Hyperpartisan political Facebook pages and websites are consistently feeding their millions of followers false or misleading information, according to an analysis by BuzzFeed News. The review of more than 1,000 posts from six large hyperpartisan Facebook pages selected from the right and from the left also found that the least accurate pages generated some of the highest numbers of shares, reactions, and comments on Facebook — far more than the three large mainstream political news pages analyzed for comparison.
Our analysis of three hyperpartisan right-wing Facebook pages found that 38% of all posts were either a mixture of true and false or mostly false, compared to 19% of posts from three hyperpartisan left-wing pages that were either a mixture of true and false or mostly false. The right-wing pages are among the forces — perhaps as potent as the cable news shows that have gotten far more attention — that helped fuel the rise of Donald Trump.
These pages, with names such as Eagle Rising on the right and Occupy Democrats on the left, represent a new and powerful force in American politics and society. Many have quickly grown to be as large as — and often much larger than — mainstream political news pages. A recent feature in the New York Times Magazine reported on the growth and influence of these pages, saying they "have begun to create and refine a new approach to political news: cherry-picking and reconstituting the most effective tactics and tropes from activism, advocacy and journalism into a potent new mixture."
The rapid growth of these pages combines with BuzzFeed News' findings to suggest a troubling conclusion: The best way to attract and grow an audience for political content on the world's biggest social network is to eschew factual reporting and instead play to partisan biases using false or misleading information that simply tells people what they want to hear. This approach has precursors in partisan print and television media, but has gained a new scale of distribution on Facebook. And while it isn't a solely American phenomenon — the British Labour party found powerful support from a similar voice — these pages are central to understanding a profoundly polarized moment in American life.
For example, in late September, Freedom Daily, a Facebook page with more than 1 million fans, scored a viral hit with a post that filled its audience with racial outrage.
The post linked to an article on the Freedom Daily website with the headline "Two White Men Doused With Gasoline, Set On FIRE By Blacks – Media CENSORED (VIDEO)." The text that accompanied the link on Facebook connected the attack to recent Black Lives Matter protests and urged people to share the post "if you're angry as hell & aren't going to take it anymore!"
Anyone clicking on the link saw a video of the altercation, with some additional commentary. "Back in the day, when people were a lot smarter and America was great, this would have been a lot different," the article said.
But nowhere in the article or Facebook post did Freedom Daily make it clear that this incident happened almost a year ago, and that it had absolutely no connection to Black Lives Matter.
The falsehoods continued from there: The altercation was actually between two people, a black man and his co-worker — and perhaps most importantly, the co-worker is not white. Court documents allege that the fight began with the co-worker throwing the first punch. Prosecutors also said the second man caught fire as a result of him coming into contact with the first man who was engulfed in flames. And finally, in spite of the headline's claim that the incident was "CENSORED" by the media, it was widely covered by Baltimore media as well as by CNN and the Daily Mail's website. (The man who allegedly set the fire, Christopher Harrison Jr., was charged with attempted first-degree murder, reckless endangerment, and first- and second-degree assault.)
But these details only stood in the way of success on Facebook. In the end, Freedom Daily's largely false post was shared more than 14,000 times, generating more than 9,000 reactions and over 2,000 angry comments on Facebook.
"Not even animals would do this," reads the most liked comment on the post. "Time to hang these people."
Pages like Freedom Daily play to the biases of their audiences — and to those of Facebook's News Feed algorithm — by sharing videos, photos, and links that demonize opposing points of view. They write explosive headlines and passages that urge people to click and share in order to show their support, or to express outrage. And in this tense and polarizing presidential election season, they continue to grow and gain influence.
"They are, perhaps, the purest expression of Facebook’s design and of the incentives coded into its algorithm," wrote John Herrman in the New York Times Magazine.
These pages are also a constant source of dubious, misleading, or completely false information.
During the period analyzed, right-wing pages, for example, pushed a conspiracy theory about a Hillary Clinton body double, recirculated an old and false story about a Canadian mayor lecturing Muslim immigrants about integration, wrongly claimed that Obama's last address at the UN saw him tell Americans they needed to give up their freedom for a "New World Government," and falsely claimed that a football player had been told not to pray by the NFL.
Left-wing pages wrongly claimed Putin's online troll factory was responsible for rigging online polls to show Trump won the first debate, falsely said that Trump wants to expel all Muslims from the US and said US women in the military should expect to be raped, claimed that TV networks would "not be fact-checking Donald Trump in any way" at the first debate, and completely misrepresented a quote from the pope to claim that he "flat out called Fox News type journalism 'terrorism.'"
The bottom line is that people who regularly consume information from these pages — especially those on the right — are being fed false or misleading information.
