OAKLAND, Calif. — For months, Twitter, Facebook and YouTube prepared to clamp down on misinformation on Election Day.
On Tuesday, most of their plans went off without a hitch. The social platforms added labels to misleading posts by President Trump and notified their users that there was no immediate outcome to the presidential race. On television, news anchors even cited fact-checks similar to those made by Twitter and Facebook.
Then came Wednesday. With ballots still being counted and the absence of a clear result, the flow of misinformation shifted away from seeding doubts about the vote to false claims of victory. Twitter rapidly labeled several tweets by Mr. Trump in the morning as being misleading about the result of his race, before later doing the same to tweets from Eric Trump and the White House press secretary, Kayleigh McEnany. And Facebook and YouTube used their home pages to show people accurate information about the election.
The actions reinforced how even a smooth performance on Election Day did not mean that the social media companies could relax, fighting a relentless flow of toxic content. In fact, the biggest tests for Facebook, Twitter and YouTube are still looming, misinformation researchers said, as false narratives may surge until a final result in the presidential race is certified.
“What we actually saw on Election Day from the companies is that they were extremely responsive and faster than they’ve ever been,” said Graham Brookie, the director of the Atlantic Council’s Digital Forensic Research Lab. But now, he said, misinformation was solely focused on the results and undermining them.
“You have a hyperfocused audience and a moment in time where there is a huge amount of uncertainty, and bad actors can use that opportunistically,” he said.
Twitter said it was continuing to monitor for misinformation. Facebook said, “Our work isn’t done — we’ll stay vigilant and promote reliable information on Facebook as votes continue to be counted.” YouTube said it also was on alert for “election-related content” in the coming days.
The companies had all braced for a chaotic Election Day, working to avoid a repeat of 2016 when their platforms were misused by Russians to spread divisive disinformation. In recent months, the companies had rolled out numerous anti-misinformation measures, including suspending or banning political ads, slowing down the flow of information and highlighting accurate information and context.
On Tuesday, as Americans voted across the country, falsehoods about broken voting machines and biased poll workers popped up repeatedly. But the companies weren’t tested until Mr. Trump — with early results showing how tight the race was — posted on Twitter and Facebook just before 1 a.m. Eastern to baselessly lash out at the electoral process.
“They are trying to STEAL the Election,” Mr. Trump posted on the sites, without being specific about who he meant.
Twitter moved quickly, hiding Mr. Trump’s inaccurate tweet behind a label that cautioned people that the claim was “disputed” and “might be misleading about an election or other civic process.” Twitter, which had started labeling Mr. Trump’s tweets for the first time in May, also restricted users’ ability to like and share the post.
On Wednesday morning, Twitter added more labels to posts from Mr. Trump. In one, he tweeted that his early leads in Democratic states “started to magically disappear.” In another message, Mr. Trump said unnamed people were working to make his lead in the battleground state of Pennsylvania “disappear.”
Twitter also applied other labels to posts that falsely asserted victory. One was added to a post by Ben Wikler, head of the Democratic Party of Wisconsin, in which he asserted prematurely that Joseph R. Biden Jr. had won the state. The Associated Press and other news outlets later called Wisconsin for Mr. Biden.
On Wednesday afternoon, Twitter also affixed context to tweets from Eric Trump, one of Mr. Trump’s sons, and Ms. McEnany when they preemptively claimed that Mr. Trump had won in Pennsylvania, even though the race there had not been called.
“As votes are still being counted across the country, our teams continue to take enforcement action on tweets that prematurely declare victory or contain misleading information about the election broadly,” Twitter said.
Facebook took a more cautious approach. Mark Zuckerberg, its chief executive, has said he has no desire to fact-check the president or other political figures because he believes in free speech. Yet to prevent the social network from being misused in the election, Facebook said it would couch premature claims of victory with a notification that the election had yet to be called for a candidate, if necessary.
On Tuesday night, Facebook had to do just that. Shortly after Mr. Trump posted about the election being stolen from him, Facebook officials added labels to his posts. The labels noted that “no winner of the presidential election had been projected.”
After the polls closed, Facebook also sent users a notification that if they were waiting to vote at a polling place, they could still vote if they were already standing in line.
On Wednesday, Facebook added more labels to new posts from Mr. Trump, checking his claims by noting that “as expected, election results will take longer this year.”
Unlike Twitter, Facebook did not restrict users from sharing or commenting on Mr. Trump’s posts. But it was the first time Facebook had used such labels, part of the company’s plan to add context to posts about the election. A spokesman said the company “planned and prepared for these scenarios and built the essential systems and tools.”
YouTube, which is not used regularly by Mr. Trump, faced fewer high-profile problems than Twitter and Facebook. All YouTube videos about election results included a label that said the election may not be over and linked to a Google page with results from The Associated Press.
But the site did encounter a problem early on Tuesday night when several YouTube channels, one with more than a million subscribers, said they were livestreaming election results. What the live streams actually showed was a graphic of a projection of an election outcome with Mr. Biden leading. They were also among the first results that appeared when users searched for election results.
After media reports pointed out the issue, YouTube removed the video streams, citing its policy prohibiting spam, deceptive practices and scams.
On Wednesday, One America News Network, a conservative cable news network with nearly a million subscribers on YouTube, also posted a video commentary to the site claiming that Mr. Trump had already won the election and that Democrats were “tossing Republican ballots, harvesting fake ballots and delaying results” to cause confusion. The video has been viewed more than 280,000 times.
Farshad Shadloo, a YouTube spokesman, said the video did not violate the company’s policy regarding misleading claims about voting. He said the video carried a label that the election results are not final. YouTube added that it removed ads from the video because it does not allow creators to make money off content that undermines “confidence in elections with demonstrably false information.”
Alex Stamos, director of the Stanford Internet Observatory, said the tech companies still had a fight ahead against election misinformation, but were prepared for it.
“There will always be a long tail of disinformation, but it will become less impactful,” he said. “They are still working, for sure, and will try to maintain this staffing level and focus until the outcome is generally accepted.”
But Fadi Quran, campaign director at Avaaz, a progressive nonprofit that tracks misinformation, said Facebook, Twitter and YouTube needed to do more.
“Platforms need to quickly expand their efforts before the country is plunged into further chaos and confusion,” he said. “It is a democratic emergency.”
