Literature review

The establishment of social networks has meant the structuring of virtual spaces where adolescents relate and share interests and experiences (Burnette et al., 2017), meeting different social needs (Anderson & Jiang, 2018; Smith & Anderson, 2018; Throuvala et al., 2019; Vannucci & McCauley-Ohannessian, 2019).

Instagram is a space where social comparison takes place through a canonized model of beauty, that is, it operates an idealised comparison (Burnette et al., 2017; Kim & Kim, 2019; Schmuck et al., 2019; Yau & Reich, 2019) that models self-concept (Kim et al., 2017; Verrastro et al., 2020). This situation is propitious for the emergence of the well-known Instagrammer, users whose function is to influence the virtual community by showing aspects of their lives and establishing themselves as marketing images of the companies that sponsor them (Boerman, 2020; Schouten et al., 2020; Weismueller et al., 2020). The linguistic simplicity and strength of the image means that it is preferred by influencers to capture an audience (Shane-Simpson et al., 2018).

On the other hand, YouTube is a social network based on audio-video content in which videos about people’s personal lives and interests (Jerslev, 2016) or celebrities (Ferchaud et al., 2018; Hartmann, 2016) are regularly presented. In other words, YouTubers are established as video bloggers who regularly post videos about their lives, interests or skills, propose challenges and participate in conversations with other users, i.e. a parasocial relationship is generated between influencers and followers (Ashman et al., 2018; de-Bérail et al., 2019; Genç & Öksüz, 2019). Nevertheless, a study with Spanish teenagers has argued that the most appreciated values are humour and leisure, that is to say, these users value their comedy and not the image of a brand that they can represent (Aran-Ramspott et al., 2018). Twitter is considered the network par excellence of information and communication, being used to debate social, environmental and political issues, rendering it a thermometer of social concerns (Harb et al., 2020; Neu et al., 2019; Peres et al., 2020).

In short, teenage influencers are teenagers who narrate and publicly show their lives (fashion looks, romantic stories, make-up skills, socio-political concerns, etc.) (Ferchaud et al., 2018; Hartmann, 2016) through social networks, and can consequently be considered child and youth celebrities (Aran-Ramspott et al., 2018). In most cases, they do not present a transgressive vision but imitate sponsors in the publications they share (Blasco-García, 2020). Their ability to reach out to the teenager gives them a privileged position from a marketing standpoint (Bakir et al., 2020; Boerman, 2020; Schouten et al., 2020). In this way, they favour the transmission of social norms based on consumerism, which not only affects the target population but also normalizes cruel optimism (Ashman et al., 2018). On the other hand, a dichotomous parasocial relationship is established between influencers and followers, so that the latter empathise to such an extent that a physical encounter is not necessary (Ashman et al., 2018; de-Bérail et al., 2019; Genç & Öksüz, 2019). For instance, Bhatia (2018) shows how YouTubers ‘discursively exploit the boundaries’, oscillating between expert and colloquial knowledge and taking advantage of their discursive competence. In this way, they show themselves to be participatory and committed, making users feel part of the YouTube community as they address them, surely read their comments on videos and so forth, thereby generating an emotional commitment (Bhatia, 2018; Burgess & Green, 2009; Lange, 2014). Specifically, Bhatia’s (2018) results show how YouTubers carefully prepare their videos, presenting some degree of strategy. Scannell (2000) argues that conversational quality is a determining factor in sociability. In short, publications on social networks are far from natural; on the contrary, they show a communicative strategy.