The nature of the falsehoods is important to note. They often take the form of claims and accusations against people, companies, police, movements such as Black Lives Matter, Muslims, or "liberals" or "conservatives" as a whole. They drive division and polarization. And in doing so, they generate massive Facebook engagement that brings more and more people to these pages and their websites and into the echo chamber of hyperpartisan media and beliefs.
What We Did
BuzzFeed News selected three large hyperpartisan Facebook pages each from the right and from the left, as well as three large mainstream political news pages. All nine pages have earned the coveted verified blue checkmark from Facebook, which gives them an additional layer of credibility on the platform.
Over the course of seven weekdays (Sept. 19 to 23 and Sept. 26 and 27), we logged and fact-checked every single post published by these pages. Posts could be rated "mostly true," "mixture of true and false," or "mostly false." If we encountered a post that was satirical or opinion-driven, or that otherwise lacked a factual claim, we rated it "no factual content." (We chose to rate things as "mostly" true or false in order to allow for smaller errors or accurate facts within otherwise true or false claims or stories.)
We also gathered additional data: Facebook engagement numbers (shares, comments, and reactions) for each post were added from the Facebook API, and we noted whether the post was a link, photo, video, or text. Raters were asked to provide notes and sources to explain their rulings of "mixture of true and false" or "mostly false." They could also indicate whether they were unsure of a given rating, which would trigger a second review of the same post in order to ensure consistency. Any discrepancies between the two ratings were resolved by a third person. That same person conducted a final review of all posts that were rated mostly false to ensure they warranted that rating. (For more detail on the methodology and some notes on its limitations, see the bottom of this article, and you can view our data here.)
In the end, our team rated and gathered data on 2,282 posts. There were 1,145 posts from mainstream pages, 666 from hyperpartisan right-wing pages, and 471 from hyperpartisan left-wing pages. The difference in the number of posts for each group is a result of them publishing with different frequencies.
Accuracy: Right vs. Left
All nine pages consisted largely of content that was either mostly true or earned a "no factual content" rating.
However, during the time period analyzed, we found that right-wing pages were more prone to sharing false or misleading information than left-wing pages. Mainstream pages did not share any completely false information, but did publish a small number of posts that included unverified claims. (More on that below.)
We rated 82 out of a total 666 right-wing Facebook posts as mostly false, for a percentage of 12.3%. Another 169 posts (25.4%) were rated as a mixture of true and false. Viewed separately or together (38%), this is an alarmingly high percentage.
Left-wing pages did not earn as many "mostly false" or "mixture of true and false" ratings, but they did share false and misleading content. We identified 22 mostly false posts out of a total of 471 from these pages, which means that just under 5% of left-wing posts were untrue. We rated close to 14% of these posts (68) a mixture of true and false. Taken together, nearly a fifth of all left-wing posts we analyzed were either partially or mostly false.
One of the most common reasons we rated a post as a mixture of true and false was because the headline and/or Facebook share line introduced misinformation or was misleading to the audience. This frequently took the form of a shared link that contained accurate body text paired with a misleading headline, likely to drive social engagement and clicks.
For example, the left-wing page Addicting Info shared an article with the headline “Trump Loses Support Of Police Union After Saying Tulsa Shooting Cop ‘Choked’ (VIDEO).” But contrary to the claim in the headline, the article makes it clear that Trump didn't lose an endorsement. The executive director of the Fraternal Order of Police merely gave a quote that was slightly critical of something Trump said.
On the right, Freedom Daily posted a link to an article from the website Yes I'm Right. It carried the headline "Australia Voted To Ban Muslims And Liberals Are Pissed." The story correctly reports on the results of a poll that asked Australians if they would support or oppose a ban on Muslim immigration to Australia. But there was no vote to ban Muslims, making the headline completely false. (Side note: As illustrated by that headline, pages on the right and the left both love to talk about how something that happened made the other side lose their minds, freak out, get totally shut down, etc.)
Alarmingly, we found examples of pages on the left and on the right presenting fake news articles as real. Two left-wing pages, Occupy Democrats and The Other 98%, posted a link to an article on U.S. Uncut that claimed the surgeon general of the US warned that drinking every time Trump lied during the first presidential debate could result in "acute alcohol poisoning." That story was an aggregation of a satirical Raw Story article with the same information, published earlier that day. ("Please do your fact-checking as responsibly as possible," joked the U.S. Uncut article that unwittingly presented false information as true.)
Right Wing News, a page with 3.3 million followers, shared a link to a story that claimed authorities in Charlotte had warned would-be rioters that their food stamps and other government benefits would be revoked if they were caught looting or rioting. That story came from the Baltimore Gazette, a fake news site.
Update: After reading this story, John Hawkins, the owner of Right Wing News, got in touch to dispute some of the mostly false ratings given to his page. As a result of his feedback, we altered four ratings, as detailed in a correction added to this story. Hawkins also published a blog post that outlines his objections and argues that the majority of our mostly false ratings for his page are incorrect. (We disagree.)
