Introduction

Abortion has become a focal point within the feminist movement, especially with recent shifts in reproductive rights legislation in certain regions globally (Meneces Avella, 2023). Since 2020, countries like Argentina and Colombia have passed laws legalising abortion on demand, while the Mexican Supreme Court affirmed the right to legal, safe, and free abortion services in early pregnancy in 2023. Similarly, the Spanish parliament has adopted legislation allowing individuals aged 16 and above to access abortion without parental consent, thereby ensuring its availability in public hospitals.

Despite abortion’s legalisation in some areas, its ethical acceptability remains disputed, especially among activists campaigning for its prohibition online (Ntontis and Hopckins, 2018). Recent developments show a global trend of nations reconsidering their stance on abortion. In Argentina, for example, the current far-right government has proposed repealing the law allowing voluntary abortion up to the 14th week.

Feminist movements have reshaped women’s rights, especially in Latin America and Spain, but legal debates on abortion have sparked a conservative backlash (Walsh, 2010). This reaction is expected as groups resist changes to the status quo, defending entrenched interests (Blais and Dupuis-Déri, 2012). Socio-political events also strengthen cohesion and identity among right-wing extremist groups (Bliuc et al. 2020).

The link between anti-abortion activism and legal reforms is well-documented (Bonet-Martí, 2020). However, conservative groups also raise women’s rights issues during stable periods to challenge past policies (Hunt, 2021). In Spain, anti-abortion campaigns were prominent during the 2010 legalisation, but the far-right party Vox has included anti-abortion stances in its agenda since 2019 (Bernardez-Rodal et al. 2022).

Although different groups conducting anti-abortion activities emerged as early as 1970, coinciding with the second wave of feminism and the legalisation of abortion in the UK and the US, they now wield unprecedented technological resources to exert international influence. Gendered harassment targeting women seeking abortions persists outside clinics and has expanded into the digital realm (Lowe and Hayes, 2019).

In this investigation, we aim to explore the nature of anti-abortion communities on the internet, considering the convergence of modernised anti-abortion discourse and the increasing presence of the far right on social media platforms. Our study focused on Spanish-speaking users for several reasons. Firstly, because digital conversations in the Global South, including Spanish-speaking regions, have received less academic attention (Waisbord, 2023). And secondly, because Spanish, the official language in 21 countries and the second most spoken globally by native speakers, spans diverse geographic, cultural, and socioeconomic contexts. Unlike country-based analysis, language brings together transnational attitudes and experiences, which can be diverse yet common.

Moreover, Spanish-speaking countries provide a compelling subject for study, given the recent victories of the feminist movement concerning reproductive rights. Hence, our goal is to attain a thorough comprehension of anti-abortion communities in Latin America and Spain, emphasising their organisational structures and communication modalities. These data complement the studies centred in frames and discourses, as detailed subsequently.

In this research, we analyse their recruitment efforts on Twitter (X), a platform closely tied to current affairs, to ascertain whether there is a correlation between the content generated by anti-abortion groups and the sociopolitical context. Twitter’s design prioritises immediacy and public interaction, making it especially suited for capturing the dynamics of social movements and controversial issues. Moreover, existing scholarship has predominantly explored the relationship between hashtag engagement and activism on this social platform, particularly utilising political tags as a primary criterion in corpus selection (Jungherr, 2015).

To address our objectives, we formulated the following research questions:

RQ1. Which communities oppose abortion prominently within Spanish-speaking countries?

RQ2. What are the characteristics of individuals participating in anti-abortion discourse?

RQ3. What role do negative emotional content and hate speech play within anti-abortion communities?

First, we will overview the digital landscape of anti-abortion activism and outline our research methodology, which integrates deep learning for qualitative analysis. Then, we present key insights into anti-abortion campaigns as networks of diverse communities, highlighting patterns of hate speech and negative narratives that unite factions in this reactionary movement.

Organisational dynamics in conservative movements

The internet’s widespread accessibility has revolutionised communication, fostering environments conducive to social engagement and civic debate. The foundational literature on the internet and social movements highlights the rise of “connective action,” characterised by more horizontal and decentralised organisation (Bennett and Segerberg, 2012). Digital networks enable more autonomous activism, enhancing coordination for political purposes (Castells, 2015). Thus, the internet is a key resource for communication, community building, and citizen engagement (Eltantawy and Wiest, 2011).

