The construction of anti-Muslim hate speech on X: an exploratory analysis of comments by media ideology in six Spanish media outlets (March and May 2025)

Construcción del discurso de odio antimusulmán en X: análisis exploratorio de comentarios según ideología mediática en seis medios españoles (marzo y mayo de 2025)

A construção do discurso de ódio antimuçulmano no X: análise exploratória de comentários de acordo com a ideologia midiática em seis veículos de comunicação espanhóis (março e maio de 2025)

Daria Mottareale-Calvanese^1*
Catalina Argüello-Gutiérrez^{2**
Ángela Martín-Gutiérrez3***}

¹ International University of La Rioja (UNIR), Spain
² International University of La Rioja (UNIR), Spain
³ University of Seville (US), Spain
^* Tenured Professor (PhD) at the Faculty of Education, International University of La Rioja (UNIR), in the Department of Teaching and School Organisation. Spain. Email: daria.mottareale@unir.net
^** Tenured Professor (PhD) at the Faculty of Education, International University of La Rioja (UNIR). Spain. Email: catalina.arguello@unir.net
^*** Lecturer Professor in the Department of Theory and History of Education and Social Pedagogy, Faculty of Education Sciences, University of Seville (US). Spain. Email: amartin9@us.es (Corresponding author)

Abstract

Social media have become key spaces for the dissemination of hate speech. Among these, Islamophobia emerges as a form of religious racism that combines cultural prejudice and exclusionary narratives. In the Spanish context, the media play a significant role in shaping the image of Islam and Muslim people. This study analyzes how anti-Muslim hate speech manifests on the social network X, considering the ideological orientation of news media outlets. A quantitative, exploratory, and descriptive approach was applied. A total of 53,787 comments posted between March and May 2025 were analyzed from the accounts of six Spanish media outlets (El País, Público, 20 Minutos, El Confidencial, ABC, and OKDiario). Messages were manually coded (intercoder reliability = 95%) according to the type and intensity of religious hate (levels 1–4), and descriptive analyses and Chi-square tests were conducted to examine differences based on ideological orientation (left, center, right). A total of 297 messages (0.6%) containing anti-Muslim content were identified. Right-leaning media (64.3%) accounted for the highest proportion, followed by centrist (33.0%) and left-leaning outlets (2.7%). Messages of low or moderate intensity predominated, with no significant differences in interaction or tone. A total of 95% combined Islamophobia and xenophobia. Although a minority, anti-Muslim discourse reflects a clear ideological media bias and a form of “camouflaged hate”, subtle and persistent, which contributes to the normalization of prejudice and highlights the need for digital literacy and moderation policies that also address implicit expressions of hate.

Keywords
Islamophobia; Hate speech; Social media; Media ideology; X; Spanish news media.

Resumen

Las redes sociales se han consolidado como espacios clave para la difusión de discursos de odio. Entre ellos, la islamofobia se configura como una forma de racismo religioso que combina prejuicios culturales y narrativas de exclusión. En el contexto español, los medios de comunicación influyen en la construcción de la imagen del islam y de las personas musulmanas. Este estudio analiza cómo se manifiesta el discurso de odio antimusulmán en la red social X, atendiendo a la ideología de los medios informativos. Se aplicó un enfoque cuantitativo, exploratorio y descriptivo. Se analizaron 53.787 comentarios publicados entre marzo y mayo de 2025 en cuentas de seis medios españoles (El País, Público, 20 Minutos, El Confidencial, ABC y OKDiario). Los mensajes fueron codificados manualmente (fiabilidad intercodificador = 95 %) según tipo e intensidad del odio religioso (niveles 1–4), y se realizaron análisis descriptivos y pruebas de Chi cuadrado para examinar diferencias según la orientación ideológica (izquierda, centro, derecha). Se identificaron 297 mensajes (0,6%) con contenido antimusulmán. Los medios de derecha (64,3 %) concentraron la mayor proporción, seguidos por los centristas (33,0%) y los de izquierda (2,7%). Predominaron mensajes de baja o moderada intensidad, sin diferencias significativas en interacción o tono. El 95% combinó islamofobia y xenofobia. Aunque minoritario, el discurso antimusulmán refleja un claro sesgo ideológico mediático y una forma de “odio camuflado”, sutil y persistente, que favorece la normalización del prejuicio y plantea la necesidad de políticas de alfabetización digital y moderación que aborden también las expresiones de odio implícito.

Palabras clave
Islamofobia; Discurso de odio; Redes sociales; Ideología mediática; X; Medios de comunicación españoles.

Resumo

As redes sociais consolidaram-se como espaços-chave para a difusão de discursos de ódio. Entre estes, a islamofobia configura-se como uma forma de racismo religioso que combina preconceitos culturais e narrativas de exclusão. No contexto espanhol, os meios de comunicação influenciam a construção da imagem do Islão e das pessoas muçulmanas. Este estudo analisa como o discurso de ódio antimuçulmano se manifesta na rede social X, considerando a orientação ideológica dos meios informativos. Foi aplicado um enfoque quantitativo, exploratório e descritivo. Foram analisados 53.787 comentários publicados entre março e maio de 2025 nas contas de seis meios espanhóis (El País, Público, 20 Minutos, El Confidencial, ABC e OKDiario). As mensagens foram codificadas manualmente (fiabilidade intercodificador = 95%) segundo o tipo e a intensidade do ódio religioso (níveis 1–4), e realizaram-se análises descritivas e testes do qui-quadrado para examinar diferenças segundo a orientação ideológica (esquerda, centro, direita). Foram identificadas 297 mensagens (0,6%) com conteúdo antimuçulmano. Os meios de direita (64,3%) concentraram a maior proporção, seguidos pelos centristas (33,0%) e pelos de esquerda (2,7%). Predominaram mensagens de baixa ou moderada intensidade, sem diferenças significativas na interação ou no tom. 95% combinaram islamofobia e xenofobia. Embora minoritário, o discurso antimuçulmano reflete um claro viés ideológico mediático e uma forma de “ódio camuflado”, subtil e persistente, que favorece a normalização do preconceito e evidencia a necessidade de políticas de literacia digital e moderação que também abordem as expressões implícitas de ódio.

