from a multimodal perspective the Instagram accounts of the presidential candidates of four Latin American countries which held elections in 2021 to identify how much of the discourse was related to controlling the pandemic. The discourse was analyzed using different taxonomies. In the correlation between the discourse and following the recommendations of the World Health Organization (WHO), Chile stood out with the highest level of pandemic compliance; Peru and Ecuador were placed in the middle, while Honduras showed little if any interest. The conclusion was that politicians focused primarily on their campaigns Marga Cabrera; Rodrigo Cisternas; Alberto J. López; Rebeca Díez 28 Anàlisi 64, 2021 Marga Cabrera; Rodrigo Cisternas; Alberto J. López; Rebeca Díez and marginally on the pandemic. The omission of COVID-19 from most publications reflected a misinformative discourse which could potentially confuse the public.


Introduction
The recent health crisis favored the use of a presidential style supported by social networks to communicate decisions, gauge the environment, and validate policies (Manfredi-Sánchez, Amado-Suárez and Waisbord, 2021), but what about presidential candidates in particular? What was their stance in the midst of campaigning? Social networks lend themselves to populist practices (Moffitt and Tormey, 2014;Bevelander and Wodak, 2019), although pandemic-related information was not always honest and consistent on the Instagram accounts of candidates who sometimes posted inaccurate news reports or items tailored to a personal goal.
Instagram was created as a multimodal information network, i.e., audiovisual and textual publications. Due to its popularity, public figures, including politicians, resorted to Instagram as a communication tool. We considered it imperative to analyze the symbolic elites of this multimodal channel where they were able to present their discourse, often with a misinformative result (Van Dijk, 2006).
Other research in political communication has also focused on Instagram (Sampietro and Sánchez-Castillo, 2020) and there have been numerous studies on the personal content of political candidates on this network, in both electoral and post-electoral periods (Lalancette and Raynauld, 2017;Larsson, 2017;Muñoz and Towner, 2017;Quevedo-Redondo and Portalés-Oliva, 2017). Campaigns based on social networks resulted from new thoughts on political communication and the management of the personal and public images of politicians in digital environments (Lees-Marshment, 2001;Paré and Berger, 2008;Marland, 2012). Instagram constituted one of the best platforms for a communicative model based on the denotative power of photographs, on the emergence of a conversation type based on portraits, selfies, and videos (Quevedo-Redondo and Portalés-Oliva, 2017), and on the combination of photography and politics (Ruiz-del-Olmo and Bustos-Díaz, 2016;López-Rabadán, López-Meri and Doménech-Fabregat, 2016).

Instagram and COVID-19 in Latin America
This research focused on misinformation in the discourse of political candidates on Instagram regarding COVID-19. A multimodal analysis was conducted of the graphic and written discourse of the main presidential candidates in a collection of Latin American countries which held presidential elections in 2021, Chile, Ecuador, Honduras, and Peru. The number of Instagram users was 7.3M in Chile, 5.2M in Ecuador where it was the fourth most used social network, 5.16M in Peru, and 0.8M in Honduras (Librero, 2020), Instagram being one of the four most-used networks in each of the four countries analyzed.
Politicians were able to use the content of their networks to promote public participation in the prevention of and the fight against the pandemic (Niknam et al., 2020;Castillo-Esparcia, Fernández-Souto and Puentes-Rivera, 2020). Research, such as a study of the use of Twitter as a presidential communication channel in the initial period of COVID-19 (Manfredi-Sánchez et al., 2021), indicated Presidents Sánchez and Bolsonaro deployed a health management communication strategy, while López Obrador and Fernández paid little attention to health policy. A study of the Instagram account of the former vice president of Ecuador, Otto Sonnenholzner, during the first emergency phase resulting from the pandemic, showed this politician used his account mainly to promote his personal image rather than having objectives linked to the national health emergency, which echoed political promotion (Lozano-Recalde, 2021).
In 2020 in Latin America, Peru had the highest number of deaths due to COVID-19 across the entire population at 118.6 per 100,000 inhabitants, while in Chile the figure was 89.8, in Ecuador 82.8, and in Honduras 38.0 (Pasquali, 2021).
From the initial declaration of the pandemic, the World Health Organization (WHO) continued issuing updated information on the disease; however, there were indications and recommendations consistent over time that were assumed by governments, national and international organizations, and the media, e.g., wash and disinfect hands frequently, cover your mouth when coughing, refrain from touching your nose or mouth, ventilate rooms well, limit stays in closed and poorly ventilated spaces, maintain respiratory hygiene, and avoid direct physical contact with others. The two measures of most interest in this study were social distancing and the use of masks, as they could both be analyzed by means of images of the candidates.

