Revista Latina de Comunicación Social. ISSN 1138-5820
“Trusted Media” on YouTube: volume and visibility of public media in search results
Los Trusted Media en YouTube: volumen y visibilidad de los medios públicos en los resultados de búsqueda
Autonomous University of Barcelona. Spain.
EUNCET Business School. Spain.
Òscar Coromina
Autonomous University of Barcelona. Spain.
Emili Prado
Autonomous University of Barcelona. Spain.
ABSTRACT
Introduction: This research aims to describe the treatment of news content coming from public media in YouTube search results. Methodology: To this end, by using the platform's API, search results were extracted for a set of 4 keywords over 60 days. Results: The analysis indicates that public media are a minority (3,70%) among the channels appearing in the search results, and they do not achieve better positions in the search results ranking. However, the algorithm proportionally selects more content from public media than from other sources, causing an overrepresentation of this type of media. Content published by public media also reappears more frequently in search results and lasts longer. Private media also receive slight benefits, at the expense of native channels and other sources, which are underrepresented in the search. Conclusions: The results suggest that YouTube selectively favors media outlets in search results, especially public ones, selecting more content from them, which indicates a curation and moderation process of content in the search aimed at limiting the appearance of misinformation. With this, YouTube would be trying to redirect audience attention to safe sources in contexts where misinformation exists, implying the invisibilization of other sources and not applying direct censorship to potentially harmful content.
Keywords: YouTube, Public media; Legacy media; Digital methods; Disinformation; YouTube API; Trusted media.
RESUMEN
Introducción: Esta investigación tiene como objetivo describir el tratamiento que reciben los contenidos informativos procedentes de medios públicos en los resultados de búsqueda de YouTube. Metodología: Para ello, haciendo uso de la API de la plataforma, se han extraído los resultados de búsqueda para un conjunto de 4 palabras clave a lo largo de 60 días. Resultados: El análisis indica que los medios públicos son minoría (el 3,70%) entre los canales que aparecen en el buscador, y no obtienen mejores posiciones en el ranking de resultados de búsqueda. Sin embargo, el algoritmo selecciona proporcionalmente más contenido de medios públicos que del resto de fuentes, lo que causa una sobrerrepresentación de este tipo de medios. Los contenidos publicados por medios públicos también reaparecen en más ocasiones entre los resultados de búsqueda y perduran más tiempo. Los medios privados también se ven ligeramente beneficiados, en detrimento de los canales nativos y otras fuentes, que se ven infrarrepresentadas en el buscador. Conclusiones: Los resultados sugieren que YouTube favorece selectivamente los medios de comunicación en los resultados de búsqueda, especialmente los públicos, seleccionando más contenido de ellos, lo que evidencia un proceso de curación y moderación de los contenidos en el buscador que persigue limitar la aparición de desinformación. Con esto, YouTube estaría tratando de desviar la atención de las audiencias hacia fuentes seguras en contextos donde existe desinformación, lo que implica invisibilizar otras fuentes y no aplicar una censura directa a los contenidos potencialmente dañinos.
Palabras clave: YouTube; Medios públicos; Medios convencionales; Métodos Digitales; Desinformación; API YouTube; Medios confiables.
1. INTRODUCTION
YouTube has managed to position itself as one more source in the range of options available in terms of the consumption of journalistic information, and in doing so has taken a part of the audience away from the “traditional” media (Casero-Ripollés, 2012; Ofcom, 2024). This shift of attention to YouTube has driven the landing on the platform of countless newspapers, radio and television channels, both public and private (Santín, & Álvarez-Monzoncillo, 2020). The presence of these media on YouTube is not only a way to maintain the relevance, impact or influence they have traditionally held, but also opens a new way to monetize journalistic production and give a second life to content previously distributed through other channels.
As these media participate in YouTube, professional material competes for visibility with content generated by amateur users and native professionals of the platform, the well-known youtubers (De-Aguilera-Moyano et al., 2019). The logics that guide this competition for visibility on YouTube are not different from those traditionally used by the media industry to establish what is relevant: this platform operates under the paradigm of the attention economy (Franck, 2019), a principle that leads to prioritize in the interface those contents that achieve greater audience retention and, therefore, offer greater profitability by allowing greater advertising exploitation. This scheme is known to those who are familiar with television, where channels traditionally compete to win shares by making a careful selection of content. However, while in television there are human teams with the capacity to judge, in addition to the profitability, the quality of the content, its ideological line, its social relevance, its cultural and informative value, in YouTube the task of selecting and ranking the content is left in the hands of a set of algorithms that disconnect the content from its meaning.
This paper focuses on the presence of the media on YouTube. Specifically, it focuses on determining the treatment received by news content from public media in the search results of this platform, while comparing it with videos published by channels belonging to private media and other types of sources. To do this, the search results are analyzed over a period of 60 days, using 4 combinations of keywords linked to controversial issues and where traditionally there has been disinformation, conspiracy or denialism. After classifying the contents and channels present in the search engine, their volume (amount of contents), their position in the ranking and social interactions (reproductions, likes and comments) are comparatively analyzed.
The results of the analysis indicate that YouTube's search algorithm tends to overrepresent media in the search engine, especially publicly funded media, which get more space in the search results. In turn, audiovisual content from public media lasts longer in the results ranking and reappears on more occasions, so it has a longer lifespan. This is despite the fact that content published by public media has less capacity to attract attention (views) and a lower amount of social interactions (likes and comments).
