Political Asymmetry in the Spread of Disinformation

Throughout the years, social media platforms have implemented various measures to reduce the spread of disinformation, from reducing the visibility of posts that share disinformation and penalizing repeat offenders to deleting posts and suspending accounts. These policies were challenged at different times for their alleged political bias, as they were considered to penalize certain political sectors more than others. But are these policies biased or do certain sectors disinform more than others?

Current evidence shows that there is indeed political asymmetry in the spread of disinformation on social media. Nonetheless, it’s important to clarify that most of these studies use data from the U.S., so they mainly analyze differences between Republicans and Democrats, or conservatives and liberals.

Conservatives Spread More Disinformation Than Liberals

---

An asymmetry has been consistently observed from 2016 to the present. Extensive research with different methodological approaches, conducted in different electoral contexts and other politically relevant scenarios, show that conservatives tend to interact with and share significantly more disinformation than liberals or moderates.

For example, 3,500 Facebook users participated in a study carried out during the 2016 U.S. presidential election, which analyzed their posts on the platform. Results showed that Republican users shared more disinformation than Democrats: 18.1% vs 3.5%. A similar pattern was observed according to their ideology: conservatives, especially those who identified themselves as “very conservative,” shared the most disinformation. Similar results were found in another study from the same electoral period that analyzed tweets from 16,442 Twitter accounts. It should be pointed out that both studies show an important limitation: they only consider posts containing links to external websites (a limited list) previously classified as disinformation by academics, journalists, and verifiers. Therefore, they do not take into consideration the entire ecosystem of disinformation. 

Other studies used broader criteria to define disinformation, which helped them to better capture the diversity of the phenomenon.

For example, a large-scale study of image-based political disinformation on Facebook collected 13,723,654 posts from 14,532 pages and 11,454 public groups from August through October 2020. This study assessed the political orientation of all posts according to the party affiliation of political figures mentioned or their stance on specific issues: images showing support for causes like LGBTQ, Black Lives Matter, minimum wage or gun control were labeled “left-wing,” while those with opposing views were considered to be “right-wing.” Results showed enormous partisan asymmetry: Right-leaning images were 5 to 8 times more likely to be false or misleading than those associated with left-wing positions.

Another study with vast amount of data, conducted during the 2020 election in collaboration with Meta, showed the same asymmetry. This research was much more robust. It gathered information from 208 million U.S. Facebook users and tracked all URLs from political news posted on the platform between September 2020 and February 2021. The analysis included the content that users could have seen in their feeds and the information with which they engaged (clicks, reactions, reshares and comments). In addition, the criteria for identifying disinformation were applied to the content of the posts themselves, rather than relying solely on whether they originated from a limited list of websites previously classified as low-quality. Although it presents methodological limitations, like all studies, its scale and analytical depth make it particularly relevant.

This pattern was also observed among all political leaders, although not in all countries. A study that analyzed 3.4 million tweets from U. S. American, British and German politicians made between 2016 and 2022 found that conservative politicians in the United States share information of lower quality than liberals, and that, over time, Republican members of the U.S. Congress were increasingly circulating news from dubious sources, thus widening the gap with their Democrat counterparts. On the other hand, there were fewer differences between right-wing and left-wing politicians in Germany and the UK, and these remained constant throughout time.

A more thorough analysis that examined 32 million tweets from parliamentarians in 26 European countries (member states and candidate member states of the European Free Trade Association), spanning six years and several election periods, offers a complementary perspective. This study classified political parties according to their level of populism, using an index that combines two aspects: people-centrism and antielitism (perception of corrupt elites). Investigators identified disinformation content while examining the URLs shared by politicians, using specialized databases to measure each party’s informative quality. Results showed that populism in itself is not linked to disinformation, right-wing parties aren’t more likely to spread disinformation and left-wing populists don’t spread more disinformation than traditional parties. Nonetheless, politicians associated with radical right-wing populist parties do spread significantly more disinformation than their traditional counterparts, suggesting that the link between populism and disinformation is specific to this form of politics.

These studies, as a whole, reveal the consistency of partisan and ideological asymmetries in the spread of disinformation. The assertiveness of this finding comes precisely from the variety of approaches used to document it: diverse data sets, various methodologies (surveys, analysis of posts, assessment of URLs), different time periods and multiple political contexts (presidential elections, pandemics and other relevant events).

All these studies have one thing in common: they determine what disinformation is based on assessments from experts — professional fact-checkers, specialized journalists and academics — who, thanks to their training and experience, are qualified to evaluate the veracity of content and reliability of sources. However, the same sectors that challenge social media content moderation policies — on the grounds of political bias — also argue that the experts assessing content are themselves liberally biased, which raises the possibility that the observed asymmetry reflects evaluator bias rather than actual differences in disinformation dissemination. 

A 2024 study examines, in a particularly thorough way, asymmetries in disinformation dissemination and addresses criticisms concerning potential biases among professional fact-checkers. Researchers analyzed 9,000 politically active Twitter users during the U.S. 2020 presidential election, classifying them based on whether they shared the hashtags #Trump2020 or #VoteBidenHarris2020, and examined the links to news websites they posted in October 2020.