Misinformation in Multimodal Discourse
We characterized misinformation as ambiguous or vague information that can generate confusion and mistrust in the receiver, can make it difficult to use the information, and can cast doubts in receivers on processing actions and decision-making (Karlova and Fisher, 2013). Discussion surrounding the terms disinformation and misinformation suffered from imprecision in their definitions (Karlova and Lee, 2011). Sometimes the terms were used interchangeably as though there was no distinction between them (Fox, 1983;Losee, 1997), or one was used as a variation of the other (Zhou and Zhang, 2007). Buckland (1991) defined misinformation as accidental or innocent, while disinformation can reveal the malicious intent of the speaker, if the receiver was aware of being misled. According to Bednar and Welch (2008) and Stahl (2006), the difference between disinformation and misinformation depended on the intention of the speaker. Misinformation occurred when the information provided was not correct but the intention of the deception was benevolent, while disinformation presupposes a malevolent intention on the part of the issuer. However, the intentionality of the speaker may be ambiguous and/ or unknown, and it was not therefore satisfactory to differentiate the two terms solely on the basis of this criterion (Karlova and Lee, 2011).
Authors such as Losee (1997) and Zhou et al. (2004) defined one of the aspects of misinformation as incomplete, distorted, or ambivalent information. In addition, Zhou and Zhang (2007) added various types of misinformation, such as concealment, ambivalence, distortion, and falsification. However, these authors did not conduct any disambiguation between the two terms misinformation and disinformation. However, Fallis (2009) concluded that misinformation may or may not be inaccurate, but must be misleading, further adding that the misleadingness of misinformation may be due to the context of the situation. This approach supported the constructivist and subjective view of information of Hjørland (2007). Thus, disinformation is not a proper subset of inaccurate information (which would be misinformation). For their part, Karlova and Lee (2011) added that misinformation could be inaccurate, uncertain (e.g., by presenting more than one possibility or choice), vague (unclear), or ambiguous (open to multiple interpretations). Information that is incomplete can also be a form of deception, often referred to as misinformation.
As discussed above, misinformation is a multifaceted concept, more complex than simply incomplete information (Karlova and Lee, 2011). Karlova and Fisher (2013) attributed much of the problem of misinformation to its ability to create confusion and mistrust among recipients, making it difficult to use the information.
The analyzed discourse was approached from a multimodal perspective, seeking to detect a process of misinformation in the graphic and textual discourse of the candidates on Instagram. Multimodal Discourse Analysis (MDA) is based on the systemic-functional linguistics of Halliday (1978). This approach understood language as being embedded in social semiotics (van Leeuwen, 2005), where meaning systems are shaped by their contextual inscription.
The multimodal perspective of discourse analysis asserts that communication is inherently multimodal. MDA analyzes the multiple modalities (language, image, or audio) that combine to create meaning in different contexts (O'Halloran et al., 2014). Various modalities beyond language come into play when creating discourse (O'Halloran, 2012). These modalities are also called semiotic resources (Parodi, 2010). Each semiotic resource is understood as a system of options that interact with each other, creating a discursive semantic unit (Menéndez, 2012).
The objective of this research was to study the multimodal narrative of candidates and to determine whether it carried implicit misinformation in the coherence of the coronavirus prevention discourse with respect to the images and text projected by candidates on their Instagram accounts. Analyzing the content of Instagram posts can help identify the thoughts and feelings of the candidates in the face of the health crisis (Niknam et al, 2020).

