Inside Twitter’s Decision to Fact-Check Trump’s Tweets
‘We knew from a comms perspective that all hell would break loose,’ says Twitter’s vice president of global communications
At 8:17 a.m. on Tuesday, Donald Trump sent a characteristically aggrieved tweet claiming that mail-in ballots were “fraudulent,” and that ballots would be stolen and forged, leading to a “rigged election.” At first, not much happened: These sorts of tweets from Trump are an everyday occurrence, and Twitter had never taken action on one before.
Within 24 hours, however, Twitter had fact-checked the U.S. president for the first time, adding a label to his tweets encouraging viewers to “get the facts” about California’s mail-in ballot plans. The label triggered fury from the White House, and the vilification of a previously obscure Twitter employee, Yoel Roth, on Fox News, which led to death threats. Trump continued to target Roth on Thursday.
Within 48 hours of the fact-check, Trump was preparing to issue an executive order that would ban federal spending on social platforms that exercise certain editorial powers and encourage the Federal Trade Commission and state attorneys general to launch investigations against them. The order could spark a reassessment of Section 230 of the Communications Decency Act, the law that underpins much of the modern Internet.
Brandon Borrman, Twitter’s vice president of global communications, talked to OneZero about the company’s decision. That conversation, coupled with information from other Twitter representatives and public statements from the company’s executives, shed light on just how the company arrived at that fateful fact-check, and the roles played by Roth and others, including CEO Jack Dorsey. It adds up to a picture of a company that knew full well what it was doing when it fact-checked the president — and what kind of reaction it would spark from the White House.
“The company needed to do what’s right, and we knew from a comms perspective that all hell would break loose,” Borrman told OneZero.
Trump’s tweet about mail-in ballots was first flagged on Tuesday morning, not by anyone at Twitter, but by one of the third-party nonprofits that partners with Twitter via its elections integrity hub. Twitter invites some of those partners to flag tweets that they worry might violate its “civic integrity” policy. (Borrman declined to disclose which nonprofit it was, in this case.)
Put in place before the 2018 U.S. midterm elections, and recently expanded, Twitter’s civic integrity policy applies special fact-checking scrutiny to tweets that might interfere with people’s participation in democratic processes. An exception to the platform’s generally laissez-faire approach to fact-checking, it’s a level of scrutiny that Twitter applies in only one other arena, which is potentially harmful misinformation about Covid-19. (Twitter has historically avoided interfering with Trump’s tweets — the company did nothing earlier this week when the president pushed the conspiracy theory that MSNBC host Joe Scarborough may have committed murder.)
Once flagged, Trump’s tweet about mail-in ballots went to an internal team at Twitter that focuses on election integrity, which is part of its broader Trust and Safety division. While that team doesn’t have a formal head, one of its members is Yoel Roth, whose title is head of site integrity. Roth started at Twitter as an intern in 2014, after finishing his PhD in communications. He is not among the company’s top executives. His main job is to work on platform manipulation, which includes spam, bots, and foreign election interference, not fact-checking. But according to multiple Twitter representatives, when a tweet is flagged for potentially violating Twitter’s civic participation policy, he’s part of the group that initially reviews it.
“The company needed to do what’s right, and we knew from a comms perspective that all hell would break loose.”
That initial review found that Trump’s tweet did not violate Twitter’s policy against tweets that attempt to manipulate or interfere in an election. If it had, the company’s policy would have dictated that it delete the tweet altogether, an action that would likely have provoked even more of a firestorm.
But opting against deleting a tweet didn’t mean that the company would have to let it go unchecked. On May 11, Twitter announced that it would begin applying labels and warning messages to disputed or misleading tweets about Covid-19, with links to further context and factual information. That announcement did not mention tweets about elections, but it did leave the possibility open, stating, “We will continue to introduce new labels to provide context around different types of unverified claims and rumors as needed.”