"In the small number of instances where I concluded that Buzzfeed was right, I made sure to address those issues with our entire staff in an effort to insure that they didn’t happen again," Hawkins wrote. "All news outlets have a responsibility to be accurate and when mistakes are made, it’s important to deal with them seriously ... If Buzzfeed wants to fact check other people, it should start being careful about getting its own facts right."
Accuracy: Mainstream Pages
Mainstream pages did not publish any mostly false content on the days we checked. We did, however, encounter one story that spread to all three mainstream pages as well as some partisan pages and remains unconfirmed to this day.
There were eight mainstream posts out of a total of 1,145 that earned the "mixture of true and false" rating. The majority of these were related to one story — the report from Politico that former President George H.W. Bush would be voting for Hillary Clinton.
Our ratings guide dictated that any posts built solely on anonymous sources or on unverified claims should be given the "mixture" rating. Since President Bush and his spokesman refused to confirm or deny the report, we rated all stories that repeated this claim the same way. Politico's story about the former president was shared more than 14,000 times from its Facebook page, making it that page's biggest hit during the period we analyzed. Overall, we saw a high number of Facebook shares for stories about the Bush voting claim. But the sample number is too small to make any larger conclusion about how unverified stories perform compared to true stories on mainstream pages.
Worst Offenders = High Engagement
Which pages shared the least credible information?
Freedom Daily, with its 1.3 million fans, was the most inaccurate and misleading page during the period we analyzed. It had the highest percentage of false posts of any page, at 23%, and also saw the same percentage of "mixture of true and false" posts. That means 46% — nearly half — of its total output during the seven days we studied was rated as false or misleading.
Not coincidentally, Freedom Daily put up impressive Facebook engagement stats. It had by far the highest Facebook engagement (defined as the total number of reactions, likes, and shares) per post among the right-wing pages we studied. It ranked third among all nine pages for its median number of Facebook shares per post. (We considered shares to be the most important individual engagement metric, as Facebook itself has said it plays an important role in determining the spread of a post.)
Occupy Democrats was the largest page we analyzed, with 4 million fans, and was rated as the least accurate left-wing page. It had 9 mostly false posts out of a total of 209, accounting for 4% of its output. Just under 16% of its posts (33) were a mixture of true and false. In the end, a fifth of its posts were false or misleading, according to our analysis.
Occupy Democrats' large number of fans means it theoretically has greater potential for engagement than the other pages. In the end, it did receive much higher Facebook engagement than any other page.
While Freedom Daily received a median of 947 shares per post, Occupy Democrats saw a median of 10,931. One factor in its impressive engagement numbers is the fact that Occupy Democrats consistently publishes native video to its page, which is an essential element in driving significant shares, likes, and reactions on Facebook.
The More Partisan or Misleading, the More Engagement
While the majority of the posts we rated from the partisan pages were mostly true, the mostly true posts typically did not perform as well as ones that were mostly false, were a mixture of true and false, or had no factual content. The more overtly partisan, misleading, or opinion-driven a post was, the more engagement the post would see, according to our data. Facebook, and the people using it, appears to reward the worst tendencies of these pages.
For example, Occupy Democrats saw a median of 7,755 shares for its mostly true posts, whereas all other post types received a median of 13,330 shares. Right Wing News — the largest right-wing page, at 3.3 million fans — received a median of 91 shares on its mostly true posts, and its other posts had a median of 568 shares.
The mainstream political news pages we analyzed received a fraction of the engagement of the partisan pages. CNN Politics was the largest mainstream page we analyzed, with more than 1.8 million fans. It had a median of 50 shares per post during the period we analyzed, the highest number for any mainstream page.
The lack of partisan content, along with an overall factual approach, may play a role in the comparably lower engagement for mainstream pages. However, there are additional important factors to consider: Mainstream pages publish with greater frequency, which means each post has less time to get traction before the next one is published; they also overwhelmingly post links, rather than mixing in a significant volume of video and photo posts, which tend to perform better on Facebook. Video and photo posts made up a significant portion of the most popular posts we analyzed.
Memes and Joke Videos
We were surprised by the number of posts that met our "no factual content" criteria. Almost 19% of all posts analyzed from partisan pages fell into this category. These posts were often images or memes that expressed strongly partisan opinions. Here, for example, are memes that generated big engagement for Occupy Democrats and Right Wing News:
These memes and jokes often contained some of the most partisan opinions, and often consisted of attacks against liberals or conservatives. Many of them also took the form of attacks on Hillary Clinton, Donald Trump, or Barack Obama. It's perhaps not surprising, then, that we found posts rated "no factual content" received a very high median number of shares when compared with other post types on the partisan pages.