Materials and Methods
This descriptive research delved into content analysis based on the monitoring of Instagram posts of the candidates most likely to be elected in four Latin American countries where presidential elections were held in 2021, Peru, Chile, Ecuador, and Honduras. The unit of analysis comprised images uploaded to Instagram along with accompanying captions. Stories, videos, or different carousels of images that accompany a publication were not included, only the first image that appeared on the post being taken into account.
The hypothesis put forward was that there was no consistency in the coronavirus prevention discourse with respect to the images and captions projected by the presidential candidates on their Instagram accounts.
We analyzed the candidates most likely to be elected according to polls, the parties to which they belonged, their Instagram accounts, and the num-ber of followers starting from the date the pandemic was declared, March 11, 2020, through to December 31, 2020. Table 1 below summarizes the relevant candidate information. An analysis sheet was constructed as an instrument for data collection on which all posts were recorded for each country and candidate from the date of the pandemic declaration until December 31, 2020. A total of 3,099 posts were analyzed including their images and the accompanying text.
The sheet identified four factors where the first refers to the relevance of the candidate on the Instagram social network, while the other three refer to the prevention of COVID-19 based on WHO recommendations: 1) Candidate Relevance Data was collected on the age of the account, the number of followers, and the start of activity by the candidate. A count was made of the total number of posts, the total number of posts during the period analyzed, and the number of posts in which the candidate themselves appear.
2) Prevention: Mask Use Analysis of the total number of posts where the candidate appeared in the period analyzed, counting the number of times they appeared alone and if they wore a mask, whether or not accompanied.

3) Prevention: Social Distancing
When the candidate appeared accompanied, we evaluated whether they wore a mask and if they complied with the recommendations on social distancing issued by the WHO. The analysis took into account whether the group comprised relatives (cohabitants) of the candidate and if it was in an open or closed space.

4) Political Discourse in the Face of COVID-19
Text and images were not combined simply by the addition or intersection of the component meanings. Bateman (2014) introduced the term meaning multiplication to refer to the creation of new meaning through the integration of images and text. Marsh and Domas (2003) created a taxonomy of text-image relationships to analyze how images and text interact. The authors developed 49 image-text relationships with the intention of developing a broad taxonomy applicable to all areas and types of documents, thereby generating a common language for the classification of image-text intent. We used the taxonomy developed by Kruk et al. (2019), based on the work of Bateman (2014) and Marsh and Domas (2003), to categorize the image-text meaning of Instagram posts. Kruk et al. (2019) analyzed a set of 1,299 Instagram posts to validate their taxonomy. The authors created three major taxonomies to categorize the sum of images and text on Instagram and emphasized the need for further research to expand and enrich the proposed classification. One of the taxonomies focused on speaker intent, while the other two (contextual and semiotic) captured different aspects of the relationship between image and caption.
A. Taxonomy of Speaker Intent: The proposal of Kruk et al. (2019) on speaker intention was based on the illocutionary acts of Austin (1962) that refer to the intention of the message and the meaning of all the communicative elements involved. In other words, they refer to the final objective of the speaker when using the various modes of communication. After analyzing and grouping representative Instagram content, they created the following 8 categories designed to categorize the intention of the author on Instagram: 1. Advocative: Defend an idea, a movement. 2. Promotive: Promote events, products, organizations. 3. Exhibitionist: Create a self-image reflecting the person, status, etc. 4. Expressive: Express emotion, adhesion, or admiration towards an external entity or group. 5. Informative: Convey information about a topic or event using factual language. 6. Entertainment: Entertain through humor, art, memes. 7. Provocative/discriminatory: Direct attack on a subject or group. 8. Provocative/controversial: Shock.

B. Contextual Taxonomy
For the contextual taxonomy, the relationship between the literal meaning of the image and the text was classified. The 49 image-text relationships identified by Marsh and Domas (2003) were grouped into three superordinate categories, which distinguished images as minimally related to text, highly related to text, or related but going beyond text. According to Kruk et al. (2019), this classification frames the image only as subordinate to the text. The authors adapted the three major contextual relationships to make them symmetrical on Instagram: 1. Minimal relationship: The relationship between image and text is literal. The meaning between the two overlaps very little. 2. Close relationship: The literal meaning of the text and the image overlap considerably. 3. Transcendent relationship: The literal meaning of one of the modalities gathers and expands the meaning of the other.
Regarding this taxonomy, it is important to note that the categories "minimal" and "close" were placed on a continuous scale indicating a semantic overlap, while the category "transcendent" indicated an expansion of meaning which cannot be captured on a continuous scale.
C. Semiotic Taxonomy Semiotic taxonomy captures the relationship between what is signified by its respective modality and its semiotics. The authors started with the distinctions made by Kloepfer (1977), modeled by Bateman (2014), and the parallel and non-parallel distinctions of Zhang, Hwa and Kovashka (2018), and composed three categories: Regarding this taxonomy, it is important to clarify that the semiotic classification is not always homologous with the contextual classification.