1.1. YouTube and the media system
Fitting YouTube into the current media ecosystem remains problematic. Although the platform's narrative styles and formats are not different from those of the television industry, its nature, codes and operational logic differ from the predecessor technology, moving from linear broadcasting and synchronous consumption to an environment of choice with an overabundance of audiovisual offerings, something similar to a logistics center for the distribution of audiovisual content of all kinds and origins (Prado, 2022). The complexity of locating YouTube in reference to television leads to labeling the platform under the heading “new media”, a catch-all in which it coexists with other digital services: social networks such as Twitter, TikTok, Instagram or Facebook, and subscription video services (SVOD) and over the top (OTT) distribution such as Netflix or HBO.
Yet despite the taxonomy, YouTube is an indistinguishable part of the media ecosystem: with more than 2 billion unique users per month, it is arguably the most consumed medium and audiovisual distribution channel on a global scale. So great is its industrial and cultural weight that to understand the current media landscape it is necessary to abandon the distinction between “old media” and “new media”, and approach the hybrid media system described by Chadwick (2017), a system where users can intervene and actively participate in the construction of the current news, by changing the traditional flows of information circulation, adding layers of bidirectionality and balancing the forces involved in the construction of the narrative, social reality and shared everyday life (Hadis, 1976; Couldry, 2019; Couldry, & Hepp, 2018). This dynamic is also known as “participatory culture” (Jenkins et al., 2015), and can be seen as a reconquest of the public sphere, hitherto monopolized by traditional media corporations and cultural industries.
These same participatory mechanisms, inherent to the very nature of social networks and platforms, can also be leveraged by players that beyond a disinterested, casual or amateur participation, operate in an organized way to influence the public sphere (Coromina, & Padilla, 2018; Khaldarova, & Pantti, 2016). This is possible given that, in the hybrid media system, information circulates making use of spaces and technologies in which there is no moderation or strict control. Lewis (2018) describes this as “alternative influence networks”: coordinated individuals or groups that achieve certain informational authority, and use these “new” spaces to disseminate narrative frames that usually differ from the official narrative. Twitter, Facebook and most especially YouTube play a fundamental role, for example, in the circulation of narratives that articulate the “alternative right” vote, both in Europe and in the United States (Allcott, & Gentzkow, 2017; Rodríguez-Serrano et al., 2019; Hosseinmardi et al., 2021; Cordero et al., 2023).
1.2. Disinformation on YouTube
YouTube has not been a platform oblivious to the problem of misinformation; in fact, it is a particularly prolific phenomenon in its environment (Castaño, 2022; Hussein et al., 2020). In 2021, Neal Mohan, YouTube's product manager, published on the platform's official blog an article entitled “Perspective: tackling misinformation on YouTube,” offering insight into what the company's policy on misinformation was (Mohan, 2021). In the article, Mohan estimates the magnitude of the problem: around 0,16%-0,18% of the content on YouTube conflicts with the platform's content policies, and 77% of potentially harmful content is removed before reaching 100 views. In the same article, Mohan recounts that in order to identify disinformation around COVID-19, disinformation was considered anything that contradicted the consensus of experts linked to the US National Center for Diseases or data from the World Health Organization (WHO). YouTube's help pages also have a section dedicated to disinformation (YouTube, n.d.-a). These pages detail that the platform uses groups of experts and external evaluators to assess the quality of the content. These human teams establish the criteria with which machine learning systems are trained to detect and eliminate potentially harmful content on a massive scale.
However, there are times and contexts in which it is not possible to establish which version of the facts is true. Faced with this situation, YouTube chooses to provide users with content from trusted sources. The platform does not detail the criteria used to establish the degree of trustworthiness of the sources, although they indicate that this process incorporates data from Google News, as well as the level of expertise of the source, the relevance and timeliness of the event (YouTube, n.d.-a). For controversial and conspiratorial content linked to scientific issues, YouTube also provides links to Wikipedia or online encyclopedic resources (see figure 1), which offer the user a much broader and reliable context (Matsakis, 2018).
The platform has also created moderation mechanisms that allow detecting “questionable” content. Videos that make use of explicitly sexual language or narratives, swearing, foul, extreme language, and sexual references or connotations are pursued, for example (YouTube, n.d.-b). These moderation criteria are not a trivial matter: among youtubers it is rumored that this type of clauses allow shadow banning, a mechanism by which YouTube penalizes, invisibilizes or censors certain content that does not conform to the moral standards of the platform, or to the ideological schemes of its moderators (Bishop, 2019). The issue of moderation on YouTube is a central debate, involving the right to freedom of speech and its limits. In the words of YouTube CEO Susan Wojcicki: Where do you draw the boundaries of free speech and, if you draw them too narrowly, are you censoring voices in society that should be heard? (Saner, 2019). Although the platform is taking steps forward in the fight against disinformation, it is sometimes complex to accurately identify potentially harmful content, not only because of the scale at which content needs to be moderated, but also because of the ideological and cultural biases of the moderators themselves, which can lead to label ironic, critical and ideologically challenging content as disinformation (Boyd, 2017).
