For fans of pro tennis, European soccer and British tabloids, the mysterious Twitter account had a lot to offer.
Beginning last year, it retweeted news, most of it in English, about Roger Federer and the Premier League, and it shared juicy clickbait on Zsa Zsa, an English bulldog that won the 2018 World’s Ugliest Dog contest.
The account, @HKpoliticalnew, and more than 200,000 other Twitter accounts were part of a sprawling Russian-style disinformation offensive from China, Twitter now says, the first time an American technology giant has attributed such a campaign to the Chinese government.
China has long deployed propaganda and censorship to subject its citizens to government-approved narratives. As the nation’s place in the world grows, Beijing has increasingly turned to internet platforms that it blocks within the country — including Twitter and Facebook — to advance its agenda across the rest of the planet.
It has done so in part by setting up accounts on the platforms for its state-run news outlets, such as China Daily, to make a public case for its views. But that is quite different from using fake accounts to manipulate opinions surreptitiously or simply to sow confusion.
“The end goal is to control the conversation,” said Matt Schrader, a China analyst with the Alliance for Securing Democracy at the German Marshall Fund in Washington.
Twitter last month took down nearly 1,000 accounts that it said were part of a state-directed effort to undermine the antigovernment protests in Hong Kong. It also suspended 200,000 other accounts that it said were connected to the Chinese operation but not yet very active. Facebook and YouTube quickly followed suit. All three platforms are blocked in mainland China but not in Hong Kong.
The 3.6 million tweets that the accounts sent represented a campaign that was less sophisticated and more hastily assembled than the one Russia carried out during the 2016 United States presidential election, researchers at the Australian Strategic Policy Institute said in a report published this month.
Instead of taking the time to cultivate plausible yet fake online personalities, the campaign’s operators appear to have simply bought accounts in the shadowy global marketplace for social media influence, where followers and retweets can be had for cheap.
The accounts posted in Indonesian, Arabic, Portuguese and other languages. They promoted hookup services, posted about Korean boy bands and retweeted messages about pop-punk music.
How the languages of the tweets shifted
“As a Hong Kong person who loves Hong Kong, I really miss the Hong Kong of before, which was developed and ruled by law,” @derrickmcnabbx wrote in Chinese on June 15. The account’s location was described as “Georgia, USA.” Before this year, nearly all of its tweets were links to pornography.
The “blunt-force” approach, the authors of the Australian report wrote, suggested that the operation was likely to have been a “rapid response to the unanticipated size and power of the Hong Kong protests rather than a campaign planned well in advance.”
A spokesman for China’s foreign ministry said he had no knowledge of the matter when asked last month whether the government was behind the accounts that Twitter and Facebook took down.
In its announcement, Twitter said little about how it determined that the accounts it removed were state directed. The company said it routinely monitored for such campaigns but declined to comment further.
The Chinese government blocks Twitter’s service in mainland China, and yet some of the accounts were operated from unblocked Chinese internet addresses, the company said. Some of the activity was traced to addresses in Beijing, according to a person familiar with Twitter’s investigation who feared retaliation from the government and requested anonymity.
There are already some signs that Twitter has not halted the Chinese campaign entirely. Nick Monaco of Institute for the Future, a think tank in Palo Alto, Calif., identified 17 accounts that had strong similarities to those that Twitter took down but which remained active.
Some of the accounts tweeted messages that matched, word for word, ones that Twitter had deleted. They used the same third-party software as many of the accounts Twitter had removed to post messages with similar themes, in what seemed to be a coordinated manner.
After The New York Times presented Mr. Monaco’s findings to Twitter last week, the company shut the accounts down but declined to say conclusively whether they had been part of the same state-backed network.
Many of the accounts originally identified by Twitter had spread pro-government messages during other public-relations crises for Beijing. Large numbers of such messages began appearing in 2017, after the exiled businessman Guo Wengui began accusing senior Chinese leaders of graft, which raises the question of why Twitter did not remove the accounts sooner.
During some of their tweet campaigns, the accounts posted primarily during the workweek, a sign that the accounts were run by employees working on the clock. For months, one account posted messages smearing Mr. Guo at 12 and 42 minutes past the hour, suggesting that the activity was automated.
Some of the accounts appear to have been started by genuine users but were later hijacked.
The first four years of posts on one account, @emiliya_naum, read like those of an ordinary American teenager.
She tweeted longingly at @justinbieber and said she twerked in her room to celebrate Barack Obama’s 2012 election victory. She cataloged her moods and mused about her crushes.
“The guy I like and my best friend hate each other … #ThisIsntGood,” she wrote in 2012.
Then, like many Twitter users, she let her account fall silent — until this summer, when she re-emerged as a cheerleader for Hong Kong law enforcement.
“Hong Kong police, way to go, we support you!” she tweeted, in Chinese. “We understand your hardships!”
It could not be determined whether @emiliya_naum was originally operated by a real person. No accounts with that name were found on Facebook, Instagram or other major social platforms.
By and large, the accounts that Twitter took down struggled to go viral with their pro-Beijing messages. Many of their most retweeted posts were links to pornography and animal videos.
Elise Thomas, one of the authors of the Australian report, said that the low level of professionalism suggested that the campaign was not the work of the People’s Liberation Army or the Ministry of State Security, which have previously been linked to Chinese cyberespionage and information campaigns.
“I would be surprised if the P.L.A. was responsible because I would expect they would be more competent than this,” Ms. Thomas said.
Russia’s social media efforts ahead of the 2016 presidential election were cannier about finding and influencing audiences in the United States. The Internet Research Agency, a St. Petersburg company with Kremlin ties, tailored its trolling operations to sow maximum discord.
Titus C. Chen, a professor who studies Chinese social media at the National Sun Yat-sen University in Taiwan, said he believed that China had created its own analogues to the Internet Research Agency, though they operated far less openly.
Turning spam bots into propaganda mouthpieces would represent a natural evolution of techniques that Beijing has long used at home.
For years, China has used armies of pseudonymous keyboard warriors to flood domestic social platforms and news sites with pro-government comments.
In 2013, the head of China’s propaganda department said that in Beijing alone, there were more than two million people working to “strengthen guidance of online opinion,” including by posting comments on Weibo, China’s Twitter-like social media platform.
Samm Sacks, a China expert at New America, a Washington-based think tank, said the clumsiness of the Twitter operations showed that China was still “out of its depth in trying to shape the international narrative.”
“What works inside China doesn’t work internationally,” she said. “I think China is probably working through that now.”
Raymond Zhong reported from Beijing, and Steven Lee Myers and Jin Wu from Hong Kong. Kate Conger contributed reporting from San Francisco, and Davey Alba and Keith Collins from New York. Wang Yiwei contributed research from Beijing.
The New York Times