Results
The taxonomies described above made up one of the factors analyzed, "Political discourse: COVID-19 situation". The other factors were comprised of those referring to COVID-19 prevention according to the standards dictated by WHO (use of a mask, and social distancing), except the first one which analyzed the relevance of the candidates on Instagram. These factors were developed ad hoc in order to integrate them with the taxonomies of Kruk et al. (2019) (figure 1) and to incorporate them (Cárcamo-Morales, 2018) into the misinformation study. On presenting the results, we followed Bateman (2014) and employed tables (<http://bit.ly/3r3YCNW>), which were presented as a key technique when performing compositional cohesion analysis (Cárcamo-Morales, 2018), in this case of image and text modalities. Tables 2 through 5 below present the data analyzed for each country.

Relevance
Despite the ongoing scourge of the pandemic, the candidates analyzed showed inconsistent and lopsided behavior with regards to following WHO recommendations. The effects of COVID-19, such as information on infection rates or prevention, did not stand out in their Instagram posts, despite it being a pre-electoral period. In most cases they were young, recently created accounts, as was the case of the candidates in Honduras. They appeared to have been opened for the campaign and none of the three showed their more personal side. Rather, they gave the impression of being managed by the respective communication teams of the candidates. Older and more personal accounts, through which it was possible to find out about the interests, hobbies, and families of the candidates, belonged to the candidates of Ecuador, Peru, and Chile. In each country, one candidate stood out for their activity on Instagram: Lasso in Ecuador, Jiles in Chile, Forsyth in Peru, and Rosenthal in Honduras. However, despite being the most active candidates, none stood out for their content related to COVID-19. Lasso dedicated 18.7% of his posts, Jiles 7.1%, Forsyth 12.8%, and Rosenthal did not dedicate a single line to the pandemic. Castro (Honduras), De Soto (Peru), and Arauz (Ecuador) with 0.8%, presented total or near total silence on the virus. The candidates who spoke the most about the pandemic and preventive measures were Pérez in Ecuador (22.4%), Lavin in Chile as prefect and mayor (33.3%), Mendoza in Peru (23.3%), and minimally Asfura in Honduras as mayor of the Central District (1.4%).

Prevention
WHO advised preventive behaviors and measures focused on the use of masks and social distancing, and above all avoiding large crowds at public events.
In Ecuador, the candidate who most adopted the recommendation to wear a mask was Arauz, who wore one in 60.5% of the posts analyzed, especially when accompanied (97.1%), despite not taking too much care with social distancing (28.0%). Arauz was followed by Pérez, who wore a mask in 45.8% of his posts, although not always when accompanied (he only wore one on 32.1% of these occasions), but he was more careful with social distancing (66.2%). Lasso was the candidate who most neglected the measures and recommendations of WHO, given that he only wore a mask in 25.8% of the posts, rarely when surrounded by people (25.8%), and he was careless with social distancing (33.3%).
The candidates in Chile followed the recommendations inconsistently. In the use of masks, both Jiles and Lavín showed a similar rate of use (64.9% and 63.1%, respectively), in addition to maintaining social distancing on more than 60.0% of occasions. Regarding the non-use of a mask with peo-ple, Lavín stood out with only 14.0% of the analyzed sample. However, of that 14.0% it would be possible to subtract 5.8% of the publications which were made in a family environment.
In Peru, Forsyth used a mask for most of his appearances (76.3%), especially when accompanied in both open and closed spaces, and also maintained social distancing (72.4%). Mendoza had a lower rate of mask use (39.5%), but when accompanied she wore it on all occasions. Both greet others according to the new normal, i.e., with a fist bump or an elbow bump. De Soto wore a mask in 60.0% of his publications, but there were only five.
In Honduras, the incidence of the virus was lower than in the other countries analyzed. Both Asfura and Rosenthal showed a low rate of mask use in their publications (29.0%, respectively) and did not maintain social distancing on 70.0% of occasions. In contrast, Castro almost always wore a face mask (70.0%), including in his profile picture, but on no occasion does he maintain social distancing from third parties.