1.3. Trusted Media on YouTube
In this context, the participation of professional media on YouTube, especially those with a certain journalistic reputation, is sponsored by the platform itself, which needs the presence of these players to provide the environment with robustness and generate a climate of information security for audiences. The presence of public and private media allows YouTube to delegate the editorial function and information gatekeeping to those who have traditionally held this role: the media specialized in the production of information. Their presence is very convenient for YouTube, since it is materially impossible for the platform to moderate the enormous amount of content published daily, and it is even more complex and compromising to establish what is disinformation. The lack of informative referents on YouTube has proved to be pressing during the pandemic (Donzelli et al., 2018; Knuutila et al., 2020; Sued, 2020), but also in electoral contexts or scientific issues, where denialism and conspiracy theories proliferate (Ribeiro et al., 2020; Tufekci, 2018). Therefore, the arrival on YouTube of newspapers, televisions, radios, magazines and media of all kinds and types of funding has allowed the platform to introduce some informative order and offer a much safer environment for its users, but it has also been something very convenient to solve the ethical and moral dilemma of moderating and drawing limits to disinformation, limits that could end up affecting freedom of speech.
Figure 1. YouTube interface where the contextual information indicator of the media “DW español” is observed.
Source: Elaborated by the authors.
In this quest to offer YouTube users a series of indicators to determine the degree of “trust” of sources and content, in 2018 the platform announced the introduction of a contextual message with additional information about the media outlet that is publishing the content (Gold, 2018). This message (see figure 1) aims to provide viewers with information about who publishes the news, its source of funding and a link to Wikipedia, where it is possible to obtain more data about the media (YouTube, 2024). This label is only awarded to media partially or fully financed with public funds, and can be found in videos published by Deutsche Welle, Radio Televisión Española, BBC News, EFE, Al Jazeera or France 24, among many others. Although YouTube does not clarify with what criteria this set of media is selected, nor is it made explicit that the label indicates a higher degree of informative trust, the decision to label only the contents of this type of media aligns with what Fotopoulos (2023) names “trusted media”.
With the presence of these media on the platform, a new competitor appears in the race to catch and monetize attention. YouTube must determine what position these contents should occupy in search results, a selective process involving a cascade of algorithms (Covington et al., 2016). As Gillespie (2014) points out, it is not innocuous how algorithms present and order results; it can affect consumer choice and thus influence decision making. Behind these ordering mechanisms may hide commercial intentions or a will to influence public opinion, all under a patina of technological neutrality and objectivity, which should lead us to reconsider the role of the platform in a much broader sociological context (Bryant, 2020; Juneja, & Mitra, 2020).
1.3.1. Background
Previous work has delved into YouTube search result ranking and, more specifically, the effects of the algorithm on this component of the platform. Rieder et al. (2018) analyze how these rankings evolve over 44 days, collecting the top 20 search results for 7 keywords linked to current news. Thereby, the team managed to identify different morphologies in the rankings: totally stable, partially stable and totally changing. Although the article warns of the presence of a large number of professional media competing with native channels of the platform for the different positions in the ranking, the team does not delve into the characteristics of these media, nor does it make any distinction between them and the rest of the channels. Nor do they analyze what type of information sources achieve a better position in search results. Even so, their work shows that the search results ranking criteria take into account complex issues that go beyond basic social interactions (comments, likes and views), something that the authors call “ranking cultures”.
On the other hand, Padilla (2022) analyzes during 20 days the first 25 search results for 6 keyword combinations linked to uninformative issues. In addition to observing and confirming the same ranking morphologies detected by Rieder et al. (2018), Padilla (2022) thematically categorizes the contents present in the search results. This allows determining that informative contents obtain slightly advantageous positions in the ranking, something that also translates into greater visibility in the search results for channels of a professional nature. However, this research does not distinguish between public and private media, so it is not possible to conclude whether YouTube gives differential treatment to media according to their source of funding.
For this reason, the aim of this article is to establish whether YouTube's algorithm prioritizes content from public media in search results over content published by private media and other types of sources. To this end, it was analyzed and compared the volume (amount of content), the position in the search engine, and the interactions (views, likes, comments) of the content found in the search results ranking, focusing on the type of channel, and explicitly distinguishing between YouTube channels belonging to public media, private media, native channels and other sources.
2. METHODOLOGY
For the analysis, four keyword combination ns have been selected that target controversial, conspiracy or misinformation topics: “vaccines”, “climate change”, “chemtrails” and “flat earth”. The selection of these words follows the line of previous works, in which the focus is on misinformation topics (Cantó, 2018; Donzelli et al., 2018; Muñoz-Pico et al., 2021; Padilla, 2022; Paolillo, 2018; Shaheed, 2019; Sued, 2020). For each of these keywords, 6 daily search result extractions are performed (one extraction every 4 hours), throughout the period from 05/20/2023 to 07/20/2023 (a total of 60 days). In each extraction, the first 50 search results and their metrics are collected. The vertical (number of results) and horizontal (time period) range exceeds that of previous work: 25 results over 20 days in Padilla (2022) and 20 results over 44 days in Rieder et al. (2018).
Extractions are created by a Python script created ad-hoc, which runs in an automated way and keeps the search results in a MySQL database. This script makes requests to the YouTube API, making use of the “search: list” method (Google for developers, 2023). The search is parameterized to limit the contents to the Spanish language. The use of the API as a way of accessing the search results allows a large amount of information to be collected in a very short time, so that all keyword combinations are collected at the same time. This method also makes it possible to perform anonymized searches, which prevents the search results from being biased by the user's consumption history. The search result ranking ordering parameter used in the API query is “relevance”, which is the same ordering system by which YouTube delivers search results in its user interface by default. This data collection method is the same one used in Rieder et al. (2018), & Padilla (2022).