Palavras-chave
Islamofobia; discurso de ódio; Redes sociais; Ideologia mediática; X; Meios de comunicação espanhóis.

1. Introduction

Social media platforms have become key spaces for the dissemination of hate speech, in which Islamophobia acquires particular relevance by combining cultural prejudice with narratives of exclusion. In the Spanish context, the media play a central role in shaping the image of Islam and Muslim people, influencing how users interpret and reproduce these discourses on digital platforms such as X (Mabdi, 2025).

Recent data from the European Union Agency for Fundamental Rights indicate that the proportion of Muslims reporting experiences of racial discrimination in the EU increased from 39% to 47% between 2016 and 2022, meaning that nearly half report experiencing racism in their daily lives (FRA, 2024). Similarly, the Report on Ireland (fifth monitoring cycle) by the European Commission against Racism and Intolerance warns of a rise in anti-immigration and, at times, anti-Muslim public discourse, in which newcomers are portrayed as a threat to national security and cultural identity (European Commission - ECRI, 2025).

Within this context, the present study examines the manifestation of anti-Muslim hate speech on X according to the ideological orientation of news media outlets. To this end, three objectives are proposed: (1) to identify the prevalence of this type of discourse, (2) to analyze its characteristics according to the ideological orientation of the media, and (3) to examine its intensity and conversational dynamics.

2. Theoretical framework

2.1. Hate speech on social media and its Islamophobic manifestation in the Spanish context

Social media platforms have become a particularly favorable environment for the spread of hate speech (Matarín et al., 2025). Platforms such as X, Facebook, and YouTube not only amplify rumors and prejudice, but also facilitate the manipulation of public opinion (Russmann & Hess, 2025). Authors such as Maarouf et al. (2024) argue that factors such as anonymity, algorithmic logic, and the viral nature of these platforms create a scenario in which hate is not only expressed, but also legitimized, normalized, and, in many cases, capitalized upon through digital attention.

To understand what is meant by hate, it is necessary to focus on the concept of hate speech, which encompasses all expressions that incite rejection or violence against a social group or against someone simply because they belong to that group (Mansur et al., 2023). Although each country may classify these expressions differently within its criminal code, the aim of this study is not to determine whether a message constitutes a crime, but rather to understand its hateful content. For this reason, the study adopts a definition of hate understood as an attitude directed toward a group or individual because of their real or perceived membership in a particular group. From this perspective, hate speech is classified according to the groups toward which it is directed. These groups are organized on the basis of characteristics commonly perceived as markers of difference: religion, race, sex, sexual orientation, ideology, among others (Sellars, 2016).

In the Spanish context, Article 510 of the Criminal Code classifies as hate crimes those acts motivated by prejudice against the victim’s real or perceived characteristics, such as race, sexual orientation, or beliefs, since such acts violate human dignity and fundamental rights, thereby threatening social cohesion and the democratic values of equality and pluralism (Spain, 2015, Art. 510). In 2023, the National Office for the Fight Against Hate Crimes (ONDOD), under the authority of the Ministry of the Interior (Ministerio del Interior, 2023), reported that hate crimes committed on the Internet and social media focused primarily on racism, xenophobia, sexual orientation, gender identity, and ideology. Among the most common criminal offenses were threats, insults, and public incitement to hatred, as well as violence and discrimination, totaling 223 cases in 2023, representing a 32% increase compared to the previous year (Ministerio del interior, 2023).

These data reveal a concerning trend in the social construction of the “other,” who is increasingly perceived as a threat (Heang, 2024). In Spain, the Muslim community reached 2,412,344 people in 2023 (Observatorio Andalusí, 2024). Nearly half are Spanish citizens (45%) and are affected by Islamophobia (Charbi, 2025), understood as any form of discrimination that restricts their fundamental rights (Runnymede Trust Institute, 2017). On social media, Islamophobic hate speech has intensified considerably in recent years, becoming a growing concern for international organizations such as the Office of the United Nations High Commissioner for Human Rights (OHCHR, 2023). This phenomenon, and its implications for those who suffer from it, can be better understood through the concepts of prejudice, stereotype, and discrimination. Allport (1958) defines prejudice as an antipathy based on an inflexible and unfounded generalization. It may manifest openly and hostilely or in more subtle ways through the exaltation of traditional values, the exaggeration of cultural differences, or the denial of positive emotions toward the stigmatized group (Pettigrew & Meertens, 1995). Stereotypes, meanwhile, operate as simplifying beliefs that confine individuals to fixed categories, whereas discrimination represents the concrete action that translates these attitudes into forms of exclusion or violence (Zamora-Medina et al., 2021).

Islamophobia, articulated through prejudice and stereotypes, is disseminated through narratives that associate Islam with backwardness, extremism, or incompatibility with democratic values (Qamar et al., 2024). Consequently, beyond fear of the Islamic religion itself, Islamophobia becomes a specific form of racism directed both at Muslim people and at any expression of Muslimness, whether real or perceived. This rejection, consistent with the definitions of blatant or subtle prejudice mentioned above, is fueled by stereotypes portraying Islam as a religion inferior to the West, incompatible with other cultures, and closer to a violent ideology than to a spiritual belief system (Mabdi, 2025).