From this perspective, we should ask ourselves why adolescents use social networks. Existing research findings suggest that adolescents engage in social sharing and self-editing to adjust their views of themselves to their ideal self. In other words, they seek idealised self-representation (Burnette et al., 2017; Schmuck et al., 2019; Stockdale & Coyne, 2020). This pursuit of the ideal can reinforce low self-esteem and low self-concept and even instigate risky eating behaviours (Burnette et al., 2017; Schmuck et al., 2019), although authors such as Stockdale and Coyne (2020) highlight how, in addition to social comparison, boredom plays a key role. Authors such as Mäntymäki and Riemer (2014) argue that social network use is mediated by hedonism, i.e. a search for pleasure and satisfaction in an online life. Others point out that it is the result of a sense of belonging and self-disclosure, key processes in identity construction at this stage (Davis, 2012). While the theory of uses and rewards and the theory of self-determination expose the relevance of controlling relationships, content, presentation and impressions, this may be the same etiology of fear of missing out (FOMO) and nomophobia (Throuvala et al., 2019). On the other hand, it is necessary to recognise how authors like de-Bérail et al. (2019) and Hartmann (2016) are based on the theory of parasocial relations (which try to satisfy those social needs that are lacking). In this sense, people with anxiety, loneliness or social difficulties rely on parasocial relationships to compensate for their difficulties. Consequently, a bond is generated between the influencer and their followers.

In terms of coherence, the state of the art shows univocal traits in terms of polarity of feeling in social networks, with the idealised vision of oneself and any situation being practically hegemonic. Thus, positive or neutral texts with a high degree of subjectivity abound (Peres et al., 2020; Reyes-Menéndez et al., 2018; Vizcaíno-Verdú & Aguaded, 2020). However, there are certain dichotomies according to the subject matter, so that body image, self-expression, travellers, digital life and startups are associated with positivity, while those related to depression, loneliness and real-world relationships have negative polarity and self-identity and anxiety are neutral (Saura et al., 2019).

One of the outstanding features of social network publications is the use of the hashtag, a short fragment headed by # which manages to increase a publication’s visibility and virality (Lipsman et al., 2012; McGoogan, 2017). Erz et al. (2020) show how the use of hashtags correlates with a person’s idealized exposure. Similarly, the presentation of consumer elements at the beginning of YouTube videos increases their market share (van-Reijmersdal et al., 2020). As for the most followed content on YouTube, Castillo-Abdul et al. (2020) refer to important gender differences, in such a way that boys stand out as ‘gamers’ while girls do so in lifestyle issues.

Research questions and objectives

It is important to examine the behaviour of adolescents in partner networks by studying their relationships with influencers. The following research questions are thus posed: Which posts do adolescents value most from their influencers? Are there differences according to the social network? What type of content is best received by the adolescent population?

The general objective of the research is to understand the behaviour of the most relevant influencers for the youth population in three of the main social networks. To this end, three specific objectives will be developed: to identify the most influential accounts at present; to carry out a sentiment analysis (polarity, agreement, irony, subjectivity) of the main contents of their publications with the greatest impact (number of likes); and to study the differences in the publications of the aforementioned accounts depending on the social network. Accordingly, the following hypotheses are proposed:

• H1. Within the widespread use of social networks among children and young people, YouTube and Instagram influencers attract more followers than Twitter influencers.

• H2. The content that most attracts teenagers differs according to the social network and its function. Twitter has an information and social awareness function, Instagram a space to present a self-idealised image of oneself, and YouTube playful and entertaining.

• H3. The content with the greatest impact finds justification in the theory of social comparison and social relations.

Dataset(s) and methods

Sample

The corpus consisted of 100 posts of Instagram, 100 tweets, and 100 YouTube videos, extracted from the accounts of international child and youth influencers with the largest numbers of followers, comprising 10 Tweeters, 10 Instagrammers and 10 YouTubers (Table 1). In this sense, the 10 publications with the highest number of likes from each influencer were chosen. To select the sample, the following were used: the platforms Hype Auditor (Instagram and YouTube) and Statista (Twitter), which provide information on the infant-juvenile influencers with the largest numbers of followers after applying the inclusion and exclusion criteria. However, it was necessary to apply inclusion and exclusion criteria to find child and youth influencers. The inclusion criteria were: a) accounts aimed at children and young people (the contents and topics worked on must be interesting for this age group and follow digital culture trends); b) content with minors (where the protagonists are children or teenagers); c) minors regarded as influencers (these channels must have a large number of followers). And, as exclusion criteria; d) family accounts of social networks (families that present their daily life, adventures and misadventures of their family nucleus with traditional content); and e) accounts aimed at educational content (accounts aimed at language learning, music, autonomy, such as Cantajuegosvevo). In this sense, today's child and youth influencers belong either to generation Z or centennials (born between 1994 and 2010) or to generation Alpha (born between 2010 and 2025).