Further studies have refined these perspectives, suggesting that networks are not simply spaces for spontaneous coordination but play a central role in shaping public spaces for participation (Gerbaudo, 2012). Social networks facilitate informal organisation, especially in disseminating information, alongside offline practices requiring more coordination (Tsatsou, 2018). Technology, therefore, works in tandem with traditional protest tactics (Treré, 2018).

Some scholars have noted that research has largely focused on progressive movements. Schradie (2019) argues that conservative organisations, with their stronger hierarchies and ideological cohesion, are more effective at utilising digital platforms than their fragmented counterparts. Right-wing echo chambers, reinforced by content-filtering algorithms, create environments where individuals primarily encounter like-minded perspectives, often without external debate (Barberá, 2020; Collins, 2024), despite ideological concerns about potential censorship by these platforms (Riedl et al. 2023).

These platforms enable interactions primarily among individuals with shared views and beliefs, reinforcing group cohesion (Sunstein, 2002). While this dynamic can amplify marginalised voices and provide a refuge for silenced perspectives (Karpowitz et al. 2009), it also fosters insular groups isolated from opposing viewpoints. As a result, the rise of counter-movements and even extremist groups in digital spaces has become more prevalent as they seek to influence political discourse (Tucker et al. 2018; Andersen, 2022). In this context, groups often seek public endorsements to amplify their message and form alliances with like-minded figures, further increasing their political activity (Hunt, 2021; Bliuc et al. 2007).

Ideological cohesion is closely linked to polarisation, especially since the disappearance of more moderate positions from the talks has been particularly salient in conservative groups (Freelon et al. 2020). Extreme right communities often connect with a narrow segment of the political spectrum, resulting in asymmetric polarisation, where divisions are not equally distributed across different ideological camps (Soares et al. 2019).

Polarisation also deepens the alignment between political and social identities, reinforcing stereotypes and negative perceptions of opposing groups (Wilson et al. 2020). In digital spaces, emotional biases are heightened (Redlawsk, 2002; Huszár et al. 2022; Chen, 2022) as algorithms prioritise morally charged and emotionally provocative content (Brady et al. 2020). Barberá and Rivero (2015) argue that a small group of highly engaged activists drives the perception of widespread polarisation on social media, making it appear more pervasive than it may be.

Tucker et al. (2018) explore the relationship between polarisation and the spread of negative emotional content and hate speech, questioning whether polarisation increases susceptibility to such content or if the proliferation of harassment fuels further division. The unregulated nature of social media, its anonymity, and its ability to foster impulsive interactions further amplify the use of offensive language and divisive rhetoric (Casero-Ripollés et al. 2023; Brown, 2018). These dynamics are particularly prominent among right-wing groups, where polarisation drives more cohesive, organised, and emotionally charged online communities.

Participation in unmoderated spaces beyond mainstream media has promoted conservative ideas online (Kasimov, 2023). This shift has led to a prevalence of emotional discourses in the digital public sphere, particularly on social media, highlighting a growing preference for beliefs over factual evidence (Han, 2022). While negative emotional content and hate speech may not always violate legal boundaries, their dissemination can foster societal radicalisation and violence (Sunstein, 2002; Cinelli et al. 2021). Tactics such as negative portrayals of adversaries, simplistic dichotomies, dehumanisation, and misinformation strategically undermine ideological opponents (Karyotakis, 2024; Tucker et al. 2018).

In the case of Twitter, researchers have observed a notable prevalence of strategies and dynamics associated with polarised discourse and segregation (Peña-Fernández et al. 2023), particularly when examining divisive hashtags (Bruns, 2021). These interactions frequently devolve into spaces dominated by personal attacks, derogatory remarks, and emotionally charged insults (Hernández-Santaolalla and Sola-Morales, 2019). This trend has been exacerbated since Elon Musk acquired the platform in 2022, as the engagement-driven ranking algorithm has increasingly amplified emotionally charged content, especially anger-laden tweets (Milli et al. 2024).