This homogeneous and reductive view of Islam, reinforced by populist discourses, is further amplified by the media, which frequently construct narratives in headlines and news reports associating Islam with insecurity, lack of integration, or cultural threat (Fuentes & Arcila, 2023). Various studies have demonstrated how the editorial line of media outlets shapes both the framing of information and audience reactions. In particular, media outlets with a more conservative or populist orientation tend to foster a greater number of comments laden with prejudice, stereotypes, and hate speech (Arce-García et al., 2024). According to the CIS (2024), media outlets are ideologically distributed, with OKDiario and ABC aligned with the political right; El País and Público with the left; and El Confidencial and 20 Minutos with the political center.

This public perception of the media makes it possible to establish certain correlations between the media orientation of the information people consume and the nature of the comments generated in digital spaces, causing Islamophobic representations to contribute to the viralization and normalization of these discourses, thereby manipulating and shaping not only public discourse but also citizens’ collective perceptions of Muslims (Šori & Vehovar, 2022). From the perspective of affect theory, hate cannot be understood solely as an individual attitude, but rather as a social emotion that circulates and becomes attached to certain bodies and collectives through discourse (Ahmed, 2004). This approach makes it possible to understand how particular representations of Islam and Muslim people acquire negative affective charge and become consolidated as persistent interpretive frameworks. The notion of “affective economies” describes precisely these processes of emotional circulation and accumulation, through which hate operates as a socially distributed and discursively sustained dynamic.

Studies such as that of Aleksandric et al. (2023) demonstrate that the initial responses within a conversation significantly influence subsequent tone and participation, highlighting the role of bystanders in amplifying or restraining the spread of messages. Similarly, Yu et al. (2025) show that certain “unexpected” participation patterns, such as when a tweet receives more replies or retweets than anticipated, reveal that the way people interact directly affects content reach. Retweets and replies therefore not only contribute to the spread of information, but also transform its visibility and credibility, confirming that conversational dynamics on X constitute a central element in the processes of virality and information circulation within the network.

2.2. Islamophobia and xenophobia as convergent discourses on social media

Islamophobia does not operate in a vacuum, but rather intersects with other forms of hate speech, especially xenophobia. Both share a logic of exclusion based on constructing the “other” as a cultural, political, or economic threat. In digital environments, and particularly on platforms such as X, these discourses reinforce one another, creating a composite narrative that stigmatizes those perceived as different, whether for religious, ethnic, or national reasons (Amores & Arcila-Calderón, 2025). The Office of the United Nations High Commissioner for Human Rights (2023) defined xenophobia as an attitude based on negative prejudices that translate into discriminatory practices against non-national groups, frequently fueled by fear of the cultural or economic “other”.

In Spain, this narrative materializes in the figure of the “foreign Muslim” even when referring to Spanish citizens of immigrant origin (Lindemann & Stolz, 2014). Recent studies demonstrate that media and political discourse tend to merge both categories, portraying the Muslim population as unintegrated, dangerous, or incompatible with Western values (Civila et al., 2020). In this context, Islamophobia operates as a form of cultural xenophobia, where rejection is based not solely on geographical origin, but also on supposedly incompatible values relative to those of the host country. The convergence of Islamophobia and xenophobia manifests not only in discourse, but also in practical consequences, including greater vulnerability for Muslims in access to public services, labor discrimination, physical assaults, and social exclusion.

These reactions cannot be understood without considering the disinhibition characteristic of digital environments, where online disinhibition encourages the use of more aggressive expressions (Heinze, 2016). Analyses of comments published in response to news about immigration or religion on platforms such as Facebook, YouTube, or X reveal recurring patterns of dehumanization and polarization. Recent research suggests that users tend to reproduce collective narratives grounded in cultural myths, perceptions of threat, and binary constructions of “us” versus “them” These narratives are reinforced by expressions already present in media discourse (Gualda, 2021)..

In some cases, comments do not express explicit hatred, but rather a veiled rejection disguised as civic concern or defense of democratic values (Carnovalini et al., 2025). This phenomenon, known as disguised hate speech, complicates its automatic detection and moderation by digital platforms (Cinelli et al., 2020). For all these reasons, identifying the issuers of hate speech on digital platforms is not a simple task, given that they do not constitute a homogeneous group. Moreover, the motivations driving them to disseminate such content online may range from ideological conviction to the pursuit of recognition, influence, or emotional release (Mao & Hu, 2025).

In response to this situation, several international organizations, including the European Union Agency for Fundamental Rights, warn in their Fundamental Rights Report that the rise of hate speech in digital environments constitutes one of the principal threats to fundamental rights, alongside violence against women and electoral manipulation (European Union Agency for Fundamental Rights, 2025).

At the legal level, Spain adheres to European guidelines on non-discrimination, criminally sanctioning public expressions of hatred against protected groups. Nevertheless, ambiguity in the application of laws, as well as inconsistencies in community standards across digital platforms, have resulted in phenomena of self-censorship and distrust among citizens. At the European level, the 2016 Code of Conduct commits digital platforms to removing clearly discriminatory content within a maximum period of 24 hours (European Commission, 2016). However, ongoing debates persist regarding the balance between content moderation and the protection of freedom of expression.