Analysis and findings

The results returned by MeaningCloud expose diversity in the polarity of feelings among social networks (Figure 1).

The network with the most critical language is Twitter, presenting a higher rate of negative than positive polarity, in contrast to Instagram and YouTube, where the opposite occurs. While Instagram’s position has some similarity to Twitter, 85% of YouTube’s content was categorized as positive, with residual negative content. As for the other variables in terms of sentiment analysis (Figure 1), the confidence index on polarity, although high, varies according to the network: Instagram registered 99.51%, Twitter 98.57% and YouTube 80.98%. As for the remainder of the variables, there are differences. Subjectivity indicates that although the most followed content on Instagram and Twitter is of an objective nature, that is, describing a fact or product, YouTube has a high rate of subjectivity, that is, the most followed content is personal opinion. YouTube consistently presents a higher level of affectivity (disagreement) than Twitter and Instagram, that is, its language is loaded with emotion, especially pleasant feelings. Similarly, it is unsurprising that YouTube also presents a higher rate of irony, ironic content on Twitter and Instagram being minimal. Consequently, H1 is accepted, as YouTube and Instagram are more active than Twitter among the younger generation.

Pearson’s correlations show strong and moderate relationships between the sentiment analysis variables of the three social networks (Table 2). However, simple forward regression tests on polarity reveal that not all variables have the same weight on the three social networks (Table 3).

Initially the polarity of Instagram is explained by 27.1% (R2=0.271, p<.000), the role of subjectivity and the level of confidence being statistically significant. That is, the number of terms referring to pleasant or unpleasant feelings is explained by the degree of subjectivity (writing of a fact or product, or personal vision or opinion) used in a message, together with the level of confidence in the success of the polarity.

Similarly, the polarity of Twitter is explained by 39% (R2=0.390, p<.000), by the level of subjectivity and the level of trust. Although YouTube shows different behaviour, its polarity is explained in 79.3% (R2=0,739, p<.000) by the agreement variable and subjectivity. That is, its polarity is determined by the number of emotionally charged words and the use of terms that denote a personal opinion or vision (Table 3).

As for the semantic analysis of the language most used on social networks, it can be said that the total sample was composed of 24,664 words in English, Spanish, Portuguese and Russian. The search for so-called stop-words was applied automatically. These words are recognised as meaningless (regardless of the number of letters they have) and must be eliminated before the semantic study.

The distribution of the most used words is a function of the level of density, generating word clouds (Figure 2). In this sense, if we analyse in depth the most frequent terms, we find differences in the themes of each social network. On the one hand, Instagram exposes a universe of terms aimed at stimulating interaction with users’ followers: “answer”, “comment”, “love”, “follow”, “like”, “photo”. In this way, users seek to promote influence among their followers, demanding attention and greater interaction while presenting themselves as subjects’ worthy of following.

As far as Twitter is concerned, the most used terms are sustainability and adjectives that express social movements or actions, such as “climate”, “crisis”, “needs” and “person”. Two hashtags even appear: “#facetheclimateemergency” and “#flattenthecurve”. The use of ‘https’ is also surprising, as it shows how Twitter is not only a text space but also a video one. Finally, YouTube exhibits behaviour similar to that of Instagram, in that the most used words refer to a conversation between a YouTuber and their followers, in which the former repeatedly launches terms to maintain attention: “cool”, “look”, “just”, “really”. In other words, influencers respond to a monologue-like narrative in which the sender refers to their audience: the content most supported by the adolescent population varies according to social network. In short, H2 is fulfilled, as the three networks studied manifest differences in their most-shared content, Instagram being a space of self-representation, Twitter of awareness of social movements and ecology and YouTube of purely playful entertainment.

On the other hand, the stream graph (Figure 3) shows the behaviour of the five most-used words in the 20 publications of greatest impact. Instagram shows how the terms “amores” (Spanish language, the translation is loves) and “vou” (Portuguese language, the translation is you), which are both used to address followers, remain constant and are interrelated with the words “comment”, “reply” and “photo”. In other words, the publications that teenagers like the most have as their backbone those that ask for help to increase the Instagrammers’ own popularity.