In total, 360 feeds were performed for each keyword combination (6x60) during the 60-day analysis period, which translates into a total of 1.440 extractions for all 4 keywords (360x4). The resulting database is composed of 71.096 analysis units (1.440x50). The margin of error in the capture process was 1,25%, which means 4 failed extractions since the YouTube API was not available (due to technical failures of the platform) at specific moments. This margin of error has no effect on the results.
The YouTube channels in the final databank were manually classified under 4 criteria: “public media”, “private media”, “native channels” and “other”. This process was carried out by visualizing the channels and contrasting the source of funding. The group “public media” includes all YouTube channels with total or partial public funding, such as RTVE, Deutsche Welle, or BBC, among others. The “private media” heading includes media financed with private capital, such as CNN or Antena 3. Channels belonging to an individual user, regardless of their professionalism, have been classified as “platform-native” channels. Finally, the group “others” includes all those channels that belong to, for example, public institutions, private companies, political parties, foundations, non-profit organizations, etc.
For this research, the quality, format or purpose of the content and YouTube channels included in the search results are not evaluated. It is not part of the research objectives to determine the presence of misinformation or to judge the quality of misinformation.
3. RESULTS
3.1. Channel population
During the period analyzed, 864 YouTube channels contributed some content to the search results for the set of keywords. Table 1 shows that 629 (73,05%) are native channels of the platform with varying degrees of professionalism, a group ranging from large content creators (youtubers) to amateur users who dabble sporadically. This categorization aims to draw a clear dividing line between this typology of channels and media companies (media outlets) specialized in the production of information that participate in the platform.
On the other hand, 126 channels (14,63%) have been identified as privately financed media, 32 channels (3,72%) correspond to publicly financed media and, finally, 77 channels (8,94%) belonging to other categories (companies, political parties, foundations, etc.) have been identified.
Table 1. Channel population classified by type.
|
|
Channels |
|
|
Quantity |
% |
|
|
Public media |
32 |
3,70% |
|
Private media |
126 |
14,58% |
|
User channels |
629 |
72,80% |
|
Others |
77 |
8,91% |
Source: Elaborated by the authors.
By focusing the attention specifically on YouTube channels operated by media outlets, the 32 channels identified as fully or partially publicly funded media outlets have a label on their videos that identifies them as publicly funded media, with a link to Wikipedia where it is possible to read about the media outlet and its record. This indicator is identical to the one shown in Figure 1. The remaining 126 media, which correspond to privately financed media, do not obtain in the interface of their videos any label that allows them to be identified as a media outlet.
3.2. Volume and position in search results ranking
Each search results ranking position can be considered as an opportunity to rank and make visible a content produced by a channel. Altogether, the total number of ranking positions available for this research is 71.096, and the algorithm must fill these positions by selecting content that matches the search criteria.
As shown in table 2, the YouTube search algorithm does not distribute the positions of this ranking proportionally for each channel typology. Publicly funded media have occupied 13,45% of the search results, although population-wise they are 3,70% of the detected channels. Their presence in the search results, in absolute terms, is close to that of the private media, which occupy 14,59% of the positions in the ranking, although the latter account for 14,58% of the channels in terms of population. On the other hand, native channels occupy 66,03% of the search results, being 72,80% of the channels. Finally, the rest of the sources occupy 5,93% of the ranking positions, although they represent 8,91% of the channels.
Table 2. Representativeness of each type of source in search results and average position.
|
|
Ranking positions occupied |
Ranking position |
|
|
Quantity |
% |
Average |
|
|
Public media |
9.560 |
13,45% |
23,45 |
|
Private media |
10.371 |
14,59% |
26,76 |
|
Native channels |
46.947 |
66,03% |
25,34 |
|
Others |
4.218 |
5,93% |
29,41 |
Source: Elaborated by the authors.
Therefore, it can be stated that there is an overrepresentation of public media in the search results ranking. The private media make up 14,58% of the channel population, and occupy 14,59% of the ranking positions, indicating a proportional presence. Nevertheless, public media occupy 13,15% of the ranking positions being 3,70% of the channels. Therefore, a small amount of public media have occupied an important part of the search results. On the other side of the scale, platform-native channels and other sources are underrepresented: the space they occupy in the results ranking does not correspond to their population.
This data also suggests that there is a high turnover of content published by native YouTube channels, which increases the number of distinct channels that have at some point participated in these rankings. According to the same logic, a small number of media (both private and public) manage to remain and reappear in the ranking on more occasions (as will be seen below).
As for the position obtained by the contents published by each source in the ranking, public media have a significantly better average position. This means that their videos are closer to the first position and, therefore, are more visible to the user, increasing their probability of consumption. However, the difference between the various sources is not wide enough to conclude that YouTube provides this type of channel with better positions in the ranking.
3.3. Algorithmic content selection and lifespan
Combined, the 864 channels identified in the previous section contributed a total of 1.329 different videos. During the 60 days analyzed, these 1.329 videos appeared multiple times among the search results, varying or maintaining their position in the ranking, either continuously or intermittently. It is important to remember that the task of selecting and positioning these 1.329 videos to fill the 71.096 result ranking slots falls to the search algorithm. The fact that the algorithm filled 71.096 ranking spaces with 1.329 videos suggests that the search results are stable over time, with an average of 53 appearances per video. The average duration in ranking for the videos as a whole is 8 days.
As shown in table 3, there are 138 (10,38%) videos from public media and 250 (18,81%) from private media outlets out of the 1.329 different videos. Another 839 videos (63,13%) came from native channels and 102 (7,67%) were published by other sources. These data indicate that YouTube's algorithm selects (in relation to its population) more content coming from public media than from private media, although the latter (as a whole) have managed to position a greater amount of different content.