Far-right extremism strategically employs hate speech, often rooted in misinformative and disinformative content, to strengthen its influence within conservative circles and advocate for strong opposition to progressive parties and institutions (Freelon et al. 2020). This strategy encompasses a range of tactics across various internet platforms, allowing adaptation to fit specific contexts (Baele et al. 2023). Within this framework, online misogyny has emerged as a significant phenomenon, necessitating further examination of the dynamics within conservative reactionary movements in contemporary society.

In summary, examining right-wing strategies on the internet reveals three key factors: increased ideological cohesion, heightened polarisation, and the strategic use of hate speech and disinformation. These characteristics differ from those traditionally associated with progressive movements, necessitating specific studies to comprehend their strategies and communication modes, which align with the logic and objectives of these conservative movements (Kasimov, 2023).

Anti-feminism and anti-abortion campaigns

The efforts of marginalised groups to achieve political and social equality on platforms often provoke a powerful backlash from dominant groups, raising questions about power dynamics and threats to democracy (Kreiss and McGregor, 2024). Horta Ribeiro et al. (2021) observe a rise in toxic and misogynistic content within online anti-feminist communities, particularly in emerging platforms like the manosphere (Ging, 2019). Online misogyny manifests through digital attacks, often involving hate speech and misinformation directed at women, especially those challenging patriarchal norms. This phenomenon closely intersects with women’s psychological and physical autonomy, encompassing various issues such as rejecting traditional marriage norms, asserting sexual agency, and advocating for abortion rights. Conservative responses vehemently oppose these advancements, perceiving them as threats to traditional moral values and depicting them as hegemonic and oppressive (Cabezas, 2022).

The discourse surrounding abortion emerges as a locus of autonomy from conventional family structures and the subjugation of female bodies (Lagos and Antezana, 2018). Anti-abortion advocacy, often mobilised by diverse groups, has traditionally been depicted as unethical or immoral, with the influence of religiosity varying across different contexts (Lowe, 2016; Cornejo-Valle et al. 2023; Kalm and Meeuwisse, 2023). While some nations exhibit strong ties between religious institutions and anti-abortion movements, others witness a distancing from religious influence, especially with the trend towards secularisation (Aguilar Fernández, 2011).

Recent research notes a rhetorical shift within anti-abortion discourse, transitioning from “pro-life” to “pro-woman” frames to present a softer and less confrontational image (Roberti, 2021). Advocates increasingly rely on sophisticated strategies that emphasise the physical and psychological risks of abortion, refraining from direct blame on women and focusing on misinformation (Patev and Hood, 2021). This transformation reflects broader trends in human rights discourse, albeit with simultaneous efforts to restrict abortion access in some countries (Roberti, 2021; Zúñiga-Fajuri, 2014). In this vein, Martin et al. (2024) describe the “narratives of marginalisation” used by progressive pro-life influencers on social networks such as Instagram and TikTok. They often align themselves directly with the establishment leaders of the movement.

Autovictimisation tactics are commonly employed by conservative anti-abortion groups, framing gender politics, including abortion, as threats to societal values and perpetuating narratives of white male supremacy (Pruden, 2021; Cornejo-Valle et al. 2023). These discourses, propagated by religious organisations and far-right groups, entrench conservative agendas and heightened polarisation in the online sphere (Ophir et al. 2023).

Methods

This study utilises an inductive approach and computational techniques to identify and characterise elements like narratives, leadership, and organisational repertoires within anti-abortion discourse. It aims to explore this conversational space and the mechanisms behind community organisation, particularly in Spanish-speaking countries, amid contradictory trends observed in recent years.

Focusing on Twitter discourse, the study analyses tweets from September 1 to December 31, 2022, using Spanish hashtags #NoAlAborto, #Provida, #SalvemosLasDosVidas, and #SiALaVida to capture the broader discourse. These hashtags, identified through an extensive scan of tweets containing the word “aborto” by a team of native Spanish speakers, reflect commonly used terms in everyday interactions. This method allows the examination of anti-abortion communities’ everyday life without major campaigns or events, emphasising normalcy over crises. Using the Twitter Academic API, a dataset of 120,094 tweets was collected.