On Twitter, there is a thematic line related to social and political actions (“leaders”) to stop climate change – “climate”, “crisis”, “#facetheclimateemergency” – mostly accompanied by videos and other “https” links. As far as YouTube is concerned, there is a strong degree of homogeneity, with the terms “like”, “gonna”, “okay”, “just” and “I’m’ succeeding each other in a regular way”. In other words, a narrative is presented in which the YouTuber interacts and seeks the complicity of their followers. In short, as in the case of Instagram, the publications with the greatest number of likes are those in which the influencers address their followers directly, seeking their direct interaction. Thus, H3 is fulfilled: the contents with the most impact are those where a parasocial relationship is established to influence followers.

In terms of hashtags, there are important differences between Twitter and Instagram. On Twitter there are two main topics: the school climate strike (“#FaceTheClimateEmergency” (30%), “#climatestrikeonline” (30%), “#fridaysforfuture” (7.50%), and “#schoolstrike4climate” (6.25%)) and COVID-19 and the social actions resulting from it (“#flattenthecurve” (6.25%), “#StayAtHome” (6.25%)). These two themes are strongly connected with the current social reality.

Even though Instagram presents greater variety and dispersion, it is striking how a significant percentage refers to brands: “#SHEIN” (12.50%), “#nyxcosmetics_en” (9.37%), “#Zalan-doStyle” (6.25%), “#LGG8X” (3.12%) and “#APPLE” (3.12%). In this case, main Instagram topics are linked with marketing of fashion and technology brands.

Discussion and conclusions

Social networks operate within adolescents’ everyday lives. The establishment of parasocial relationships between followers and influencers (Instagrammers, YouTubers and Tweeters) is of considerable importance for teenagers today. The differences between the social networks Instagram, Twitter and YouTube operate according to the different functions they perform, the results being consistent with previous publications (Throuvala et al., 2019; Vannucci & McCauley-Ohannessian., 2019).

Although all social networks are regularly used by the adolescent community (Anderson & Jiang, 2018; Smith & Anderson, 2018), the level of active participation seems to be mediated by the social network in question. In this regard, the results of fans and number of likes agree with previous studies indicating that YouTube is the most used platform among this population (Anderson & Jiang, 2018; Ofcom, 2017).

With respect to the sentiment analysis, similar results to previous investigations have been obtained. The contents of the study have considerable positivity or neutrality while also presenting significant subjectivity (Peres et al., 2020; Reyes-Menendez et al., 2018; Vizcaíno-Verdú & Aguaded, 2020). However, it should be noted that there are differences among the social networks, with Twitter being the most critical and objective and YouTube being the most positive and subjective. The results of the correlations and regressions specify how the polarity presents disparate actions. In the cases of Instagram and Twitter, the polarity is explained by “subjectivity” in moderate percentages. By contrast, the polarity of YouTube is the result of “agreement” and “subjectivity” in a high percentage. In summary, the sentiment analysis shows that, in the case of YouTube, messages are charged with “positive” emotions and personal views, as opposed to Twitter and Instagram.

On the other hand, we agree with Saura et al. (2019) that posts with a “real-world relationship” theme, especially those related to environmental policy violations, have a strong negative charge. In the same way, the categories body image, self-expression, travellers, digital life and startups have a positive or very positive polarity, being common in videos about the YouTubers’ lifestyles or in Instagram publications. Similarly, the semantic analysis of the word densities (word clouds), stream graphs and hashtags most used in publications with greater impact, has revealed significant differences among the social networks. Firstly, we agree with Blasco-García (2020) that posts do not present a transgressive image, but follow the fashion of the moment. It is necessary to make an important reflection on the use of the hashtag. The use of this technique increases visibility and virality, as can be seen in the results of this research, in line with previous studies (Lipsman et al., 2012; McGoogan, 2017).

