The analysis, published in Frontiers in Communication, combined AI-assisted filtering with human validation, multilabel thematic coding and statistical testing of user responses. GPT-4o mini was used to identify potentially relevant posts and later classify replies, but researchers manually reviewed every flagged post, independently reassessed ambiguous cases and audited a random 10% of the rejected material. Two researchers coded the thematic content, with a blind 50-post overlap producing a Cohen’s kappa of 0.72 and observed agreement of 0.83. The study is observational and describes the content and reception of posts; it does not establish that Musk’s rhetoric caused changes in public trust, political behaviour or the quality of information.

Musk’s most frequent references were to broad categories such as “media,” “legacy media,” “mainstream media” and “press,” rather than to individual journalists or isolated reporting errors. Generic labels enabled broad claims about bias, political coordination and informational manipulation, while identifiable outlets supplied concrete examples for those claims. X was mentioned more often than Facebook, LinkedIn or YouTube. The identifiable media organisations were also geographically concentrated: 153 references concerned United States outlets and 42 British outlets, while continental European organisations appeared much less frequently. The corpus therefore captures a discourse centred primarily on Anglo-American media conflicts during a period of intense electoral polarisation.

Bias and credibility criticism dominated the thematic analysis, appearing in 59.2% of the 500 relevant posts. Journalism was represented as selective, dishonest, partisan or structurally incapable of neutrality, rather than merely occasionally fallible. Accusations of disinformation and propaganda appeared in 35% of posts, often depicting media organisations as intentionally concealing information, coordinating messages or manipulating elections. Claims about electoral interference appeared in 13.4%. Against this construction, X was associated with participation, direct communication and freedom from institutional filters. Musk’s statement after the election that “You are the media now” exemplified the inversion: instead of journalists distributing information to an audience, platform users were presented as journalists capable of producing and circulating it themselves.

The analysis identified a clear normative asymmetry between journalism and X. References to traditional media were dominated by delegitimising frames involving manipulation, propaganda, bias and deception, whereas references to X were linked to openness, participation, innovation and communicative freedom. Technological transformation and media innovation occurred in 40.2% of X-related discourse but only 7.6% of media-related discourse. Disinformation was framed in the opposite directions: media organisations appeared as producers and distributors of distorted narratives, while X was portrayed as a corrective environment in which users could challenge reporting. Even humour served contrasting purposes, mocking media organisations through satire in one case and celebrating virality and citizen participation in the other.

Citizen journalism nevertheless did not eliminate mediation in the platform model. Musk’s rhetoric contrasted professional and editorial selection with platform-based mechanisms such as user participation, rebuttals and Community Notes. But X remained an algorithmically organised infrastructure that selected, ordered and distributed content, and Musk controlled the platform promoting itself as a substitute for legacy media. The study therefore interprets “cutting out the middleman” as a relocation rather than the removal of mediation. The locus of gatekeeping shifts from professional journalists toward users and platform systems, while the owner of that system supplies the normative account of the shift and retains control over the infrastructure on which it occurs.

User responses broadly converged with Musk’s position without producing a uniformly supportive discussion. Researchers collected 42,066 replies through stratified sampling from 3,233,112 interactions associated with the selected posts. The model classified 25,520 replies, or 60.67%, as agreement and 15,342, or 36.47%, as disagreement; 2.86% lacked a sufficiently identifiable evaluative position. Support was especially concentrated around posts alleging political and electoral manipulation or attacking media bias and credibility. Posts presenting X as a direct alternative and discussing technology, artificial intelligence or communication transformation showed relatively more disagreement, suggesting that rejection of institutional journalism was more mobilizing than enthusiasm for the proposed platform substitute.

The statistical analysis found statistically significant associations between agreement and both thematic orientation and inferred professional background, but their effect sizes were limited. Cramér’s V was 0.0485 for thematic categories and 0.1100 for professional categories. Inferred gender was not significantly associated with agreement at the conventional threshold: the p-value was 0.0620 and the effect size was only 0.0144. Professional inference was possible for 7,043 profiles, while 81.73% of the sample lacked sufficient occupational information. Profiles linked with design and creative work, research, activism and socio-political analysis tended toward greater disagreement, whereas some managerial, family-oriented, retirement-related and property-associated profiles showed relatively more agreement. These patterns cannot establish the identities or occupations of respondents.

Several constraints make the findings narrow but informative. The study covers one highly polarised political transition rather than Musk’s wider record or X users in general. It measures what appeared in a sampled set of posts and replies, not the truth of claims about media bias, propaganda or electoral manipulation. Gender and profession were estimated from self-presentational profile text, and automated classification can misread irony, slang, memes, evolving jargon and context-dependent remarks. Researchers addressed these issues through manual review and reclassification, but some interpretive uncertainty remains. The central result is a description of rhetorical organization: Musk framed journalism as politically compromised and X as a corrective, participatory space, while users generally endorsed the criticism more strongly than the alternative, and the platform’s owner simultaneously promoted and governed the mechanism presented as replacing professional mediation.

Subject of Research: Analysis of Elon Musk’s anti-media and citizen-journalism rhetoric on X during the 2024 US presidential transition

Article Title: Cutting out the middleman? Elon Musk and the rhetoric of citizen journalism in platform-mediated communication

Article References: Peña Alonso, U., Meso Ayerdi, K., Peña Fernández, S., & Morales i Gras, J. (2026). Cutting out the middleman? Elon Musk and the rhetoric of citizen journalism in platform-mediated communication. Frontiers in Communication, 11, Article 1901417. https://doi.org/10.3389/fcomm.2026.1901417

Image Credits: AI Generated

DOI: 10.3389/fcomm.2026.1901417

Keywords: Elon Musk, X platform, citizen journalism, anti-media rhetoric, media legitimacy, platform governance, information authority, social media analysis, AI-assisted research, Cutting, middleman, Elon