Two complementary methods based on distinct versions of Social Network Analysis (SNA) were employed to analyse these discussions. SNA was chosen to uncover communication patterns, leadership, and influence within the discourse, shedding light on how these communities organise and engage on social media. Initially, a network was constructed from interactions in the tweets, including retweets, replies, and mentions. The Louvain algorithm from the igraph library for Python (Csárdi and Nepusz, 2006) was used for community detection within the mentions network analysis. The network was visualised using Gephi (Bastian et al. 2009), highlighting clusters and their interactions, which helped identify central actors and key connections. Data was cross-tabulated and analysed with Microsoft’s PowerBi. To enhance insights from each cluster, several Natural Language Processing algorithms were employed, including:

  • The XLM-T-Sent-Politics model (Antypas et al. 2023) is used for sentiment analysis of tweets, capturing emotional tones and viewpoints.

  • Hate speech analysis uses the dehatebert-mono-english model (Aluru et al. 2020), with tweets translated into English via Google Translate to detect instances of intolerance.

  • User profile identification, including religious activists and politicians, uses the deberta-v3-small model (Reimers and Gurevych, 2019) as a zero-shot classifier to determine community user types.

  • Gender identification via the gender_guesser library in Python (Lead Ratings, 2016) categorises users as “male,” “female,” or “unknown.”

  • Text genre classification using the X-GENRE classifier model (Kuzman et al. 2023) categorises tweet styles into “information”, “news”, “opinion”, and “promotion”.

The Social Network Analysis utilised the Louvain algorithm to automatically and unsupervisedly categorise users into clusters. This community-detection method, aligning with an inductive approach common in qualitative strategies like grounded theory (Nelson, 2020), reduces observer bias. Our method emphasises users’ conversational behaviours, grouping them based on patterns. Various node-level metrics, such as input and output degrees, quantify a user’s influence within their community, indicated by mention volume. The dataset was enriched with advanced NLP variables like hashtags, locations, and tweet content, and it was organised in an interactive Power dashboard for comprehensive user behaviour analysis.

In addition to the visual analysis from PowerBi, various statistical techniques explored relationships among key variables. Cramér’s V quantified the strength and significance of associations among categorical variables, ranging from 0 to 1, with higher values indicating stronger associations. The Student’s t-test compared means between two independent groups to assess the disparity in quantitative variable impacts across qualitative groups. Cohen’s d measured effect size and practical significance, ranging from 0 to 1, to provide insight into the magnitude of the effect. These analyses were executed using the Scipy library for Python (Virtanen et al. 2020).

Results

Users’ connections and communities

The mentions network developed for this study consists of 42,312 nodes, representing users involved in the conversation either by mentioning others or being mentioned themselves (Fig. 1). These nodes are connected by 80,772 weighted edges, indicating the frequency of interactions between users. The weights of these edges vary from 1 to 311, reflecting the intensity of the mentioned relationships.

Fig. 1: The largest communities in the network.
figure 1

Source: Own elaboration (PowerBi).

Full size image

On average, each user has either mentioned or been mentioned 3.8 times (weighted average degree) and is in contact with approximately 1.9 unique users (average degree). This suggests that the network is relatively dispersed and lacks cohesion, with nodes having limited connections to each other. Such a structure tends to favour strong leadership, a characteristic observed in several communities analysed, where a single user stands out for receiving numerous mentions. Further indicating the network’s lack of cohesion and poor connectivity between nodes, the nodes within our network are connected by an average of 7.2 steps.

By applying the Louvain community detection algorithm implemented with igraph, a total of 574 communities were identified. However, only 12 of these communities account for 1% or more of the total nodes, representing 88.06% of the entire network. The modularity statistic for these identified communities is 0.65, indicating relatively insular communities where most edges exist within rather than between communities.

Discursive cohesion in users’ profiles

The examination of Twitter interactions on the anti-abortion discourse revealed several distinct communities within the conversation graph. These clusters are not only different in their actors but also in their narratives. Negative emotional content is present in 48% of original tweets, while hate speech is detected in 23.56%. This toxic tone acts as background noise in the conversation. Upon juxtaposing the data with the sentiment, hate speech, and gender classifications of tweets, it became evident that these communities exhibit varying levels of negative framing tweets (“A child is never a mistake”) and hate tweets (“Against the killing of the innocent”) (Table 1). Notably, communities 2, 0, and 14 show the highest concentration of negative framing tweets, while their prevalence of hate speech marks communities 2, 1, 5, and 18. Thus, the findings suggest that negative framing and hate are not always correlated.