Based on all the foregoing, the hypothesis of this study holds that anti-Muslim religious hate messages on X vary according to the ideology of the news media outlet to which the news refers, such that comments responding to news from right-wing media exhibit a greater frequency and intensity of hateful expressions than those responding to news from left-wing media. Building on this hypothesis, the objective of the present study is to examine how anti-Muslim hate speech manifests on X according to the ideological orientation of Spanish media outlets.

3. Methodology

The study was conducted using a quantitative, exploratory, and descriptive approach (Hernández-Sampieri & Mendoza Torres, 2018) with the aim of analyzing anti-Muslim hate speech in Spanish digital media on the platform X. Data analysis employed descriptive statistics, including frequencies and percentages, as well as chi-square tests to examine differences according to media ideology and message characteristics.

The research focused on X, selected because of its relevance to news consumption in Spain. Data were collected using three TIER Basic licenses from X over a continuous two-month period between March and May 2025.

A systematic collection of comments on news posts published on X was conducted using six nationally circulated Spanish media outlets (ABC, El Confidencial, OKDiario, El País, Público, and 20 Minutos), selected for their relevance within the Spanish digital news landscape and their ideological positioning along a broad left–right spectrum.

The ideological classification of the media outlets was established on the basis of previous studies on media bias and audience perceptions (Masip et al., 2020). For the specific statistical analyses of this study, the outlets were grouped into three analytical categories: left, center, and right. This classification considered both previous empirical evidence and public perceptions regarding the ideological orientation of newspapers such as El País, ABC, Público, El Confidencial, and 20 Minutos, as reported in recent public opinion studies (CIS, 2024).

The dataset consisted of 53,787 messages, with no missing values. These messages underwent standard preprocessing procedures, including the removal of non-textual elements and content normalization, in order to ensure subsequent analysis (Ruíz-Iniesta et al., 2024). The corpus was manually coded by two previously trained annotators following a structured protocol for hate speech identification (De Lucas Vicente et al., 2022). Intercoder reliability reached 95%. Each message was classified according to the presence of hate speech, its type, its intensity (Levels 1-4), and the religion toward which it was directed.

Hate intensity was classified into four levels according to the type of attack and its potential repercussions, ranging from uncivil expressions to threats (De Lucas Vicente et al., 2022). Level 1 includes statements presented as factual claims, without aggressive language, but with the clear intention of stigmatizing a group, for example: (“And yet they still have 20,000 Muslims crossing daily”). Level 2 includes abusive expressions in which negative intentions or actions are attributed to a group without reaching the level of direct insults, thereby reinforcing negative stereotypes (“Don’t drink alcohol, but let a Moor rob you and occupy your home”). Level 3 encompasses direct verbal violence, including insults and explicit humiliation based on group membership (“Ramadan means shit”). Finally, Level 4 includes implicit or veiled threats, messages that, while not directly violent, suggest a real danger to the person or group being referred to (“Genocidal enemies of the faith are not negotiated with; they are annihilated”). In ambiguous cases, the higher intensity level was assigned.

For the main analysis, only messages classified as religiously motivated or xenophobic and directed against Muslims were selected. Religious affiliation categories were structured according to the religions with the largest number of adherents based on CIS data (2024). Through this process, 297 anti-Muslim messages were identified, constituting the basis of the study.

4. Results

The distribution of the total number of recorded messages (N = 53,787) by media outlet was as follows: El País (26.1%, n = 14,061), ABC (21.0%, n = 11,269), Público (18.9%, n = 10,185), OKDiario (15.0%, n = 8,089), 20 Minutos (10.5%, n = 5,661), and El Confidencial (8.4%, n = 4,522).

Regarding the ideological orientation of the media outlets, the sample was characterized by a greater representation of left-wing media (45.1%, n = 24,246), followed by right-wing media (36.0%, n = 19,358) and centrist media (18.9%, n = 10,183). Of the total messages analyzed, 19,121 (35.5%) were identified as hate messages, whereas 34,666 (64.5%) did not contain hate speech. The distribution of hate messages varied considerably across media outlets, with Público showing the highest proportion (39.5%) and El Confidencial the lowest (29.9%).

Among the hate messages analyzed, approximately 70% of the dataset was classified as Political or General hate speech. Although Misogynistic and Sexual hate categories were individually less frequent, they appeared notably in combinations, especially with Political or General content. Other combinations (involving three or more codes) represented approximately 10% of the messages, reflecting the intersectional nature of hate speech within this corpus.

Finally, the sample was filtered to include only those categories corresponding to religious (3) and xenophobic (4) messages that displayed some level of hatred toward Muslims. Messages meeting these criteria were categorized as anti-Muslim messages. A new dichotomous variable referring to anti-Muslim messages (1 = “Yes” / 0 = “No”) was created and used in the main analyses of this study. Based on this filtering process, a total of n = 297 messages were identified and selected for the principal analysis.

Subsequently, a description of the characteristics of anti-Muslim hate messages was conducted. These 297 identified messages correspond to 0.6% of the original sample. It is important to note that the majority of these messages were classified as religious or xenophobic hate speech (94.9%, n = 282), with small proportions containing mixed categories involving misogynistic or political elements in addition to religious/xenophobic hate.

Addressing the first objective (O1), the distribution of messages by media outlet was analyzed (Table 1). The chi-square results revealed significant differences in the prevalence of anti-Muslim messages across the different media outlets (χ² = 686.402, df = 5, p < .001, Cramer's V = .209). ABC and 20 Minutos exhibited the highest proportions, whereas El País and Público showed the lowest. It is noteworthy that OKDiario and El Confidencial did not register any anti-Muslim messages.