Table 3. Repetition rate and lifespan of the contents of each type of channel.
|
|
Number of different videos |
% Different videos |
Average number of videos per channel |
NUM. Appearances |
Average repetition rate |
Lifespan (days) |
|
Public media |
138 |
10,38% |
4,31 |
9.560 |
69 |
12 |
|
Private media |
250 |
18,81% |
1,98 |
10.371 |
41 |
7 |
|
Native channels |
839 |
63,13% |
1,33 |
46.947 |
41 |
7 |
|
Others |
102 |
7,67% |
1,32 |
4.218 |
56 |
9 |
Source: Elaborated by the authors.
However, in turn, private media also receive more attention from the algorithm than the platform's native channels, with private media making up 14,58% of the channel population, but contributing 18,81% of the content. On average, the algorithm will select 4,31 videos per public media channel, compared to 1,98 videos per private media channel, 1,33 per native channel and 1,32 for other types of channels.
The fact that the 1.329 different videos identified during the period analyzed appear in successive feeds suggests that the contents have a repetition rate and a lifespan. Considering the number of appearances of these videos in successive feeds in the search results ranking (see table 3), content from a public media has a repetition rate of 69 times, compared to 41 times for private media and native channels. Consequently, the survival rate for content from public media is 12 days, compared to 7 days for content from private media and native channels.
Based on this data, it is possible to state that the search algorithm tends to keep videos from public media longer in the search results. It also selects proportionally more content from this type of source: more than 4 videos per channel. The algorithm also selects a higher amount of private media content, 1,98 videos per channel, although these videos are not maintained for the same amount of time among the search results and, therefore, there is a higher rotation.
While this behavior of the algorithm could be a reflection of the productivity of each YouTube channel, the fact that content from public media remains in the search results for 12 days suggests that the algorithmic choice is not linked to productive cycles of the media, as they tend to the constant resonance of current news, and therefore a higher rotation of content would be expected.
3.4. Interactions
The work published by Rieder et al. (2018) conceptualizes different “ranking cultures”, a way of understanding search results that goes beyond the popularity metrics that are usually analyzed (likes, comments, views), and that manifests the existence of much more complex algorithmic criteria intervening in the content selection process. However, although these metrics do not intervene directly in the ranking, they still provide information on the interaction that users have had with the content.
Table 4. Average number of views, likes and comments by type of channel.
|
|
Average views per video |
Average number of Likes per video
|
Average number of comments per video |
|
Public media |
848.840 |
10.515 |
1.855 |
|
Private media |
1.023.269 |
13.604 |
1.608 |
|
Native channels |
2.459.078 |
80.563 |
3.258 |
|
Others |
488.617 |
15.909 |
810 |
Source: Elaborated by the authors.
Table 4 shows that content published by YouTube channels belonging to public media obtain less audience attention when compared to private media and user-generated content channels. The average number of views of content published by a native YouTube channel is three times that of content published by public media, and twice that of private media. Something similar happens with the number of likes and comments: native channels far exceed the figures for private and public media.
These data confirm, first of all, what Rieder et al. (2018) point out in their research, and that is that basic interaction metrics are not an indicator for determining what is relevant for the algorithm, and therefore do not decisively impact the ranking of search results. It also confirms something that YouTube's own engineers have openly stated: users do not have a special predilection for content from media with journalistic “authority”, despite the fact that these appear (as seen in previous sections) more frequently in the search engine. Consumption tends to favor content generated by youtubers and amateurs, a behavior that the platform itself has already identified and recognized (Parker, 2020).
4. CONCLUSIONS
YouTube has drawn a line that divides the informative content coming from channels operated by public media and the rest of the channels that participate in the platform. At the time of this research, only public media have the distinctive label that informs users about the type of source that broadcasts the information (see example in figure 1). This label does not explicitly advertise a higher quality of information, but it is inevitable that this element is considered by users as an indicator of trust. This interface label is closely linked to what Fotopoulos (2023) refers to as Trusted media, but it is important to note that the public funding model alone does not guarantee independence, impartiality or news quality, nor is it indicative of the degree of trust that the public should place in the media.
Based on this initial distinction, the aim of this article is to establish whether YouTube's algorithm prioritizes content from public media in search results over content published by private media and other types of sources. For this purpose, content volume, ranking position, and interactions (views, likes, comments) of the videos that emerge in the search results were analyzed and compared, focusing on channel typology, and distinguishing between YouTube channels belonging to public media, private media, native channels and other sources.
The results of this analysis indicate that channels belonging to public media are a minority population in the search results. Thirty-two channels belonging to public media were found among the search results, which represents 3,70% of the channels that the algorithm deemed relevant for the searches performed throughout the entire period analyzed. On the other hand, 126 channels belonging to private media were found, representing 14,58% of the population of channels that at some point appeared in the search engine.
These 32 channels (3,70%) occupied 13,45% of the search result ranking positions. Therefore, YouTube's search algorithm tends to select proportionally more content from public media than from other sources. In addition, videos from public media reappear in search results more often and last longer. Such behavior leads to an overrepresentation of the public media among the search results, which, although they do not obtain a more advantageous average position than the rest of the sources, they do manage to occupy a large amount of space in the ranking, and to do so for a longer period.