Table 1 Presence of negative and hate tweets in the communities.
Full size table

The predominant content across most communities tends to be opinionated or argumentative, aligning with an activist and highly politicised discourse. On this point, the opinions most applauded with likes are the one that disempowers the mother’s body from the moment of fertilisation (“It’s not just my body, it’s my daughter”) and the view that emphasises women’s reproductive rights disappear with the right to life granted to the embryo (“Despite their small size, NO ONE can deny them the right to live”).

Another critical aspect entails identifying user types within each pertinent community (Table 2). The analysis reveals a prominent presence of religious and political activists across the examined conversations. Particularly, religious activists are prevalent in communities 18, 5, 7, and 8, with notable accounts such as @porlavida2014, @ObispoMunilla, and @LAM_Oficial. Conversely, political engagement is more pronounced in communities 2, 5, and 0, featuring actors like @Ferantonia28, @cfelmerv, @Freealwaysss, @Pipistrella11, @Antoniovillar73, and @libertaaddexp76. Although the representation of users classified as politicians is slightly lower, it remains significant, notably in communities 0, 1, and 5. Conversely, the presence of users identified as journalists is less conspicuous, though in certain communities, such as 5, 18, and 10, their presence reaches or approaches 1.5%.

Table 2 Presence of users classified by different criteria in each community.
Full size table

From a geographical standpoint, it is noteworthy that most communities predominantly consist of users from a single Spanish-speaking country, indicating a trend of national homophily in the discourse. Only communities 3 and 18 demonstrate communication patterns with a more transnational dynamic. An example of the geographical dispersion between communities is the following tweet, referring to the demonstrations in Mexico following the legislative change: “The march had a strong turnout, yet I wished for broader participation from Catholics and society at large. Our common adversary remains unchanged. #Notoabortion. https://t.co/LnqVoCYCXC”. Gender analysis reveals a male predominance in the dialogue, with communities such as 14 and 8 (precisely noticeable for their hate speech contents) comprising over 70% of male participants. However, this does not preclude the possibility, as elucidated below, of certain communities having female leadership, such as 2, led by conservative Chilean activists.

In the statistical analysis employing Student’s t-test, evidence emerges of a relationship between the classification of tweets as hateful and identification with certain user types. Tweets categorised as hateful exhibit, on average, higher scores on variables associated with religious activists (N = 43,300, p-value < 0.001; Cohen’s d = 0.17), politicians (N = 43,300, p-value < 0.001; Cohen’s d = 0.07), and political activists (N = 43300, p-value < 0.001; Cohen’s d = 0.15). Although the effect sizes are small, they remain consistent and significant, suggesting a trend wherein hate tweets are marginally more associated with these activist figures. Conversely, hate tweets are significantly less linked with journalists (N = 43300, p-value < 0.001; Cohen’s d = −0.07), as indicated by a lower mean in this rating category.

These emotional trends can be seen, for example, in tweets published by Bishop Munilla, who justifies his oratory at the doors of abortion clinics in Alicante (Spain) showing a decontextualised video of a birth by caesarean section or a photograph of him and other catholic activists in the street: “On this #AllTheFaithfulDefused day, I joined the 40 Days For Life campaign, standing in prayer outside an abortion clinic in Alicante. If abortion is deemed progressive, then embracing the law of the jungle would be the pinnacle of democracy!”. Other contents link the concepts of abortion and mother with murder and crime, such as those of the Colombian Catholic Movement Lazos de Amor Mariano –“Our Father, Hail Mary, and Glory Be for all the babies who are daily murdered in their mothers’ wombs. #NoToAbortion”− and @porlavida2014 –“If killing a newborn is a crime, why is it permitted in the womb? #ChooseLife”.

