Disinformation in Spanish is prolific on Facebook, Twitter and YouTube despite vows to act
Last year, US lawmakers urged the CEOs of major tech companies including Facebook, Twitter and YouTube to do more to combat disinformation spreading in Spanish, warning that inaccurate information on key issues such as vaccines and the presidential election was proliferating on their platforms.
“There is significant evidence that your Spanish-language moderation efforts are not keeping pace, with widespread accounts of viral content promoting human smuggling, vaccine hoaxes, and election misinformation,” the lawmakers wrote in a July 2021 letter. “Congress has a moral duty to ensure that all social media users have the same access to truthful and trustworthy content regardless of the language they speak at home or use to communicate online.”
More than a year later, and with the midterm elections fast approaching, advocates say these social media platforms are still falling short on policing such content – particularly when it comes to non-English languages.
With Spanish-speaking voters making up a significant part of the US electorate – Latino voters constituted the second largest voting block in the 2020 presidential election – the failure to eradicate misinformation in Spanish from social media platforms amounts to aiding and abetting disenfranchisement, said Mariana Ruiz Firmat of advocacy group Color of Change.
“This kind of nonchalant approach, where companies turn their heads away from the threat, shows how little they value protecting or caring about Latinx users who rely on their platforms to gain crucial access to information about voting,” said Ruiz Firmat.
Experts say misinformation narratives in Spanish often mirrors those seen in English, falling into the two main categories: politics, or health and vaccines.
The most urgent narrative being tracked by researchers is what is being called “the big lie” – the baseless theory that Joe Biden did not win the 2020 election legitimately.
That claim has become widely-believed on the right, with 70% of Republican voters backing the “stolen” election theory, according to a recent Politifact study. And it is continuing to spread on social media, in English and Spanish, preemptively creating doubt about legitimacy of the midterm vote and alarming experts.
“When this lie proliferates, it undermines trust in democracy and lowers the likelihood people will vote,” said Jessica J González, co-chief executive officer of civil rights group Free Press. “People are less likely to vote if they think their vote doesn’t matter, that their vote isn’t counted, or that there’s major corruption and fraud in a system.”
An August report by Media Matters for America found many Spanish-language videos pushing the big lie were still up on social media platforms, despite policies prohibiting them.
The report detailed violating content including baseless claims that voting machine glitches in 2020 allowed one candidate to win a swing state, allegations of fake ballots lending to another candidate’s win, and more claims that dead people voted in large enough numbers to change the 2016 election results.
Media Matters identified three Spanish-language YouTube channels that have violated content policies numerous times but remain online, with a combined subscriber count of more than 880,000.
Free Press and a coalition of other civil rights groups say they’ve been pushing big tech to take misinformation in English and other languages seriously for months, but have found the companies are not responding quickly or thoroughly enough.
González says her anti-misinformation coalition Change the Terms has attempted to engage with major
tech companies including Meta, TikTok and YouTube parent company Google regarding Spanish misinformation on their platforms – but has not seen sufficient concrete action.
Specifically, the group asked YouTube for more information on how well it is policing Spanish election misinformation and claims it did not receive it, and asked Facebook to outright ban theories that the 2020 election was stolen on its platforms. The platform has not, according to the advocates.
In fact, many of the theories posted in 2020 remain prevalent on major platforms, said Jacobo Licona, a disinformation research lead for Equis Labs, a polling firm focused on Latino voters.
“There are still Spanish-language posts active today from November 2020 that promote election lies with no warning labels,” he said. “A lot of these narratives are being recycled, and a lot of the original narratives persist.”
The continued proliferation of misinformation in Spanish is due, at least in part, to a lack of investment experts say – including failing to hire human moderators fluent in these languages or training artificial intelligence on the languages.
In 2021, former Facebook employee turned whistleblower Frances Haugen confirmed in a Congressional hearing that that platform invests significantly in moderation in the US, but that 87% of misinformation spending at Facebook is on English content when only 9% of users are English speakers.
“Facebook invests more in users that make them more money, even though danger may not be evenly distributed based on profitability,”
Haugen told lawmakers.
An internal Facebook memo released by Haugen revealed the company assessed its ability to detect antivaccine rhetoric and misinformation as “basically non-existent” in non-English comments. Facebook has since made improvements to Spanish misinformation prediction models, said Meta spokesperson Aaron Simpson, and they are now working at a level of precision similar to English for content the company sends to fact-checkers for review.
Tech companies say they have been working at tackling those discrepancies ahead of the midterms, with measures to combat misinformation – including in Spanish.
Facebook made improvements to Spanish misinformation prediction models since the Haugen revelations, said Meta spokesperson Aaron Simpson, and Meta, Facebook’s parent company, is now working at a level of precision similar to English.
The company says it works with 90 independent fact-checking organizations around the world to review and rate viral misinformation in more than 60 languages on Facebook and Instagram. Of the 11 companies it works with in the US, six review content in Spanish, said Simpson.
“We’ve invested heavily to combat Spanish misinformation on our platforms, and it mirrors our strategy to address English misinformation,” Simpson said. “We remove Spanish-language voter interference content, and we connect people with authoritative information in Spanish through our voting alerts and voting information center.”
Meta has also invested $5m in media literacy initiatives ahead of the midterms, including fact-checking services on WhatsApp after the app was identified as a large source of misinformation in 2020.
Youtube says it applies its misinformation policies globally, “and we apply them consistently regardless of the language”, according to spokesperson Elena Hernandez.
YouTube also employs humans to assist its artificial intelligence led moderation system, with more than 20,000 people around the world working to review and remove content that violates its policies, including Spanishspeaking employees, Hernandez said. She declined to share how many employees are able to moderate non-English languages.
Twitter is working with organizations such as National Association of Latino Elected and Appointed Officials and MiFamiliaVota to promote voter registration, a spokesman said, in addition to misinformation efforts in English and Spanish. It is also working with Spanish Language Disinformation Coalition to inform its policies on misinformation ahead of Midterms.
“Our goal is to preserve space for robust debate while ensuring people have the context and control they need to make informed decisions about the content and accounts they see and engage with on Twitter,” said Twitter spokesperson Lauren Alexander.
Twitter declined to share how many human moderators it employs, or how many of them speak Spanish.
Meanwhile, TikTok has launched an election center in more than 45 languages to flag inaccurate content and connect users to authoritative information, said Ben Rathe, a TikTok spokesperson.
“We take our responsibility to protect the integrity of our platform and elections with utmost seriousness,” he said. He declined to share how many human moderators the company employs.