Social Media's Hidden Power Structure Revealed

Social media platforms promise open democratic spaces where anyone can participate in political discussions, but new research reveals a starkly different reality. A study examining Arabic-language conversations about Hezbollah on X (formerly Twitter) shows that while thousands of users participate in posting content, audience attention is overwhelmingly concentrated among a tiny minority of accounts. This finding s the notion that social media creates level playing fields for political discourse and reveals how influence becomes concentrated in ways that mirror traditional media hierarchies.

The key finding from this research is dramatic in its inequality. The study analyzed 15,767 tweets posted by 8,148 users between March 1 and March 8, 2026, and discovered that the top 1% of users—just 81 accounts—capture 61.5% of all engagement. This concentration becomes even more extreme when looking at broader groups: the top 5% of users capture 90.6% of total engagement, while the top 10% capture 96.2%. These numbers reveal that despite thousands of participants in the conversation, nearly all attention flows to a small fraction of users, creating what researchers call a 'pronounced concentration of attention' within what appears to be an open discussion.

Ology behind these involved collecting data from X using automated queries through the Apify cloud platform. Researchers retrieved Arabic-language tweets containing references to Hezbollah, applying filters to exclude retweets and restrict to Arabic-language content. After collecting data through multiple overlapping runs to improve coverage, the final dataset contained 15,767 tweets from 8,148 unique users. Researchers then classified users into two categories: media-labeled accounts (identified through media-related keywords in account metadata) and non-media users. This classification allowed comparison between institutional media accounts and ordinary participants, with engagement measured as the sum of likes, reposts, and replies for each tweet.

Show several important patterns beyond the overall concentration of attention. While non-media users account for 89.6% of participants and produce 79.9% of all tweets, they receive lower engagement per tweet (30.84 interactions) compared to media accounts (41.32 interactions). This means media posts generate roughly 34% more engagement per tweet despite representing only 10.4% of users. The composition of the most engaged users reveals media accounts are substantially overrepresented—while they make up just 10.4% of all users, they account for 29.6% of the top 1% most engaged accounts. Manual inspection of these highly engaged accounts shows they include institutional media outlets, journalists, political commentators, activists, official spokesperson accounts, and several large-following parody accounts, most with follower counts exceeding 10,000 and more than half exceeding 100,000 followers.

Of these are significant for understanding how political discourse actually functions on social media platforms. The research reveals a clear disparity between participation and attention—while conversations may involve thousands of users, visibility and influence remain concentrated among a small group. This pattern suggests that social media platforms, despite their appearance as open forums, may actually reinforce existing power structures rather than democratizing political discussion. The overrepresentation of media accounts among highly engaged users indicates that traditional media institutions maintain significant influence even in supposedly decentralized online spaces, while the concentration of attention among accounts with large follower bases suggests that existing audience reach plays a crucial role in shaping what gets seen and discussed.

Several limitations of the study should be considered when interpreting these . The classification of media accounts relied on keyword-based approaches applied to account metadata, which means some individual journalists with media-related descriptors were classified as media while others were not. The researchers acknowledge this classification should be understood as capturing accounts presenting themselves as media-related actors rather than representing a precise separation between institutional outlets and individual users. Additionally, the study examined only one week of data (March 1-8, 2026) and focused specifically on Arabic-language discourse about Hezbollah, meaning patterns might differ for other topics, languages, or time periods. The exclusion of quote tweets from the engagement metric also means certain types of interaction were not captured in the analysis, though researchers made this choice to maintain consistency in measuring interactions.