From deepfakes to breathless TikTok videos, a wave of disinformation has hit Taiwan's voters ahead of Saturday's presidential election, aimed overwhelmingly at candidates China opposes.
Experts and Taiwanese officials say this campaign is linked to Beijing, which has made no secret of its disapproval of frontrunner Lai Ching-te, whose Democratic Progressive Party (DPP) rejects the Chinese claim that Taiwan is its territory.
One Chinese hashtag mocking Lai got more than 8.5 million views, and the responses to posts and videos against the DPP were replete with derogatory comments and conspiracy theories.
"She's a foreigner," commented one TikTok user on a video featuring DPP opponents loudly claiming Lai's running mate Hsiao Bi-khim was secretly a US citizen -- and thus ineligible.
Hsiao has repeatedly said she gave up her US citizenship years ago, and AFP found her name on a US government list of people who have renounced their nationality.
Still, the unfounded claim about Hsiao's US citizenship has been the most persistent social media allegation during the election campaign.
Experts say it illustrates a Beijing-linked push to use disinformation or partially true claims to discredit any politicians who do not accept Taiwan as a part of China.
Beijing has ramped up military pressure on Taiwan in recent years, but analysts say it is also attempting to sway public opinion away from pro-independence views.
"(China-led) information campaigns are sustained, systematic and produce huge volumes of misinformation -- anything to discredit... the legitimacy of Taiwan's democracy," Jonathan Sullivan of the University of Nottingham, told AFP.
"Taiwan faces an extremely well-resourced and motivated rival for 'hearts and minds'."
Beijing has dismissed the allegations as "rumours and hype".
Starting as early as May 2022, accounts posing as Taiwanese users began disseminating videos and political memes in a "sustained and coordinated effort", the US research firm Graphika said in a recent report.
Lai has accused China of using "all means to interfere with this election", including disinformation.
China has long opposed the DPP, whose Tsai Ing-wen became president in 2016. She considers Taiwan a sovereign state and does not accept China's claim.
With the DPP candidate the frontrunner, China has framed the vote as a choice between war and peace.
It is a theme echoed in much of the disinformation or misleading content identified by AFP, other fact-checking organisations and Taiwanese authorities: the DPP is acting against Taiwan and at the behest of the ill-intentioned United States.
While many of these videos include straightforward stitching together of commentary by DPP opponents, some now feature deepfakes.
One such fraudulent video emerged in November, showing Lai speaking to the press with the audio noticeably altered to make it sound like he was praising his party's Beijing-friendly opponents.
Taiwanese authorities swiftly took it down and described it as an attempt to influence voters.
An AFP Fact-Check investigation found four anti-DPP videos that first appeared on Douyin, followed by multiple clones on TikTok with hundreds of thousands of views.
One video featuring footage of people criticising Lai over a railway project appeared on Douyin on December 19 and then landed on TikTok the same day, timestamps collected by AFP showed.
Another indication that the videos first appeared on Douyin is the use of simplified Chinese characters in captions.
While mainland China uses those characters, Taiwan uses the traditional script.
The videos were viewed by a far larger audience on TikTok.
One got nearly 20 times more views on the platform than Douyin, according to AFP Fact Check.
"These messages and videos often come from China, and most of the content is about distrust of some government policies," Charles Yeh, founder of the fact-checking group MyGoPen, told AFP.
AFP, along with more than a dozen fact-checking organisations, is paid by TikTok to verify videos that potentially contain false information.
TikTok then removes some videos if the information is shown to be false by AFP teams, or flags them as misleading to users on the platform.