Concerning public engagement metrics, tweets labelled as hateful receive lower numbers of impressions, replies, retweets, quotes, and likes than tweets labelled as non-hateful. Notably, the disparity in the number of retweets (N = 43300, p-value < 0.001; Cohen’s d = −0.97) and ‘likes’ (N = 43300, p-value < 0.001; Cohen’s d = −0.94) is particularly pronounced. This trend indicates a negative reception of hateful content and a reluctance to interact with it. Regarding content generation, there is a notable incidence of automation in communities 18 and 10, where approximately 5% of users engage in content repetition practices.

In the quantitative analysis of categorical variables and their relationship with the hate speech label, the application of the chi-square test reveals statistically significant associations in all instances. Utilising Chi-Squared Tests and Cramér’s V as a measure of the strength of the association, it becomes apparent that the user’s community (N = 43300, p-value < 0.001; Cramér’s V = 0.43) exhibits the strongest correlation with the hate speech label. This underscores that the community to which a user belongs serves as a significant predictor of whether their speech is categorised as hate speech, indicative of distinctive communication patterns within specific communities.

Furthermore, the relationship between country and hate speech label is particularly noteworthy (N = 43300, p-value < 0.001; Cramér’s V = 0.23), which may indicate cultural or regional disparities in the expression of hate speech (i.e., a higher prevalence of hate speech in Chile or Mexico compared to what would be expected by chance). Conversely, the various tweet genres appear to be moderately correlated with hate speech (N = 43300, p-value < 0.001; Cramer’s V = 0.19; i.e., tweets categorised as “promotional” or “other” tend to exhibit higher levels of hate content). In contrast, user gender or propensity to repeat content exhibit weaker associations (N = 43300, p-value < 0.001; Cramér’s V = 0.09 and 0.03, respectively), suggesting that these factors are unreliable predictors of hate speech.

Additionally, the relationship between country and hate speech label is noteworthy (N = 43300, p-value < 0.001; Cramér’s V = 0.23), possibly indicating cultural or regional disparities in hate speech expression (i.e., the prevalence of hate speech in Chile or Mexico exceeds what would be expected by chance). Conversely, different tweet genres appear to be moderately linked to hate speech (N = 43300, p-value < 0.001; Cramer’s V = 0.19; i.e., tweets categorised as “promotional” or “other” tend to contain higher levels of hate content). In contrast, user gender or content repetition tendencies exhibit weaker associations (N = 43300, p-value < 0.001; Cramér’s V = 0.09 and 0.03), suggesting they are unreliable predictors of hate speech.

Characteristics of communities

A detailed analysis of larger communities will provide insights into the main communication strategies, identity markers, and interaction dynamics shaping the abortion discourse on this platform. Community 3 stands out as the most prominent, encompassing 41.95% of the nodes, spearheaded by the user @porlavida2014, emerging as a leading advocate for anti-abortion or “pro-life” viewpoints. This community is characterised by its sustained activity, particularly between October and December. It predominantly comprises Latin American users who engage in retweets and mentions of the community’s primary users. Notably, the user descriptions within this community prominently feature religious, conservative, and pro-traditional family affiliations originating from diverse countries. The content predominantly consists of activist messages imbued with both mobilisation and persuasion. It is noteworthy that, unlike other clusters discussed below, the framing of the most widely shared messages in this cluster remains positive and eschews confrontation or disqualification.

Community 2 emerges as the second largest community, comprising 9.08% of the nodes, lacking clear leadership but distinguished by the engagement of several conservative Chilean women. Sporadic bursts characterise this group’s activity, which is geographically concentrated in Chile. Unlike the preceding community, the tweets within this cluster typically exhibit a more negative tone, with a significant portion containing hate speech (i.e., 38.18% of the tweets). The content predominantly features anti-abortion slogans and is marked by its aggressiveness, eschewing attempts to persuade opponents.

Community 0, the third largest group, encompasses 8.71% of the nodes and comprises leaders affiliated with the Vox party in Spain, as well as a significant presence of anonymous activists and a few local officials. Most of their tweets adopt a negative stance towards abortion, albeit with a lower frequency of hate speech compared to the second community. The content predominantly revolves around anti-abortion slogans and reactions to news items, lacking a strong emphasis on deliberative discourse.