Although not to the same extent, private media (14,58% of channels) are also slightly benefited by the algorithm. When compared to platform-native channels and other source typologies, private media have a higher probability of their content appearing in the search results ranking. They manage to occupy 14,59% of the positions in the search engine, a figure proportionate to their population. Finally, the platform's native channels and other sources are underrepresented, as the percentage of space they occupy in the search engine is below their channel population (see table 1, 2, and 3).
The data suggest that the selection made by the algorithm is not necessarily linked to the frequency of publication and the productive capacity of each channel typology. On the one hand, in both media categories (public and private) there are large media corporations with the capacity to compete with each other. On the other hand, the algorithm has selected a higher total amount of videos from private media. These private media have also achieved a higher turnover of their content in the search engine, suggesting a strong link to current news. In contrast, public media (although overrepresented) position a smaller amount of videos overall, but achieve greater stability and survival. Indeed, this aging suggests the content is not necessarily subject to current events.
Finally, audience views and social interactions with these contents have been analyzed. The results indicate that users prefer to consume content produced by native channels and private media (see Table 4). This finding confirms that the algorithmic criteria for content selection goes beyond basic social interactions (Rieder et al., 2018), but also calls into question the principles of the attention economy. With the overrepresentation of public media YouTube is populating search results with content that gets lower audience attention, thereby losing the opportunity to position other videos with greater ability to attract and monetize audiences' attention.
All data indicate that YouTube is performing a content curation task in search results, intervening directly to draw users' attention to safe content, with the aim of avoiding controversy or new reputational crises linked to the dispersion of misinformation. This work is carried out by overrepresenting the media, especially public media, and underrepresenting the platform's native channels. Responsibility and accountability is a central element in this issue, and it is easy to offload informational responsibility onto these external actors, as both public and private media are recognizable organizations.
The presence of the media on YouTube should also be understood as a symptom of the acceptance of change and the adaptation of large media corporations to a new audiovisual ecosystem. It is a demonstration that their languages, formats and productive dynamics (what is understood as “television”) are not limited to the original technology, but are components of an audiovisual expression that admits multiple forms of distribution (Prado, 2022). YouTube's identification of content from public or semi-public media with a specific label should be interpreted as a self-interested move by the platform in the face of its manifest inability to offer a controlled information environment. To some extent, this can be understood as a recognition by YouTube that native content creators (youtubers) are not recognized by default as having the capacity to create content with the quality, credibility and trust that is recognized to the media, both public and private, and that youtubers alone cannot build a complete and pluralistic account of current news.
This research is limited to a period of 60 days and 4 keyword combinations linked to uninformative or controversial issues. Although its scope exceeds that of previous works, it is necessary to further investigate the issue, expanding the number and thematic variety of keywords, as well as the period of analysis. This would make it possible to extend the platform's performance to other areas, such as audiovisual entertainment, where public media also actively participate through their YouTube channels. It would also be necessary to monitor and compare the behavior and routines of YouTube channels belonging to public and private media, a work that would help to better understand to what extent their production processes affect the visibility of their content on the platform, in addition to providing information on their productive capacity.
Finally, it is considered necessary to investigate the effects that may be caused by the fragmentation and decontextualization of public service news content on YouTube, where it is chopped up to fit the consumption rhythms and narrative styles of this platform. This may encourage audiences to consume unstructured information, far from the editorial order that characterizes formal spaces: a television news program, a newspaper or a radio bulletin.
5. REFERENCES
Allcott, H., & Gentzkow, M. (2017). Social Media and Fake News in the 2016 Election. Journal of Economic Perspectives, 31(2), 211-236. https://doi.org/10.1257/jep.31.2.211
Bishop, S. (2019). Managing visibility on YouTube through algorithmic gossip. New Media & Society, 21(11-12), 2589-2606. https://doi.org/10.1177/1461444819854731
Boyd, D. (March 27, 2017). Google and Facebook Can’t Just Make Fake News Disappear. Wired. https://www.wired.com/2017/03/google-and-facebook-cant-just-make-fake-news-disappear/
Bryant, L. V. (2020). The YouTube Algorithm and the Alt-Right Filter Bubble. Open Information Science, 4(1), 85-90. https://doi.org/10.1515/opis-2020-0007
Cantó, P. (2018). Ni te fumigan ni quieren provocar sequía: vuelve la conspiración de los 'chemtrails'. El País. https://verne.elpais.com/verne/2018/03/02/articulo/1520005564_270533.html
Casero-Ripollés, A. (2012). Beyond Newspapers: News Consumption among Young People in the Digital Era. Comunicar, 20(39), 151-158. https://doi.org/10.3916/C39-2012-03-05
Castaño, J. (January 14, 2022). Los bulos y las teorías negacionistas campan a sus anchas por YouTube. La Vanguardia. https://bit.ly/3V2J678
Chadwick, A. (2017). The Hybrid Media System: Politics and Power (2nd ed.). Oxford University Press.
Cordero, G., Alonso-Muñoz, L., & Casero-Ripollés, A. (2023). Believe it or not. The role of ideology as a predictor of trust in mainstream and social media during COVID-19. Revista Latina de Comunicación Social, 82, 1-20. https://doi.org/10.4185/rlcs-2024-2141
Coromina, Ò., & Padilla, A. (2018). Análisis de las desinformaciones del referéndum del 1 de octubre detectadas por Maldito Bulo. Quaderns del CAC, 21(44), 17-26.
Couldry, N. (2019). Media: Why it matters. John Wiley & Sons.