Discourse in the Face of COVID-19
In the analysis of the taxonomy of speaker intention related to COVID-19, most of the messages of the candidates were informative and expressive, especially from those already holding public office. In the case of Pérez (Ecuador), 53.9% were messages with an informative intention taxonomy, which was not surprising given the function of the candidate as a provincial prefect. The same happened with Forsyth (Peru) who, as mayor of La Victoria, visited several groups to distribute masks and, at the same time, to inform about the risks of the pandemic as well as the importance of preventive measures. The case of Lavín was similar, with 144 publications related to COVID-19 with informative intention (64.6%) or to promote events (26.4%). In the case of Jadue, mayor of a populous district of Santiago de Chile, he published 4.3 times fewer than his peer Lavín and his main intention was promotive (43.3%) and to a lesser extent informative (33.3%).
Informative publications were also in the majority for candidates such as Lasso and Pérez (Ecuador), Mendoza (Peru), and Jiles and Jadue (Chile). The latter also stood out in expressive publications, similar to the candidates from Ecuador Lasso (48.9%) and Pérez (30.3%), and the Peruvian Mendoza (20.0%). However, the publications of Chile were more advocative, especially Jiles (39.6%).
Cases without statistical significance were Arauz (Ecuador), whose only post on Instagram corresponded to an informative message, and Asfura, the only candidate in Honduras who mentioned the pandemic, once with an informative intention and the other time with an expressive intention.
In the contextual taxonomy, differences were found between candidates from the same country, as in the cases of Chile and Ecuador. In Chile, the posts of Jiles focused on minimal (62.3%) and close (35.8%) relationships and those of Lavín stood out for significant relationships (62.5%). In Ecua-dor, the posts of Pérez demonstrated the most significant relationships, with 51.3%, as opposed to the messages of Lasso, where their significance represented only 20.1%. In Peru, the candidates stood out for posts with a significant relationship (Forsyth with 52.6% and Mendoza with 50.0%) and a close relationship (Forsyth with 42.1% and Mendoza with 40.0%). Finally, in Honduras, the only two posts by Asfura had a close relationship, highlighted by the absence of images or comments.
Regarding the semiotic taxonomy, most of the information corresponded to the additive and parallel categories for all of the candidates from the four countries analyzed. The candidates from Ecuador stood out in the parallel category with Lasso (69.8%) and Pérez (48.7%), as well as the candidate Jiles from Chile (62.3%). In comparison, Forsyth from Peru had equal percentages for parallel and additive information (47.4%), comparable to his opponent Mendoza. Additive publications predominated in Chile with Lavín (87.5%) and Jadue (43.3%), as well as with Pérez in Ecuador (47.4%). Divergent publications had little relevance across all candidates.