In line with O2, message frequency was compared according to media ideology, as shown in Table 2. The results revealed differences in the prevalence of anti-Muslim messages according to the ideological orientation of the media outlets (χ² = 208.813, df = 2, p < .001, Cramer's V = .064). Right-wing media outlets exhibited the highest proportion, followed by centrist outlets, whereas left-wing media reported the lowest proportion of anti-Muslim messages.

Subsequently, the distribution of anti-Muslim message intensity according to both media outlet and ideology was analyzed. In general, these messages corresponded mostly to low-intensity hate, as shown in Figure 1. Specifically, Level 1 intensity was the most frequent, accounting for 39.7% of the messages, whereas Level 2 represented 35.4%. Level 3 accounted for 22.9%, and Level 4 for 2.0%. This pattern suggests that most hate messages in the dataset represented relatively moderate rather than extreme forms of hateful content. However, chi-square analyses did not reveal a significant association between intensity and media outlet, nor between intensity and media ideology (ps > .202).

Finally, in order to address O3, conversational dynamics were examined by distinguishing between initial messages (posts initiating a conversation) and replies (interventions within threads). Results showed that 99.3% of the messages corresponded to interventions within threads, with no differences between media outlets (ps > .508). In addition, the number of replies received by these messages was analyzed, yielding the distribution shown in Table 3. Most posts received no replies, whereas only 30 messages received at least one response.

The distribution of intensities shows that conversational dynamics remained within a low- or intermediate-intensity register (Table 4). This pattern indicates that mild forms of anti-Muslim hate accumulate through insinuations or irony which, although seemingly harmless, create a hostile and persistent environment. No significant differences were found between ideological blocs, nor between comment position and intensity: the key factor lies not in who emits the message, but in how it is inserted into the conversation. Anti-Muslim discourse operates in a reactive and repetitive manner, contaminating threads without the need to escalate tone, which makes it more difficult to detect and more effective through repetition.

Although this pattern was observed, behavior did not vary according to the media outlet or its ideology, nor according to message intensity (ps > .508).

Furthermore, as shown in Table 5, the majority of posts did not receive any “likes” whereas fewer than 20% obtained at least one. No differences were observed according to media outlet, media ideology, or message intensity (,ps > .331).

5. Discussion and conclusions

The present study aimed to characterize anti-Muslim hate in responses to news content published by media outlets on X. The results show that anti-Muslim messages represent a relatively small proportion of the overall corpus; however, their distribution is not homogeneous. Regarding convergence with xenophobia, the vast majority of anti-Muslim messages were coded as “pure” religious/xenophobic hate, that is, with limited presence of combinations involving misogynistic or political components. This finding supports the argument that Islamophobia and xenophobia operate as overlapping frameworks that construct the “other” as a cultural threat, although their frequency within left-wing media remains residual. This result is consistent with the work of Amores and Arcila-Calderón (2025), who demonstrate how these discourses mutually reinforce one another in digital environments. Likewise, the narrative of the “foreign Muslim” even when referring to Spanish citizens, exemplifies how rejection is not limited to religion but also incorporates elements of national and cultural exclusion, (Lindemann & Stolz, 2014).

In this regard, the prevalence analysis indicates that the phenomenon exists, but does not dominate the broader conversation; in other words, the anti-Muslim subset represents only a small proportion of the total dataset. Moreover, its intensity in our analysis proved to be low. Nevertheless, its potential for symbolic harm appears substantial because of its persistent nature and its apparent discursive “respectability” (uncivil/malicious tone), which facilitates its circulation and normalization. Although this could be interpreted positively in terms of the absence of extreme outbursts, several authors warn that the real danger lies in the persistence of normalized hate speech expressed through subtle forms (Allport, 1958; Maarouf et al., 2024). As Heang (2024) points out, the perception of otherness as a threat is constructed cumulatively through everyday messages that reinforce stereotypes and prejudice. This pattern suggests that expressions of anti-Muslim hostility may respond more to processes of contextual activation than to a constant presence within digital conversation. In other words, certain news topics or narrative frames may function as discursive triggers that concentrate this type of messaging, rather than reflecting a uniform climate of rejection across all content (Ahmed, 2004).

In accordance with the proposed objectives, it was observed that 20 Minutos and ABC concentrated the majority of the messages (first objective). Thus, the digital newspapers categorized in this study as right-wing concentrated most of these messages, followed by centrist outlets, whereas those identified as left-wing media registered the lowest levels (second objective). It is important to emphasize that the ideological classification used in this study is based on the aggregation of continuous measures of media positioning drawn from previous literature, meaning that categories such as “left” or “right” encompass outlets with internal ideological nuances (Guerrero-Solé et al., 2022; Masip et al., 2020). This pattern nuances the initial hypothesis by highlighting that media outlets classified as centrist also occupy a relevant position within the observed dynamics. Specifically, when newspapers are analyzed individually, 20 Minutos and ABC stand out for their higher prevalence of anti-Muslim messages, compared to El País and Público, which show very low proportions, and OKDiario and El Confidencial, which registered no cases within the analyzed subset.

The evidence partially supports the initial hypothesis: right-wing media outlets exhibited the highest proportion of anti-Muslim messages; however, centrist media surpassed left-wing outlets and approached right-wing levels, introducing an important nuance into the ideological landscape. Likewise, the results confirm that Islamophobia and xenophobia operate as convergent discourses, jointly constructing the figure of the “other” as a cultural and social threat, although their prevalence in left-wing media remains very low compared to right-wing and centrist outlets. The detected pattern is consistent with literature warning about the fusion of both categories within media and political discourse, where Islamophobia acquires a xenophobic dimension by questioning not only religion, but also national belonging (Civila et al., 2020; Lindemann & Stolz, 2014). These findings invite a reconsideration of strictly binary interpretations of the media ideological spectrum. Rather than functioning as completely differentiated blocs, the data suggest possible areas of discursive convergence within certain news frames, implying that not only the ideological orientation of the outlet should be considered, but also the types of topics and informational approaches presented, as these may shape audience responses.