The fourth community, constituting 7.10% of the nodes, is Community 10, distinguished by its focus on Mexico and spearheaded by Catholic and conservative figures and organisations. The sporadic activity predominantly occurs in early October −when a demonstration took place in the Mexican capital against the decriminalisation of abortion sanctioned a month earlier− with tweets striking a balance between positive and negative tones and a lesser prevalence of hate speech compared to other communities. The primary messages aim to galvanise political opposition against abortion rights in Mexico, with limited deliberative content.

Additional smaller communities, led by activists and politicians of diverse nationalities, exhibit an abundance of conservative and Catholic anti-abortion markers, featuring predominantly negative discourse with a minimal presence of hate speech. The narrative centres on political and social mobilisation against abortion rights, exhibiting varying intensity and focus across different communities.

Community 8, led by the Bishop of Orihuela and HazteOir in Spain, features strong Catholic influences and a fluctuating narrative against abortion centres, with some hate speech. Community 14, comprising Colombian activists, adopts a negative, alt-right stance against leftist and abortion rights movements, marked by significant hate speech. Community 1, organised by Chilean activists, espouses a right-wing, pro-traditional family narrative with conspiracy elements, resulting in a predominantly negative and viral discourse.

Community 5, led by Argentine Catholics, critiques the Pope while promoting anti-abortion arguments, maintaining a largely negative tone. Community 18, featuring Catholic communication leaders, presents a primarily positive anti-abortion discourse that incorporates prayers and economic arguments but also includes notable hate speech. Community 13, led by a journalist, focuses on a parliamentary narrative with a consistent publishing schedule and minimal hate speech. Lastly, Community 9 advocates for progressive discourse, critiquing the “pro-life” movement and promoting equality with minimal hate speech.

Discussion

This study has sought to investigate the landscape of Spanish-speaking anti-abortion communities on Twitter, revealing a diverse array of groups with varied political backgrounds and narratives. This observation aligns with existing research on the online far-right ecosystem, which characterises it as “vast, dynamic, multidimensional, and heterogeneous in terms of ideology and practices” (Baele et al. 2023, p. 1601). The analysed network underscores the complexity associated with anti-abortion advocacy, which encompasses diverse and sometimes contradictory trends, ranging from communities associated with religion to online misogynistic groups that instrumentalise anti-abortion discourse. Consequently, the generalisability of the anti-abortion conversation is limited. Thus, the primary contribution of this study lies in identifying certain trends within the Spanish-speaking online environment.

These findings offer valuable insights into the organisational dynamics within these communities, diverging from the horizontal visions typical of progressive social movements (Freelon et al. 2020). Only a select few of the numerous communities of the debate demonstrate significant organisational structure and sustained activity (RQ1). Cohesion among different communities is modest, with interactions predominantly occurring internally rather than across different groups: there is no repertoire of joint or organised mobilisation, nor is there any flow of interaction leading to an amplification of the anti-abortion discourse, even within the proven national homophily.

The absence of cohesion facilitates the emergence of influential leaders who attract the bulk of interactions, acting as focal points for discourse cohesion. Previous research has identified alliances between these groups and political figures (Hunt, 2021), and we extend this observation by noting that these alliances often adopt a hierarchical structure (Blais and Dupuis-Déri, 2012). Leaders influence many users in this structure rather than foster a more collective form of organisation, as previously observed in investigations on reactionary groups (Kasimov, 2023). Consequently, this hierarchical arrangement gives rise to diverse discourses and narratives opposing abortion. It remains to be seen whether this top-down dimension of activism responds more to the lack of organisation of the anti-abortion communities −which have entered the digital space after pro-rights organisations− or whether it represents a new form of hierarchical activism, reflecting the logic of traditional institutional conversationalists such as the Catholic Church in the online environment.

Secondly, the profiles engaged in anti-abortion discourse predominantly consist of religious activists, politicians, and political activists (RQ2). Despite previous investigations claiming the separation of anti-abortion discourse from religion (Aguilar Fernández, 2011), this study demonstrates that religion continues to play a significant role, particularly in Latin America. Male participation in the discourse is prevalent. This specific data reveals that anti-abortion discourse is part of the manosphere, which has been extensively investigated in recent years (Ging, 2019) and, in this instance, demonstrates its strategic objective of controlling women’s bodies and reproductive rights. However, the data does not permit confirmation of a direct relationship between gender and the use of hate speech.