Couldry, N., & Hepp, A. (2018). The mediated construction of reality. John Wiley & Sons.
Covington, P., Adams, J., & Sargin, E. (2016). Deep Neural Networks for YouTube Recommendations. En S. Sen, & W. Geyer, RecSys '16: Proceedings of the 10th ACM Conference on Recommender Systems (pp. 191-198). https://doi.org/10.1145/2959100.2959190
De-Aguilera-Moyano, M., Castro-Higueras, A., & Pérez-Rufí, J. I. (2019). Between broadcast yourself and broadcast whatever: YouTube’s homepage as a synthesis of its business strategy. Profesional de la Información, 28(2). https://doi.org/10.3145/epi.2019.mar.06
Donzelli, G., Palomba, G., Federigi, I., Aquino, F., Cioni, L., Verani, M., Carducci, A., & Lopalco, P. (2018). Misinformation on vaccination: A quantitative analysis of YouTube vídeos. Human Vaccines and Immunotherapeutics, 14(7), 1654-1659. https://doi.org/10.1080/21645515.2018.1454572
Fotopoulos, S. (2023). Traditional media versus new media: Between trust and use. European View, 22(2), 277-286. https://doi.org/10.1177/17816858231204738
Franck, G. (2019). The economy of attention. Journal of Sociology, 55(1), 8-19. https://journals.sagepub.com/doi/10.1177/1440783318811778
Gillespie, T. (2014). The Relevance of Algorithms. En T. Gillespie, P. J. Boczkowski, & K. A. Foot (Eds.), Media Technologies: Essays on Communication, Materiality, and Society. The MIT Press. https://doi.org/10.7551/mitpress/9780262525374.003.0009
Gold, H. (February 3, 2018). YouTube to start labeling vídeos posted by state-funded media. CNN. https://money.cnn.com/2018/02/02/media/youtube-state-funded-media-label/index.html
Google for developers. (2023). Search:list. https://bit.ly/444tEJZ
Hadis, B. F. (1976). La construcción social de la realidad [Reseña del libro La construcción social de la realidad, por P. L. Berger y T. Luckmann]. Desarrollo Económico, 15(60), 641-643. https://doi.org/10.2307/3466656
Hosseinmardi, H., Ghasemian, A., Clauset, A., Mobius, M., Rothschild, D. M., & Watts, D. J. (2021). Evaluating the scale, growth, and origins of right-wing echo chambers on YouTube. Proceedings of the National Academy of Sciences, 118(32), 1-8. https://doi.org/10.1073/pnas.2101967118
Hussein, E., Juneja, P., & Mitra, T. (2020). Measuring Misinformation in Vídeo Search Platforms: An Audit Study on YouTube. Proceedings of the ACM on Human-Computer Interaction, 4, 1-48. https://doi.org/10.1145/3392854
Jenkins, H., Ito, M., & Boyd, D. (2015). Participatory Culture in a Networked Era: A Conversation on Youth, Learning, Commerce, and Politics. John Wiley & Sons.
Juneja, P., & Mitra, T. (2020). Investigating the impact of personalization on misinformation present in algorithmically curated content in YouTube.
Khaldarova, I., & Pantti, M. (2016). Fake News: The narrative battle over the Ukrainian conflict. Journalism Practice, 10(7), 891-901. https://doi.org/10.1080/17512786.2016.1163237
Knuutila, A., Herasimenka, A., Au, H., Bright, J., Nielsen, R., & Howard, P. N. (2020). Covid-related misinformation on YouTube: The spread of misinformation vídeos on social media and the effectiveness of platform policies. COMPROP Data Memo. https://n9.cl/kqdpr
Lewis, R. (2018). Alternative Influence: Broadcasting the Reactionary Right on YouTube. Data & Society. https://datasociety.net/library/alternative-influence/
Matsakis, L. (March 13, 2018). YouTube Will Link Directly to Wikipedia to Fight Conspiracy Theories. Wired. https://www.wired.com/story/youtube-will-link-directly-to-wikipedia-to-fight-conspiracies/
Mohan, N. (August 25, 2021). Perspective: Tackling Misinformation on YouTube. YouTube Official Blog. https://blog.youtube/inside-youtube/tackling-misinfo/
Muñoz-Pico, H. P., León Anguiano, B., & García Martínez, A. N. (2021). Representación del cambio climático en YouTube: Un análisis cuantitativo de los vídeos más populares. Palabra Clave, 24(1), 1-32. https://doi.org/10.5294/pacla.2021.24.1.5
Ofcom. (2024). Children and parents: media use and attitudes report 2023. https://n9.cl/bvvk75
Padilla, A. (2022). Análisis del fenómeno de la desinformación en YouTube: Volumen, visibilidad y elección en los resultados de búsqueda [Tesis de doctorado, Universitat Autònoma de Barcelona]. http://hdl.handle.net/10803/688436
Paolillo, J. C. (2018). The Flat Earth phenomenon on YouTube. First Monday, 23(12). https://www.firstmonday.org/ojs/index.php/fm/article/view/8251
Parker, T. (May 10, 2020). YouTube CEO: Users don’t like “authoritative” mainstream media channels but we boost them anyway. Reclaim The Net. https://n9.cl/25bfq
Prado, E. (2022). Retos para los legacy media y el servicio público en el ecosistema audiovisual convergente. En M. Francés, R. Franquet, & G. Torres (Coords.), Retos de la televisión pública ante la multidifusión digital (pp. 57-87). Gedisa.