These findings are consistent with previous research indicating that editorial orientation shapes the framing of information and, consequently, the nature of the comments generated by users (Fuentes & Arcila, 2023). In this sense, the data reinforce the argument that the ideological orientation of media outlets plays a decisive role in the amplification of prejudice.

The fact that centrist media outlets also displayed significant proportions of anti-Muslim messages introduces an additional relevant nuance. Although the theoretical framework of this study, as well as the main hypothesis, suggested that right-wing media would concentrate the majority of these expressions (Šori & Vehovar, 2022), the results also indicate that dynamics of stigmatization may emerge from supposedly neutral positions. This may be partially explained by public perceptions that place certain centrist media outlets in an ambivalent position, capable of reproducing both narratives of inclusion and exclusion depending on the informational context (Guerrero-Solé et al., 2022).

In response to the third specific objective, the results show that the intensity of anti-Muslim hate was concentrated primarily at low or moderate levels (Levels 1 and 2, “uncivil” and “malicious”), with no significant differences across media outlets. Furthermore, higher-intensity messages did not generate a greater volume of replies within threads. This finding reinforces the idea of the proliferation on X of “disguised hate” that is, subtle and normalized expressions that are more difficult to identify both algorithmically and normatively than explicit attacks. In this respect, the results align with Pettigrew and Meertens’ (1995) notion of “subtle prejudice” whereby rejection manifests indirectly, covertly, or under the guise of civic concern. From an analytical perspective, this predominance of low-intensity forms also poses a methodological and conceptual challenge, as it shifts the focus away from easily identifiable extreme expressions toward ambiguous, socially tolerated discursive manifestations with greater potential for normalization, thereby contributing to their persistence and circulation without provoking immediate rejection in public interaction.

The predominance of low values in the distribution of hate message intensity across the analyzed media outlets, suggesting a problem of normalized subtle hate rather than isolated extreme outbursts, constitutes a significant contribution, as it encourages greater attention to normalized forms of discrimination which, although less visible, produce cumulative effects on social cohesion. In this regard, the evidence once again resonates with Cinelli et al. (2020), who emphasize the difficulty of identifying and moderating disguised hate within digital environments. Furthermore, the findings reinforce concerns expressed by international organizations such as the Office of the United Nations High Commissioner for Human Rights (2023) and the European Union Agency for Fundamental Rights (2025), both of which identify hate speech as a direct threat to democratic values and social cohesion. Indeed, although the majority of analyzed tweets did not generate conversation and only a small minority provoked interaction, revealing low audience participation and asymmetric engagement with news content (Carnovalini et al., 2025), this dynamic may nevertheless foster polarization, emotionally exhaust audiences, and enable the silent spread of hate even in the absence of visible responses, thereby persistently reinforcing toxic behaviors.

As a synthesis of the empirical results, it is important to highlight that high-intensity hateful expressions were found to be a minority, whereas uncivil or malicious forms of low or moderate intensity predominated. This pattern indicates that anti-Muslim hostility in the analyzed spaces manifests primarily as repeated and normalized discourse, rather than as explicit radical outbursts. Likewise, the overall low prevalence of anti-Muslim messages and their concentration within particular conversational contexts suggest the possible influence of contextual and situational factors, the confirmation of which would require longitudinal studies.

From an interpretive perspective, these patterns reinforce the importance of analyzing hate not only in its explicit forms, but also in its implicit and socially normalized manifestations. Given the exploratory nature of the design and the temporal delimitation of the sample, the results should be interpreted cautiously and do not allow for broad generalizations; nevertheless, they open new avenues of research regarding the circulation of expressions of rejection within digital informational contexts.

Among the principal limitations of the study is the fact that the sample was restricted to six news media outlets on X (formerly Twitter) and to a specific two-month period, which limits the generalizability of the findings to other media outlets, platforms, and broader contexts. Although message coding was conducted under structured protocols and with specific training, manual coding always entails a degree of subjectivity, especially in complex categories such as subtle hate or intersectional combinations. The analysis focused on associations, which limits the ability to control for relevant confounding factors such as news topic, political context, or the potential intervention of automated actors (bots). In addition, the study did not model in detail the impact of the platform’s algorithmic architecture, moderation policies, or network diffusion dynamics (retweets, quote tweets), factors that could significantly alter the prevalence and circulation of hate speech. Finally, the operationalization of media ideological orientation through three broad categories (left, center, right) represents a simplification of a more complex ideological continuum. Although this decision was based on previous estimations of media bias and audience affinity (Guerrero-Solé et al., 2022; Masip et al., 2020), future studies could employ continuous measures or more refined typologies whenever sample size permits.

Future research would benefit from implementing longitudinal designs capable of observing the evolution of hate speech across different moments, especially during politically charged contexts such as electoral campaigns or international crises. Likewise, comparative studies across digital platforms (e. g., YouTube, TikTok, Facebook) could help evaluate the role of algorithmic design in amplifying hate speech. It would also be useful to deepen intersectional analyses by examining how factors related to religion, gender, and race intersect within experiences of discrimination. Similarly, future work could incorporate network analyses and studies of diffusion dynamics (retweets, quote tweets, communities) in order to map who amplifies subtle hate and how it circulates among different ideological bubbles. Finally, research in this field could advance toward greater standardization of coding schemes for subtle hate and its intersectional combinations, together with the release of annotated Spanish-language corpora and the development of metrics addressing the persistence and contagion of uncivil tone, linking these phenomena to broader social processes such as social cohesion and public opinion climates.