Twitter’s internal team reviewed Trump’s tweet again, this time through the lens of whether to apply its new labeling mechanism to an election-related tweet for the first time. By Tuesday afternoon, it had decided to recommend doing so — a recommendation that would have to go up the chain to the very top of the company for approval. Borrman said he could not provide insight into the deliberations that led to that call, and neither he nor anyone else on the communications team was involved in them.
The next step was to take the recommendation to Del Harvey, Twitter’s vice president of trust and safety, and Sean Edgett, the company’s general counsel. Normally, both of them report to Vijaya Gadde, Twitter’s global lead for legal, policy, and trust and safety and one of Dorsey’s most trusted deputies. But Gadde was on maternity leave, so it fell to Harvey and Edgett to decide whether to send it back for further review, or accept it and bring it straight to Dorsey and the other C-level executives for approval. Harvey and Edgett agreed with the team’s assessment that a warning label and link to a fact-check was the right approach to Trump’s tweet.
By mid-afternoon Tuesday, the recommendation was in front of Dorsey and his executive team. They gave the go-ahead, and only after that did they loop in Borrman and Monique Meche, Twitter’s VP of global public policy and its top liaison to government, Borrman told me. The system is set up that way to keep enforcement decisions independent from the teams responsible for PR and government relations, he noted. (In contrast, Facebook routes critical policy decisions through policy chief Joel Kaplan, who is also the company’s main man in Washington, an arrangement that its former chief security officer recently criticized.) Borrman and Meche did not push back.
At about 4:30 p.m. ET, Twitter added a warning message to Trump’s tweet. Prefaced with an exclamation point icon, it read, “Get the facts about mail-in ballots.” That language echoed the labels it had been placing on misleading coronavirus tweets, which said, “Get the facts about COVID-19.” The message linked to a Twitter Moment — a collection of tweets curated by human editors at Twitter — that carried the boldface headline, “Trump makes unsubstantiated claim that mail-in ballots will lead to voter fraud.”
Critics were quick to point out that the messaging on the tweet itself was ambiguous as to whether it was endorsing or disputing Trump’s claims, while the Moment seemed an awkward format for correcting the record. Wired called it “a bit of a mess.” But there was no mistaking its significance: After three and a half years of taking a hands-off approach to the president’s incendiary Twitter account — which it belatedly justified on the grounds that they’re “newsworthy” and even more belatedly buttressed with a policy creating exceptions for elected officials — the company had intervened for the first time.
Borrman said he expected that Trump would respond, and that lawmakers might call for changes to Section 230 of the Communications Decency Act, the crucial policy that gives internet platforms broad leeway to host and moderate users’ content without being held liable for it. All of those things quickly transpired. What the company did not anticipate, Borrman acknowledged, was the right seizing on an individual Twitter employee, who previously had little public profile, as the person responsible for the decision.
But that’s what soon happened, as a New York Post reporter unearthed a series of extremely ill-considered tweets by Roth from several years ago, which included stereotyping of “flyover states” as full of racist Trump voters and called members of the Trump administration “Nazis.” By Wednesday morning, Roth had been called out on Fox News by Trump aide Kellyanne Conway, and on Thursday his picture was splashed on the cover of the New York Post. On Thursday afternoon, Trump himself had mentioned Roth in an angry tweet, linking to his handle so that his supporters would know exactly whom to go after.
Meanwhile, Trump vowed to take action against social media companies, with news quickly breaking that he was preparing to sign an executive order that would “make it easier for federal regulators to hold companies such as Twitter Inc. and Facebook Inc. liable for unfairly curbing users’ speech,” according to a draft viewed by the Wall Street Journal.
On Wednesday evening, Dorsey said in a series of tweets that he took responsibility for the decision to add a label to Trump’s tweets and stood by it, adding, “Please leave our employees out of this.” For its part, Facebook declined to take action on an identically worded Trump post about mail-in ballots. CEO Mark Zuckerberg appeared on Fox News on Wednesday to implicitly criticize Twitter’s fact-check and insist that his own platform does not want to be an “arbiter of truth.”
Twitter never wanted to be an arbiter of truth either, but thanks to a decision made in a matter of hours on Tuesday, the company is now the face of Trump’s war on social media.