Humor videos also fell into this category and were the source of some of the biggest hits, particularly for left-wing pages. This Occupy Democrats scrape and repost of a Daily Show segment with Trump supporters received more than 1 million shares:
Echo Chambers and Polarization
One thing we noticed when trying to fact-check posts was that these pages, and the websites connected to them, largely aggregate information from elsewhere. That wasn't surprising. What was notable was that the right-wing pages almost never used mainstream news sources, instead pointing to other highly partisan sources of information. Even if the key information they were covering originally came from a mainstream source, they almost always linked to other partisan sources, which in turn often did the same.
For example, Right Wing News published and shared a story with the headline "NFL Boycott In FULL EFFECT! 44% Of Americans Will Stop Watching Football If Players Continue To Kneel… [VIDEO]." Its article linked to a story from Young Conservatives, which in turn linked to a Breitbart story, which itself was an aggregation of a poll conducted by Yahoo.
This was a contrast with the left-wing pages, which frequently pointed to articles from mainstream sources. Some may view this as validation of the long-held view of conservatives that mainstream media has a liberal bias. It's also important to note that the right-wing pages we analyzed had a much higher percentage of false and misleading information compared to the mainstream and left-wing pages.
Based on our analysis, we found the hyperpartisan right-wing echo chamber to be more polarized than its counterpart on the left, and our sense is that this likely contributes to the tendency for right-wing Facebook pages to promote false and misleading information.
Devoted readers of these pages likely experience a version of the same echo chamber effect. The more they read and engage with these pages, the more Facebook will show them this content in their News Feeds. The more they click on the hyperpartisan websites, the more Google will show them search results from these sources. The result is that over time people will likely become more polarized because algorithms and friends continue to feed them information that pushes them further in this direction. This "group polarization" phenomenon is well-documented and has been shown to exist in studies of Facebook users.
The group of BuzzFeed News reporters who conducted this analysis found that in just a few days, our News Feeds and search results began to shift and align with the type of content we were checking. "The most interesting thing is that after a few days of fact-checking right-wing pages, my Google results started skewing to right-wing sites," said one team member.
The reality is that people who frequent these hyperpartisan pages on the right and on the left exist in completely different segments of the online world, rarely interacting with or seeing what the other side is seeing. The more they rely on these pages for information, the more polarized they will likely become — and the more their worldviews will be based on information that is misleading or completely false.
CORRECTION
The fake news story about the surgeon general of the US warning that drinking every time Trump lied during the first presidential debate could result in "acute alcohol poisoning" originated on Raw Story. We incorrectly said it originated on National Report, but their hoax was published after the Raw Story piece.
Right Wing News received a median of 91 shares on its mostly true posts, and its other posts had a median of 568 shares. We originally said it had a median of 87 shares on its mostly true posts, and its other posts had a median of 521 shares. The data for Right Wing News was recalculated after we realized we incorrectly rated three posts from that page. One post was rated mostly false that was in fact a mixture of true and false; one was rated mostly false that actually had no factual content; and one was rated mostly false that was in fact mostly true. We subsequently corrected a fourth post from mostly false to a mixture of true and false. The graphics in this story were also updated to reflect the corrected Right Wing News totals. To read a response to this story from Right Wing News, click here.
More on Our Methodology and Data Limitations
Each of our raters was given a rotating selection of pages from each category on different days. In some cases, we found that pages would repost the same link or video within 24 hours, which caused Facebook to assign it the same URL. When this occurred, we did not log or rate the repeat post and instead kept the original date and rating. Each rater was given the same guide for how to review posts:
Mostly True: The post and any related link or image are based on factual information and portray it accurately. This lets them interpret the event/info in their own way, so long as they do not misrepresent events, numbers, quotes, reactions, etc., or make information up. This rating does not allow for unsupported speculation or claims.
Mixture of True and False: Some elements of the information are factually accurate, but some elements or claims are not. This rating should be used when speculation or unfounded claims are mixed with real events, numbers, quotes, etc., or when the headline of the link being shared makes a false claim but the text of the story is largely accurate. It should also only be used when the unsupported or false information is roughly equal to the accurate information in the post or link. Finally, use this rating for news articles that are based on unconfirmed information.
Mostly False: Most or all of the information in the post or in the link being shared is inaccurate. This should also be used when the central claim being made is false.
No Factual Content: This rating is used for posts that are pure opinion, comics, satire, or any other posts that do not make a factual claim. This is also the category to use for posts that are of the “Like this if you think...” variety.
In gathering the Facebook engagement data, the API did not return results for some posts. It did not return reaction count data for two posts, and two posts also did not return comment count data. There were 70 posts for which the API did not return share count data. We also used CrowdTangle's API to check that we had entered all posts from all nine pages on the assigned days. In some cases, the API returned URLs that were no longer active. We were unable to rate these posts and are unsure if they were subsequently removed by the pages or if the URLs were returned in error.