Author contribution

Daria Mottareale-Calvanese: Conceptualization, Formal Analysis, Investigation, Methodology, Project Administration, Resources, Software, Supervision, Validation, Visualization, Writing – Original Draft, Writing – Review & Editing. Catalina Argüello-Gutiérrez: Conceptualization, Formal Analysis, Investigation, Methodology, Resources, Software, Validation, Visualization, Writing – Original Draft, Writing – Review & Editing. Ángela Martín-Gutiérrez: Data Curation, Formal Analysis, Investigation, Methodology, Resources, Software, Validation, Visualization, Writing – Original Draft, Writing – Review & Editing. All authors have read and agreed to the published version of the manuscript.

Conflicts of interest

The authors declare no conflicts of interest.

Funding

This article is the result of the research project HatemediaReligion (PC-24-0050), funded by Pluralismo y Convivencia Foundation.

Data availability

The dataset corresponding to the Supplementary Material supporting the findings of this study was published in the Zenodo repository and can be accessed at https://doi.org/10.5281/zenodo.15789730.

References

Ahmed, Sara. (2004). The Cultural Politics of Emotion. Edinburgh University Press.

Aleksandric, Ana; Singhal, Mohit; Groggel, Anne; & Nilizadeh, Shirin. (2023). Understanding the Bystander Effect on Toxic Twitter Conversations. In Proceedings of The Web Conference (WWW) ’23, April 30śMay 4, 2023, Austin, Texas, 6 p. https://doi.org/10.48550/arXiv.2211.10764

Allport, Gordon. (1958). The nature of prejudice. Addison‑Wesley.

Amores, Javier; & Arcila-Calderón, Carlos. (2025). What is the nature of the racist and xenophobic hate speech disseminated on social media in Southern Europe? Topic modelling of anti-immigration messages posted on X and YouTube in Spain, Italy and Greece. Communication & Society, 38(1), 163-180. https://doi.org/10.15581/003.38.1.013

Arce-García, Sergio; Said-Hung, Elias; & Montero-Díaz, Julio. (2024). Unmasking coordinated hate: Analysing hate speech on Spanish digital news media. New Media & Society, 27(10), 5848-5868. https://doi.org/10.1177/14614448241259715

Carnovalini, Filippo; Rodà, Antonio; & Wiggins, Geraint. (2025). Popularity Bias in Recommender Systems: The Search for Fairness in the Long Tail. Information, 16(2), 151. https://doi.org/10.3390/info16020151

Centro de Investigaciones Sociológicas [CIS] (2024). Barómetro de marzo 2024. CIS. http://bit.ly/48VA6I1

Charbi, Mohamed. (2025). La imagen del Islam en España: la islamofobia y los discursos políticos. L’Archétype, 3(spécial), 161-174. https://doi.org/10.34874/PRSM/archtype-56685

Cinelli, Matteo; Quattrociocchi, Walter; Galeazzi, Alessandro; Valensise, Carlo; Brugnoli, Emanuele; Schmidt, Ana; Zola, Paola; Zollo, Fabiana; & Scala, Antonio. (2020). The COVID-19 social media infodemic. Scientific Reports, 10, 16598. https://doi.org/10.1038/s41598-020-73510-5

Civila, Sabina; Romero-Rodríguez, Luis; & Civila, Amparo. (2020). The demonization of Islam through social media: A case study of# Stopislam in Instagram. Publications, 8(4), 52. https://doi.org/10.3390/publications8040052

De Lucas Vicente, Alberto; Römer Pieretti, Max; Izquierdo, Damaso; Montero-Diaz, Julio; & Said-Hung, Elias. (2022). Manual para el Etiquetado de mensajes de odio. figshare. https://doi.org/10.6084/m9.figshare.18316313.v4

España. (2015). Ley Orgánica 1/2015, de 30 de marzo, por la que se modifica la Ley Orgánica 10/1995, de 23 de noviembre, del Código Penal [Artículo 510]. Boletín Oficial del Estado, 77, de 31 de marzo de 2015, páginas 27070 a 27124. https://www.boe.es/eli/es/lo/2015/03/30/1/con

European Commission. (2016). EU Code of Conduct on Countering Illegal Hate Speech Online. European Commission. http://bit.ly/47cryuT

European Commission against Racism and Intolerance. (2025). ECRI report on Ireland (fifth monitoring cycle). Council of Europe. https://bit.ly/4nvAwK5

European Union Agency for Fundamental Rights. (2025). Fundamental Rights Report 2025: Testing times for fundamental rights (Informe sobre discriminación, racismo y discurso de odio en línea). https://fra.europa.eu

FRA - Agencia de los Derechos Fundamentales de la UE. (2024). Being Muslim in the EU: Experiences of Muslims. Publications Office of the European Union. https://fra.europa.eu/en/publication/2024/being-muslim-eu

Fuentes, Cristina; & Arcila, Carlos. (2023). El discurso de odio islamófobo en las redes sociales. Un análisis de las actitudes ante la islamofobia en Twitter. Mediterranean Journal of Communication, 14(1). https://doi.org/10.14198/MEDCOM.23044

Gualda, María Estrella. (2021). Metaphors of invasion: Imagining Europe as endangered by Islamisation. In Europe: Continent of Conspiracies (pp. 54-75). Routledge.