The study used three complementary methods to measure the spread of disinformation. First, they compared the URLs shared by users with multiple website trustworthiness ratings. Professional fact-checkers and journalists analyzed 60 domains; 283 domains were rated by Ad Fontes Media; 3,216 by Media Bias/Fact Check; and 4,767 domains were rated by aggregating ratings from various fact-checkers and academics. In all cases, they found the same pattern: users that shared Trump’s hashtag spread content from websites significantly less trustworthy than those who shared Biden’s hashtag.

However, since these results were based on tests from professional fact-checkers and journalists, researchers implemented a second method designed specifically to minimize any political bias: they used trustworthiness ratings created by politically balanced groups of laypeople. They surveyed 970 demographically representative Americans, who rated the trustworthiness of 60 news websites on a five-point scale and reported their partisan preference (Democrat or Republican, with no neutral option). For each outlet, they then calculated balanced ratings by calculating separately the assessments from Democrats and Republicans, and then averaging those two average ratings, thus giving equal weight to both groups. Results were practically identical: using these politically balanced ratings, the average quality of domains shared by Trump users was 2.17 s.d. lower than people who used Biden hashtags. Even when creating a purposefully right-biased quality measure by only using the trustworthiness ratings of Republican laypeople, the pattern remained: Trump’s users shared content 1.29 s.d. less trustworthy than Biden’s. To contextualize this difference, the median Trump hashtag poster shared four times more links to low-quality websites compared with the median Biden hashtag poster.

The study also verified that this pattern was not unique to the method of political classification, finding equally high correlations when the users’ ideology was predicted based on the accounts they followed or news sites they shared.

Finally, considering that previous analysis used the website’s quality as an indirect indicator of the accuracy of its content, researchers implemented a third method that directly assessed the veracity of specific posts, qualified as inaccurate by professional fact-checkers and politically biased groups of laypeople. Once again, conservative users shared significantly more inaccurate URLs than liberals. Additionally, during a large-scale survey experiment conducted in 16 countries, participants were presented with COVID-19 claims (eliminating potential exposure confounds, and presented without source attribution) and measured according to their sharing intentions. Once again, they found a significant correlation between conservatism and average sharing intentions for inaccurate claims.

A recent 2025 study revisits the objection that the identification of misleading content relies on professional journalists and fact-checkers by analyzing evaluations of tweets made through X’s Community Notes program. In this case, identifying misleading content is not the job of professional fact-checkers, but the result of a consensus among a diverse community of X users. Community Notes allow users to mark posts as potentially misleading and write explanatory notes that other users vote on. An open code algorithm requires an agreement among users with diverse perspectives to validate the note and flag the post as misleading. This method, based on crowd-sourced assessments, is less likely to be accused of bias than fact-checkers. Researchers examined all English-language notes written between January 2023 and June 2024, a total of 218,382 proposed Community Notes, and the results are conclusive. Once again, 2.3 times as many posts by Republicans are flagged as misleading compared to posts by Democrats. This finding cannot be conferred to differences between X users, since — at the time of data gathering — there were no significant Republican overrepresentations. 

The evidence presented is clear: party asymmetry in the spread of disinformation on social media is observed through time and in different contexts, even when the assessment comes from the user community, and cannot be attributed to evaluators’ political biases.

Literature Consulted

---

• Guess, A., Nagler, J., & Tucker, J. (2019). Less than you think: Prevalence and predictors of fake news dissemination on Facebook. Science Advances, 5(1), eaau4586. https://doi.org/10.1126/sciadv.aau4586
• Grinberg, N., Joseph, K., Friedland, L., Swire-Thompson, B., & Lazer, D. (2019). Fake news on Twitter during the 2016 U.S. presidential election. Science, 363(6425), 374–378. https://doi.org/10.1126/science.aau2706
• González-Bailón, S., et al. (2023). Asymmetric ideological segregation in exposure to political news on Facebook. Science, 381(6656), 392–398. https://doi.org/10.1126/science.ade7138
• Lasser, J., et al. (2022). Social media sharing of low-quality news sources by political elites. PNAS Nexus, 1(4), pgac186. https://doi.org/10.1093/pnasnexus/pgac186
• Mosleh, M., Yang, Q., Zaman, T., et al. (2024). Differences in misinformation sharing can lead to politically asymmetric sanctions. Nature, 634, 609–616. https://doi.org/10.1038/s41586-024-07942-8
• Renault, T., Mosleh, M., & Rand, D. G. (2025). Republicans are flagged more often than Democrats for sharing misinformation on X’s Community Notes. Proceedings of the National Academy of Sciences, 122(25), e2502053122. https://doi.org/10.1073/pnas.2502053122
• Yang, Y., Davis, T., & Hindman, M. (2023). Visual misinformation on Facebook. Journal of Communication, 73(4), 316–328. https://doi.org/10.1093/joc/jqad020
• Törnberg, P., & Chueri, J. (2025). When do parties lie? Misinformation and radical-right populism across 26 countries. The International Journal of Press/Politics. Advance online publication. https://doi.org/10.1177/19401612241311886