Simmering tension between Democratic presidential candidates Bernie Sanders and Elizabeth Warren appeared to burst into the open on social media on Wednesday, with the "#NeverWarren" hashtag gaining media attention by becoming the No. 1 trend on Twitter.
But the data paint another picture.
The hashtag was used by many well-followed accounts that sought to downplay division between Sanders and Warren, in turn boosting the trend and further fueling the belief that the candidates and their supporters were at odds.
The top three most engaged-with tweets on the #NeverWarren hashtag called for unity and urged people to resist divisiveness between the two liberal candidates, highlighting how social media remains an imperfect — and easily manipulated — part of the political scene.
Analysis of the #NeverWarren hashtag by NBC News using the social media analytics tool Hoaxy showed that the top tweet Wednesday morning was posted by the author and pastor John Pavlovitz.
"Anyone posting #NeverWarren gets blocked immediately," he tweeted.
Pavlovitz later took down the tweet, citing its impact on the hashtag's spread.
Users imploring others to ignore the negative hashtag — as well as others, such as #WarrenIsASnake — wound up giving momentum to the appearance of a growing rift between Warren and Sanders supporters. It's a prime example of what disinformation researchers call "accidental amplification."
The hashtags trended after Warren said in a statement and, later, on Tuesday night's debate stage, that Sanders once told her that a woman couldn't win the presidency. Sanders denied having made the comment, and his supporters used the hashtags to attack Warren as a liar, a motif that has come up repeatedly as the two candidates compete over their progressive bona fides. Both senators have passionate supporters, but Sanders' fans have developed a reputation for trafficking in particularly aggressive online vitriol.
But social media is particularly geared to make minor divisions seem bigger than they are.Kanishk Karan, a researcher at the Atlantic Council's Digital Forensics Lab, which tracks disinformation campaigns, said it's easy for people to see a hashtag and assume it's trending because it's popular or because they agree.
"Anyone who just read the hashtag and not the tweets themselves wound up conflating a negative sentiment that's less popular with the actual sentiment," Karan said. "Users telling people not to believe a hashtag — it doesn't fix the idea of what the conversation is."
Twitter users had tweeted the #NeverWarren hashtag 62,000 times by 3 p.m. ET Wednesday , and CNN, The Hill, Daily Wire and Newsweek all wrote about it, saying supporters were playing into a "feud" or "meltdown."
While some Twitter users genuinely appeared to promote the divisive messages, the second-most-retweeted post about the #NeverWarren hashtag Wednesday morning came from Mehdi Hasan, a columnist for the investigative journalism website The Intercept, who excoriated users who agreed with it.
"If you're tweeting in support of this ludicrous #NeverWarren hashtag, you're not only dumb but you're also telling the world that you're ok with kids in cages and bans on Muslims," Hasan said.
Divisive hashtags can be started on purpose to bait high-profile users into denouncing them, Karan said. The tactic was frequently deployed during the 2016 election, notably in hashtag campaigns by groups like WikiLeaks.
Karan said fixing the problem comes down to tech companies, who he said need better systems to give better context around hashtags, and users, who should think twice before they tweet — even if they mean well.
"The important thing is: How do you talk about it? You want to add value and depth in this conversation, but if you retweet this keyword, you're adding to it," Karan said.