A month before Georgia voters are set to decide which party will control the U.S. Senate, online disinformation about the state’s elections is kicking into overdrive. The full force of 2020’s most effective fake news tactics is hitting the state: fears of widespread voter fraud, allegations of violence and claims about candidates’ socialism. But despite its much-touted efforts to add warning or fact-checking labels to election disinformation, Facebook is failing to do exactly that on more than half of the questionable posts related to the Georgia races, according to a new analysis by Avaaz, a nonprofit that tracks online disinformation.
The group found that Facebook failed to add fact-checking labels to at least 60% of a cross-section of Georgia-related election misinformation that reached thousands of voters. An analysis of more than 200 posts promoting a dozen false claims about the Georgia elections in both English and Spanish showed how the posts followed a now well-worn playbook of amplifying disinformation that has proven effective in recent months. This includes falsehoods about widespread voter fraud, fake rumors about acts of violence targeting African-American voters, and allegations that Democratic Senate candidate Raphael Warnock had “celebrated” Cuban leader Fidel Castro. This sample of false claims was widely shared and garnered more than 643,400 interactions, despite the clear debunkings of each by Facebook’s own fact-checking partners.
In November’s presidential election, Biden flipped Georgia blue by a razor-thin margin of 12,670 votes, or 0.25%, for the first time in almost two decades. Recent polls show that the Jan. 5 runoffs, which will decide if Republican Sens. David Perdue and Kelly Loeffler will keep their seats or lose to their Democratic opponents, will also be tight. If Democrats manage to win both seats, it will put the control of the Senate in Democrats’ hands.
In such a competitive and high-stakes contest, a deluge of online disinformation undermining faith in the election, targeting key constituencies, or spreading viral false claims about candidates could make a real difference, analysts say. This fall “we saw disinformation actors targeting a bunch of swing states, and now they’re putting all their resources into one, so all that disinformation can have a heightened, more extreme impact,” Avaaz Campaign Director Fadi Quran tells TIME. “Even though it’s the same kind, even though Facebook’s failures are the same, the consequences are much larger.”
Facebook did not respond to TIME’s request for comment.
If false claims are identified by Facebook’s fact-checking partners, which include Politifact and several news organizations, the company says it uses “technology to identify near-identical versions” and label them. Between March and Nov. 3, Facebook says it added warning labels to more than 180 million posts that included election misinformation.
Critics of Facebook’s election efforts say the company’s rollout of fact-checking labels was mainly a good public relations move, and not the most effective way to debunk misinformation for the platform’s users. In the months before the election, Facebook said it would add labels to election disinformation, including to “newsworthy” posts from politicians spreading false claims, and remove posts that break the company’s rules on voter interference.
But other features developed to combat election disinformation that many experts find more effective, including retroactively notifying users that they had seen false news, directing them to specific fact checks instead of neutral labels, and demoting “hate bait” meant to invite a flood of angry comments and shares, were vetoed by Facebook executives who feared they would disproportionately apply to right-wing websites and provoke backlash from conservative users, according to the New York Times.
The Georgia analysis shows how Facebook applies its fact-checking or warning labels inconsistently to nearly-identical pieces of misinformation. For example, a Nov. 11 segment of Tucker Carlson’s show on Fox News that was shared on Facebook, which pushes false claims about dead people voting in Georgia as evidence of widespread fraud, now has a label warning of “partly false information.” The same segment shared by the Team Trump Facebook page, which has over two million followers, has only been given a neutral label stating that voting by mail and in person both have “a long history of trustworthiness in the U.S.” The same claims, when written out in a Spanish-language post shared more than 300 times, have no label at all.
Other widely amplified false claims pushed by President Donald Trump, Republican politicians and conservative outlets about Georgia – for instance, that the consent decree prevented officials from verifying signatures on absentee ballot envelopes, that election workers had been spotted throwing out ballots, that Democrats had volunteers going door to door to “cure,” or fix, their ballots, and that Republican votes had been found in a dumpster – showed similarly inconsistent labeling.
Top election officials in Georgia have forcefully pushed back on these claims, many of which have been perpetuated by Trump and other Republican leaders amid a recount of votes in the state’s presidential race requested by the president. These attacks, which have been amplified on Facebook and other platforms, have led to death threats and harassment of election officials. A Twitter thread accusing a 20-year old technician working for Dominion Voting Systems of altering ballots led to his identity being made public and threats that he “should be hung for treason.”
“Someone’s going to get hurt, someone’s going to get shot, someone’s going to get killed,” Gabriel Sterling, who works in the office of the Republican secretary of state, said at a Tuesday news conference, appearing visibly angry. “It’s not right.”
The disinformation about Georgia’s runoffs also shows how both politicians and partisan outlets have become adept at getting through loopholes in Facebook’s policies. In one case, Republican candidate Kelly Loeffler shared a Fox News story about how a church where Democratic candidate Raphael Warnock had been a youth pastor received Cuban president Fidel Castro in 1995. She said this was evidence he was “the most dangerous and radical candidate in America.” According to the company’s own fact-checking partner PolitiFact, “there’s no evidence Warnock was involved in arranging the speech or in welcoming Castro.”
While Loeffler is protected under Facebook’s political speech policy, which exempts politicians from third-party fact checks, the post did not even receive a generic election label warning users of unverified claims. Neither did the “vast majority” of the shares of her misleading claim as it was amplified across the platform, according to Avaaz. The official Georgia GOP page further spread the unsubstantiated claim that Warnock had personally been involved, calling him “true radical” who had “celebrated Castro.” Breitbart posted a misleading article quoting Loeffler about Warnock’s “Marxist ideology” as “exposed” by the Castro visit, which was then also widely shared. None of these posts were issued election labels from Facebook.
Another false claim posted on a Facebook page that was shared more than 2,000 times warned that the NAACP “issued a warning in Georgia that white supremacists and other fringe groups will be targeting black men for violent acts this weekend.” It received a real fact-check label, calling it a false claim and directly linking to a Politifact article debunking it, but posts that shared the copy-pasted text did not, according to Avaaz’ analysis.
This is likely just the beginning of what’s to come in Georgia as the state becomes ground zero in the ongoing 2020 election saga, says Avaaz. “Facebook’s failure to act on such content could further harm trust in the election process and influence both voter turnout and election behavior…[and] have the potential to suppress turnout from key constituencies in the special election,” the report says.