Ribeiro, M. H., Ottoni, R., West, R., Almeida, V. A. F., & Wagner Meira, W. M. (2020). Auditing radicalization pathways on YouTube. En M. Hildebrandt, & C. Castillo, FAT* '20: Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency (pp. 131-141). https://doi.org/10.1145/3351095.3372879
Rieder, B., Matamoros-Fernández, A., & Coromina, Ò. (2018). From ranking algorithms to ‘ranking cultures’: Investigating the modulation of visibility in YouTube search results. Convergence, 24(1), 50-68. https://doi.org/10.1177/1354856517736982
Rodríguez-Serrano, A., García-Catalán, S., & Martín-Núñez, M. (2019). Estrategias narrativas audiovisuales de desinformación en YouTube de la nueva extrema derecha europea. Profesional de la información, 28(3), https://doi.org/10.3145/epi.2019.may.11
Saner, E. (August 10, 2019). YouTube’s Susan Wojcicki: 'Where's the line of free speech – are you removing voices that should be heard?' The Guardian. https://n9.cl/bi9ojn
Santín, M., & Álvarez-Monzoncillo, J. M. (2020). YouTube en la prensa española: Un modelo por definir. Profesional de la Información , 29(1). https://revista.profesionaldelainformacion.com/index.php/EPI/article/view/73320
Shaheed, N. M. (2019). Conspiracy Theories and Flat Earth Vídeos on YouTube. The Journal of Social Media in Society Fall, 8(2), 84-102.
Sued, G. (2020). El algoritmo de YouTube y la desinformación sobre vacunas durante la pandemia de COVID-19. Chasqui. Revista Latinoamericana de Comunicación, 1(145), 163-180. https://doi.org/10.16921/chasqui.v1i145.4335
Tufekci, Z. (10 de marzo de 2018). YouTube, the great radicalizer. The New York Times.
YouTube. (2024). Panel de información sobre el medio. https://bit.ly/3P94Zhl
YouTube. (n.d.-a). ¿Cómo aborda YouTube la desinformación? https://bit.ly/3TnZePf
YouTube. (n.d.-b). Política sobre el lenguaje vulgar. https://bit.ly /4a4OKts
AUTHORS’ CONTRIBUTIONS, FUNDING AND ACKNOWLEDGEMENTS
Authors’ contributions:
Conceptualization: Padilla, Adrián; Coromina, Òscar, & Prado, Emili. Software: Padilla, Adrián Formal analysis: Padilla, Adrián. Drafting-Preparation of the original draft: Padilla, Adrián. Revision: Padilla, Adrián; Coromina, Òscar, & Prado, Emili. Supervision: Coromina, Òscar, & Prado, Emili. All authors have read and accepted the published version of the manuscript: Padilla, Adrián; Coromina, Òscar, & Prado, Emili.
Funding: This research has no external funding.
AUTHORS:
Adrián Padilla
Autonomous University of Barcelona.
EUNCET Business School.
D. in Audiovisual Communication and Advertising from the Autonomous University of Barcelona. His studies focus on the social and cultural impact of digital platforms and social networks. He specializes in the use of digital Methods and computational methods for the collection and analysis of data from the digital environment. He teaches digital analytics, audience analysis, big data and data visualization at UAB and EUNCET Business School (UPC).
Índice H: 3
Orcid ID: https://orcid.org/0000-0001-7668-1322
Scopus ID: https://www.scopus.com/authid/detail.uri?authorId=57204037262
Google Scholar: https://scholar.google.es/citations?user=zKICdGIAAAAJ&hl=es
ResearchGate: https://www.researchgate.net/profile/Adria-Padilla-2/research
Òscar Coromina
Autonomous University of Barcelona.
Dr. Òscar Coromina, Serra Húnter Professor in Audiovisual Communication and Advertising, combines research and education, guiding future professionals. He has coordinated master's degree programs and has participated in conferences on Digital Marketing in international roles. His research addresses the impact of digital marketing on society and social network dynamics, exploring emerging trends in research methods. His latest work on Digital Marketing as Digital Methods was published in Big Data and society, with other contributions on platforming creative cultures in Convergence, First Monday and Social Media & Society.
Índice H: 10
Orcid ID: https://orcid.org/0000-0002-6306-4154
Scopus ID: https://www.scopus.com/authid/detail.uri?authorId=57195587707
Google Scholar: https://scholar.google.com/citations?user=CIDIvJUAAAAJ&hl=en
Emili Prado
Autonomous University of Barcelona.
Professor Emeritus of Audiovisual Communication and Advertising at the Autonomous University of Barcelona, where he has been Dean and Director of the Grup de Recerca en Imatge, So i Síntesi, and of the television observatories Euromonitor and Usamonitor and of the social networks OXSI. He has also been founding Dean of the Faculty of Communication Sciences at the University of Santiago de Compostela. He has been a visiting professor and researcher at the Universities of Quebec in Montreal, California Berkeley, New York, Bordeaux, Pisa, Sao Paulo and several Spanish universities.
emili.prado@uab.cat
Índice H: 26
Orcid ID: https://orcid.org/0000-0003-4871-2472
Scopus ID: https://www.scopus.com/authid/detail.uri?authorId=56452662700
Google Scholar: https://scholar.google.com/citations?user=JTWsTA4AAAAJ&hl=es
ResearchGate: https://www.researchgate.net/profile/Emili-Prado/research
Academia.edu: https://uab.academia.edu/EmiliPrado