Geographical location also serves as a significant organising principle in anti-abortion communication with most communities and countries. This distribution corroborates prior research emphasising the strong national component of contemporary misogyny and far-right movements (Bonet-Martí, 2020). Legislation undoubtedly exerts significant influence over the discourse surrounding abortion. However, it is crucial to underscore that anti-abortion sentiments frequently persist beyond legal mandates, as observed in Spain. Despite existing legislation governing abortion, entrenched anti-abortion attitudes endure within the country. Conversely, our empirical investigation unveils ongoing deliberations surrounding a prospective constitutional reform in Chile, with considerations advocating for constraints on abortion rights. These empirical instances elucidate the enduring and dynamic landscape of anti-abortion discourse within Latin America.

Thirdly, communities exhibit varying degrees of negative narratives and hate speech (RQ3). Certain communities (2, 0, and 14) display a high concentration of negative framing tweets, while others (2, 1, 5, and 18) are notable for their prevalence of hate speech. Inspired by previous investigations identifying different actors in anti-abortion campaigns (Cornejo-Valle and Pichardo Galán, 2017; Kalm and Meeuwisse, 2023), this study discerns three main groups based on user profiles and specific discourse.

Firstly, conservative Catholics, found in communities 3, 10, and 14, prioritise religious, conservative, and pro-traditional family values in their discourse. They maintain a balanced tone, with less hate speech, and focus on opposing abortion through prayers and economic arguments. Secondly, right-wing activists, represented by communities 2, 1, and 5, display intermittent activity marked by aggressive anti-abortion rhetoric and high virality, often accompanied by hate speech. Finally, the institutional far-right, seen in communities 0, 8, and 18, consists of anonymous activists, local officials, and political parties with conservative leanings but not necessarily Catholic affiliation. This group has a stronger presence in the media.

Different community types engage with distinct narratives. Activist profiles focus on mobilising populations and spreading extreme arguments. At the same time, political parties aim to become reference points in anti-abortion struggles, possibly explaining the use of bots by communities like 18. The study highlights that negative framing and hate speech persist in contemporary abortion debates, influenced by the extreme right and online misogyny, aligning with previous research on the polarising role of activists (Barberá and Rivero, 2015).

The classification offers a framework for understanding anti-abortion communities, but due to their diversity, it is not universally applicable. These loosely connected communities show that anti-abortion discourse unites various organisations despite lacking a rigid structure, challenging its categorisation as a counter-movement.

However, it illuminates the dialectical relationship between anti-feminist counter-movements and feminist demands, demonstrating adaptation and transformation according to different contexts. Additionally, it reveals that activism against women’s rights tends to be reactive rather than proactive, responding to achieved or ongoing rights advancements. Despite recent advances in reproductive rights in some Western countries, it is important to note that our research was not conducted on significant dates for the feminist movement. Although the International Day for the Elimination of Violence Against Women falls on 25 November, none of the hashtags analysed are related to this specific event, nor is this date marked by antifeminist movements. Nevertheless, we observed messages exhibiting considerable cohesion and polarisation. Therefore, we conclude that political events catalyse this movement (Bliuc et al. 2020), while its intrinsic characteristics remain inherent to the counter-movement itself, giving it permanent coherence.

This study has limitations. The research duration between September and December 2022 is relatively short for collecting high-quality data to interpolate. It does not quantitatively analyse the frames of pro-life and pro-choice movements, leaving an opportunity to explore their link to hate speech and negative framing in future research. The focus on textual information limits the examination of visual narratives, suggesting a multimodal approach for future studies. Additionally, the methods used are insufficient for investigating hoaxes, highlighting the need for further exploration of their connection to hate speech and disinformation.

Future research could also examine far-right groups’ use of alternative platforms (Collins, 2024; Martin, 2024) and extend the analysis to other language communities. Comparative studies, such as those involving French- or Arabic-speaking populations, could reveal cultural differences in abortion debates. Exploring the potential link between social media extremism and offline mobilisations would also be valuable. Finally, observing the rhetoric of public figures like Javier Milei and Nayib Bukele, who oppose reproductive rights, could provide further insight. This work has proved that anonymity does not characterise the main actors in the conversation.