However, there are profound differences between Twitter and Instagram in terms of the use of hashtags: whereas on Twitter its use responds to a social movement, on Instagram it pertains to fashion, cosmetics and technology brands, as also noted in previous research (Bakir et al., 2020; Boerman, 2020; Schouten et al., 2020). Indeed, Instagram is much more than a mirror to proclaim an ideal self, but rather a platform to engage an audience, based on communication that encourages interaction: “respond, picture, comment, love”. In other words, the results are consistent with Shane-Simpson et al. (2018), who demonstrate how this social network is used to capture an audience, while its outreach to adolescents renders it a marketing tool (Bakir et al., 2020; Boerman, 2020; Schouten et al., 2020).

Besides that, if interest is focused exclusively on word density (word clouds) and stream graphs, similar communicative features can be identified, although with different themes. At first glance, one can note how Instagram refers to a world of self-representation, where social comparison is generated through selfies, attending to a canon of beauty (Burnette et al., 2017; Schmuck et al., 2019; Verrastro et al., 2020; Yau & Reich, 2019). Followers show their emotional attachment to publications in which Instagrammers ask them to collaborate or increase their influence. On another note, Twitter deals with social changes and actions, which turn out to be the thermometer of social concerns in the case of adolescents, for instance climate change (Neu et al., 2019; Peres et al., 2020). Teenagers present the social value generated by this age group as well as their ability to reason and defend their ideas and values. YouTube collects action verbs needed for humorous storytelling by both gamers and lifestyle bloggers, consistent with Aran-Ramspott et al. (2018) and Castillo-Abdul et al. (2020). The most relevant publications are those in which the YouTuber asks for attention and collaboration. The use of this type of expression coincides with the results of Bhatia (2018) and Scannell (2000) in such a way as to show a communicative strategy to achieve commitment and active participation. The use of terms that facilitate the establishment of communication among Instagrammers (“respond”, “comenten” (translation comment), “loves”, “photo”, “vou” (translation you) and YouTubers (“I’m”, “gonna”, “okay”, “just”, “like”) responds to the need to establish a parasocial relationship.

In this way, an affective bond and emotional commitment is generated between influencer and follower (Bhatia, 2018), which is rewarded by followers giving “likes”. In this sense, these results coincide with previous studies that either explain this relationship or the extent to which these influencers normalise certain behaviours (Ashman et al., 2018; de-Bérail et al., 2019; Hartmann, 2016). In other words, the publications with the greatest scope are those in which a celebrity seeks their followers’ participation, which, in the eyes of authors such as de-Bérail et al. (2019) and Hartmann (2016) could be indicative of high rates of loneliness and social difficulties, leading them to seek refuge in parasocial relationships.

This study is not without its limitations. First, the context of the pandemic must be taken into account, as this element may have increased connectivity for adolescents. On the other hand, it was not possible to find studies with which to make a true comparison of the results due to the speed with which new communication mechanisms and platforms are incorporated. On the other hand, the tools used are constantly changing and developing. Therefore, it is necessary to replicate this research in the coming years with more up-to-date and cutting-edge software. Similarly, it is necessary to generate longitudinal studies involving the study of a larger number of publications. Finally, given the pseudonymisation and the margin for lies inherent to social networks, bigdata research in this field has a limitation in the study of attitudes. This is because we will never be able to access what these people really think and feel, but only their behaviour in networks. For this reason, it seems interesting to carry out a study of the behaviour of influencers and their followers on all their profiles on all the social networks they have, including emerging channels such as TikTok or Twitch. In this way, it would be possible to analyse whether it is a behaviour adapted to the characteristics of the network used, such as market demand, or a behaviour maintained in networks and over time, more similar to a study of attitudes. It would be interesting to carry out this study in a few years’ time and observe changes in behaviour between today’s adolescents and those in the future: will face-to-face social relationships have disappeared? Did social rules and social behaviour change in adolescents? Is the concept of friendship changing in adolescence? Do adolescents feel lonely and therefore establish parasocial relationships through social networks? Are adolescents’ capacities for social relationships diminishing? What role do social networks such as Instagram play in marketing and attracting teenagers as customers? And as advertisers or marketers? Will new behaviours emerge linked to new social networks, with new features, such as TikTok or Twitch? As for the direct implications of this research, it is necessary to highlight how it has allowed us to understand what aspects are most valued by the digital community of children and youth influencers. Likewise, the behavior of influencers in this age group is different depending on the social networking sites used. (1)