Guerrero‑Solé, Frederic; Mas-Manchón, Lluís; & Aira, Toni. (2022). The impact of the far-right on Twitter during the 2019 Spanish elections. Cuadernos. info, (51), 223-245. http://dx.doi.org/10.7764/cdi.51.27471

Heang, Ratha. (2024). Digital Nationalism: Understanding the Power of Digital Technology in the Rise of Nationalism, National Identity, and National Narratives. Open Access Library Journal, 11(9), 1-26. https://doi.org/10.4236/oalib.1112129

Heinze, Eric. (2016). Hate speech and democratic citizenship. Oxford University Press.

Hernández-Sampieri, Roberto; & Mendoza Torres, Christian Paulina. (2018). Metodología de la investigación. Las rutas cuantitativa, cualitativa y mixta. Editorial Mc Graw Hill Education.

Lindemann, Anaïd; & Stolz, Jörg. (2014). Use of Islam in the definition of foreign otherness in Switzerland: A comparative analysis of media discourses between 1970–2004. Islamophobia Studies Journal, 2(1), 44-58. https://doi.org/10.13169/islastudj.2.1.0044

Maarouf, Abdurahman; Pröllochs, Nicolas; & Feuerriegel, Stefan. (2024). The virality of hate speech on social media. Proceedings of the ACM on Human‑Computer Interaction, 8(CSCW1). https://doi.org/10.1145/3641025

Mabdi, Helen. (2025). Islamofobia, cultura y prejuicio: Una lectura crítica desde la filosofía política. Editorial Comares.

Mansur, Zainad; Omar, Nazlia; & Tiun, Sabrina. (2023). Twitter Hate Speech Detection: A Systematic Review of Methods, Taxonomy Analysis, Challenges, and Opportunities. IEEE Access, 11, 16226–16249. https://doi.org/10.1109/ACCESS.2023.3239375

Mao, Yuanyi; & Hu, Bo. (2025). From trolling victimization to reactive trolling: Moderated mediation effects of online disinhibition and motivations. Social Science Computer Review, 43(1), 27-47. https://doi.org/10.1177/08944393241250013

Masip, Pere; Suau, Jaume; & Ruiz-Caballero, Carlos. (2020). Percepciones sobre medios de comunicación y desinformación: ideología y polarización en el sistema mediático español. Profesional de la información, 29(5), 1-13. https://doi.org/10.3145/epi.2020.sep.27

Matarín, Eva; Gómez, Tomás; & Rodríguez‑Peral, Daniel. (2025). Propagation of hate speech on social network x: Trends and approaches. Social Inclusion, 13. https://doi.org/10.17645/si.9317

Ministerio del Interior (2023). Informe sobre la evolución de los delitos de odio en España 2023. Ministerio del Interior, Gobierno de España. http://bit.ly/4oLZC7f

OACNUDH - Oficina del Alto Comisionado de las Naciones Unidas para los Derechos Humanos. (2023). UN Human Rights Chief condemns rise in hatred. https://bit.ly/4bCII5Z

Observatorio Andalusí. (2024). Informe especial sobre islamofobia 2023. Unión de Comunidades Islámicas de España (UCIDE). http://bit.ly/49mR16b

Pettigrew, Thomas; & Meertens, Roel. (1995). Subtle and blatant prejudice in Western Europe. European Journal of Social Psychology, 25(1), 57–75. https://doi.org/10.1002/ejsp.2420250106

Qamar, Ayesha; Irtaza, Sadaf; & Raza, Syed. (2024). Islamophobia in Hollywood Movies: Comparative Analysis of Pre and Post-9/11 Movies. Journal of Development and Social Sciences, 5(1), 529-537. https://doi.org/10.47205/jdss.2024(5-I)48

Ruíz-Iniesta, A., Blanco-Valencia, X., Pérez Palau, D., Gregorio-Vicente, Ó., Cubillas-Mercado, J., Montero-Díaz, J., & Said-Hung, E. (2024). Informe final sobre el scrapeo de datos brutos obtenidos en medios informativos digitales españoles en X (Twitter), Facebook y portales web. https://recercat.cat/handle/2117/439452

Runnymede Trust Institute. (2017). Islamophobia: Still a challenge for us all. Runnymede. https://www.runnymedetrust.org/

Russmann, Uta; & Hess, Andreas. (2025). The management of uncivil and hateful user comments in Austrian news media. Journalism Practice, 19(2), 427-446.https://doi.org/10.1080/17512786.2023.2189152

Sellars, Andrew. (2016). Defining hate speech. Berkman Klein Center for Internet & Society. https://doi.org/10.2139/ssrn.2882244

Šori, Iztok; & Vehovar, Vasja. (2022). Reported User-Generated Online Hate Speech: The ‘Ecosystem’, Frames, and Ideologies. Social Sciences, 11(8), 375. https://doi.org/10.3390/socsci11080375

Yu, Yulin; Chen, Houming; Romero, Daniel; & Dhillon, Paramveer. (2025). Signals in the Noise: Decoding Unexpected Engagement Patterns on Twitter. Proceedings of the ACM on Human-Computer Interaction, 9(7), 1-33. https://doi.org/10.1145/3757657

Zamora-Medina, Rocío; Garrido-Clemente, Pilar; & Sánchez Martínez, Jorge. (2021). Análisis del discurso de odio sobre la islamofobia en Twitter y su repercusión social en el caso de la campaña «Quítale las etiquetas al velo». Anàlisi: Quaderns de comunicació i cultura, 65, 1-19. https://doi.org/10.5565/rev/analisi.3383