Conclusions
According to their Instagram accounts, candidates seemed more concerned about the usual problems of the countries (economy, minorities, farm workers, weather disasters, corruption, new constitution, etc.) than reporting on preventive measures related to the pandemic. Several publications showed actions contradictory to WHO recommendations. Although mask use increased throughout 2020, its use was not a consistent practice among the candidates analyzed. For example, one might have thought the coronavirus had not reached Honduras given that only one candidate, Asfura, talked about COVID-19 in two publications, but appeared without a mask in 71.2% of the posts and without maintaining minimum social distance (73.3%).
The omission of COVID-19 from the various publications, together with the poor compliance by some candidates with the health recommendations, projected a misinformative discourse that could potentially confuse the public through imprecise and vague information, even more so when their multimodal messages were posted in situations that were clearly described as having a high risk of contagion. Candidates could generate distrust and confusion because there was no correlation between the discourse on the pandemic and the compliance with or reinforcement of WHO recommendations. In Ecuador for example, Lasso, the candidate with the most followers (121,000), with the most activity in the study period (743 posts), and with the most publications on the coronavirus (18.7%), appeared without a mask in 72.0% of the images and without maintaining social distancing (66.7%).
The behavior of the candidates in the face of COVID-19 was not consistent in all publications. Our hypothesis was confirmed: there was a lack of consistency in the discourse of coronavirus prevention with respect to the images and texts projected by the candidates on their Instagram accounts. Also, the majority of publications referring to the coronavirus did not refer to prevention/education, but rather to the promotion of the image of the candidate. The misinformation detected referred to imprecise, vague, and sometimes even contradictory information (Karlova and Lee, 2011;Karlova and Fisher, 2013). The multimodal discourse of candidates in the face of the pandemic was mostly misinformative regarding COVID-19, as it could generate confusion and mistrust in the recipients. As the results showed, there were variations among the countries analyzed and among the candidates. Moreover, the misinformative effect lay in the modality of the image. It was in the photographs the candidates shared on Instagram that contradictions were observed, e.g., whether or not the candidates talked about the pandemic, or if the prevention measures against COVID-19 recommended by WHO were not always followed. Therefore, despite not detecting intentionality on the part of the issuer (which would be considered disinformation), the misinformation was accidental or innocent (Buckland, 1991) and could potentially generate distrust and confusion in the receiver.
In addition to being a serious health problem, the pandemic was also a political issue as government decisions impacted discourse and the public images of the candidates. In this respect, the speeches, recommendations, or relative positions regarding the pandemic served to inform the voters and showed how the candidates assumed relevance and presence from various positions, such as supportive, critical-denunciating, managing, or passive-indifferent. In this sense, the configuration of a message did not respond to a taxonomy of absolute intention, as other intentions could always be interpreted behind a message and associated discourse, such as the case where the multimodal discourse on the pandemic served as a broader means of communication than a mere reference to the pandemic.
Beyond the classification of the taxonomy of the speaker's intention, four message styles could be identified in the posts of the candidates related to the pandemic, which were understood and contextualized in the particular context of each country. In Ecuador, in general, a supportive-managing communicative style was observed, i.e., they tended to connect with the emotions of the users (expressive category), demonstrated concern for the situation of those affected, and, via Instagram, demonstrated concrete actions to alleviate and improve the conditions of those affected (informative category). In Chile, the communication style was critical-denunciating. This was evidenced in the allusions to the "ineffective" management of the pandemic by the opponents of the current Government (provocative category). The style was also advocative, arguing and demonstrating how the management of the candidate could potentially better handle the situation (advocative taxonomy). Honduras showed a passive-indifferent communicative style, where politicians had a passive or directly null attitude both in the management of the pandemic by the Government and in their own actions. In Peru, the predominant style in the posts was management-advocative, especially those politicians who were running for the presidency from their current positions (mayors or governors). The discourse was oriented towards addressing the pandemic from the action of the person involved, demonstrating how they were managing and/or helping to address the problems of their jurisdiction (informative category).
In this aspect, note that the predominance of posts with minimal relation (in the contextual taxonomy) and parallel (in the semiotic taxonomy) indicated little recognition of the textual modality in Instagram, instead entrusting the delivery of meaning to images. However, it was striking how some candidates employed the textual semiotic resource to take their posts further, while others only reinforced an image through the textual format.
Likewise, the results of the analysis showed that a minimal contextual taxonomy relationship was not necessarily negative. In the triangulation of results, such as fewer posts and communications about the pandemic with more social activity while accompanied by other people (social distancing and greater use of masks), note that Arauz (Ecuador), Jiles (Chile), Forsyth (Peru), and Castro (Honduras) reinforced their messages visually. They represented the incorporation of the measures into their behavior that went beyond the formal discourse (texts), i.e., they did not talk about the measures, but they put them into practice in precautionary situations, particularly the use of masks. This behavior might contribute to the assumption of protective measures as a permanent and daily fact. Therefore, although the text and image had a minimal relationship, as a taxonomic evaluation they were the ones who most respected the rules in a proselytizing context.
The misinformative discourse could not be considered as a fulfilled discourse, at least not in a clear-cut way. The analysis showed that the results can vary in the sense that the shaping of the discourse of the candidate had many elements in its composition and in the end not all of them could be fulfilled. The misinformative behavior of the candidates varied in relation to time, such as the context of the communicative situation and the relationship of the candidate with different people. In this sense, there were messages which in their multimodal construction were completely clear and coherent (image, text, situation, and all the precautionary measures), but in others they were only half fulfilled or even varied in similar situations.
Of the total number of posts analyzed, only 1.0% in Honduras, 10.0% in Peru, 14.0% in Ecuador, and 17.3% in Chile were related to the management of the pandemic. That is to say, in the speeches of the candidates the importance given to the pandemic issue did not reach 20.0%. Moreover, in most cases it was simply an informative message, especially in the case of candidates who already held public office. The conclusion was that politicians focused a great deal on their campaigns and little on the pandemic. The omission of COVID-19 from most publications projected a misinformative discourse that may have potentially confused the public.