Meta, TikTok, YouTube and LinkedIn didn't act on 77% of misinformation reported during Áras race

No action was taken on about 87% of non-AI election-related disinformation.

SOCIAL MEDIA COMPANIES failed to act on the vast majority of electoral disinformation reported by researchers ahead of last year’s presidential election, a new report shows.

The report also criticised the companies for not demonstrating an understanding the context around the election, which was held in October last year.

The 2025 election, which resulted in Catherine Connolly becoming the tenth President of Ireland, saw the use of AI-generated deepfake videos and smear campaigns, as well as a coordinated drive on social media to encourage people to spoil their votes. 

The research, published today by the Ireland hub of the European Digital Media Observatory (EDMO), showed that no action was taken on 72 of the 85 items that were reported for breaking the platforms’ own rules on election-related disinformation.

EDMO’s report only looked at platforms that agreed to take part in a programme that allowed their researchers to flag election-related misinformation and disinformation with them directly.

Those platforms included LinkedIn, Facebook, Instagram, YouTube and TikTok.

Other platforms, such as Elon Musk’s X, were not included in the research.

X was fined €120 last year for breaching the Digital Services Act (DSA), which came into force in 2024 in part as a response to election interference. The act regulates how large social media companies must respond to and reduce the risk of misinformation.

The Journal FactCheck is an EDMO Ireland partner.

Some platforms, including Facebook and LinkedIn, didn’t take any action at all on the types of posts reported to them.

YouTube only took down reports of election disinformation under the different category of “AI-generated election disinformation”. 

The EDMO report suggests that many of the platforms were poorly prepared for misinformation ahead of the election.

“During meetings, platform representatives often appeared unfamiliar with the context of the unfolding election,” the report reads. “Many were unable to articulate the rationale for content-moderation decisions.”

Election misinformation

The risk of disinformation on social media during democratic elections is real and documented. In 2016, large amounts of misinformation were spread ahead of the US presidential vote — which was amplified by Kremlin-instigated propaganda campaigns — as well as the Brexit referendum.

Multiple instances of major disinformation campaigns have been documented in elections since, including the 2021 Stop the Steal campaign in the United States that resulted in the 6 January riots.

Other recent campaigns have included the use of AI-generated audio and videos in an attempt to undermine electoral events.

In the 2025 presidential election in Ireland, an AI-generated deepfake video appeared to show Catherine Connolly withdrawing from the race days before the election.

The fake news report, posted on social media, was quickly debunked — in part due to the uploader’s account featuring earlier, less convincing AI-generated videos.

In actuality, Connolly did not withdraw and would go on to win the election.

The campaign period also saw an online smear campaign making personal allegations against one of the candidates, Jim Gavin, by Kieran Kelly, a businessman who is currently based in Asia and is known for spreading misinformation on social media.

The election run up also featured a campaign to secure a nomination by MMA fighter Conor McGregor, who used his celebrity status to spread multiple false claims about the Irish presidency, including that presidents can unilaterally call elections, and about the election process.

McGregor had claimed he had secured enough nominations to run for president, though this never came to pass. He misleadingly claimed that only candidates approved by the government can run for the office. 

Connolly, the ultimate winner, ran without nominations from the government.  

The presidential election also saw multiple instances of abuse against candidates and their campaigns, including one incident, filmed and shared online, in which a man allegedly punched a canvasser.

The man was arrested and the legal case against him is ongoing.

The presidential election

EDMO partners, including The Journal FactCheck, monitored social media for election misinformation.

Potential misinformation was flagged with researchers at Dublin City University (DCU), including analyst Aidan O’Brien and Dr Eileen Culloty, associate professor at the DCU School of Communications.

The researchers then checked to make sure these examples broke the platforms’ own rules or terms of service, and were given direct channels to report this content onto the platforms themselves.

Content was monitored throughout the election campaign to see whether the platforms took action after potential misinformation was reported, including whether a warning was added to posts containing misinformation or whether they were removed entirely.

In most cases, no action was taken.

Less than half of the 177 items that were reported to the platforms during the campaign had any form of action taken on them, according to EDMO’s report.

However, this varied by category. Two posts were reported for incitement to violence: one on Instagram and one on YouTube; both of these were acted upon.

Nine of the 13 “harassment and abuse” posts were acted on, and the platforms responded to all 49 reports of illegal content.

Reports of illegal content generally consisted of photos shared on social media of filled in ballots — which were usually spoiled — taken at polling stations. It is illegal to take or share images from polling stations because doing so is deemed to compromise the integrity of elections in Ireland.

In contrast, the platforms only took action against 13 of the 85 pieces of election-related disinformation that was not generated by artificial intelligence.

LinkedIn, Facebook, and YouTube took no action on these at all.

Nine of the videos researchers reported for breaking YouTube’s rules had been made by monetised accounts, meaning the uploaders could make money from these views.

There were an additional 26 reports of AI-generated election disinformation — half of which were acted upon.

Taken together, reports of AI and non-AI disinformation were acted on only about 23% of the time, according to the report. 

LinkedIn did not act on any of the seven reports of content breaching its rules that it received ahead of the election. Other platforms’ responses ranged from 44% (Facebook) to 59% (Instagram) — both of which are owned by Meta.

LinkedIn said that they actually had removed one piece of content that was reported to them, but that this was not included in EDMO’s project scope. They said that the other reported content had not broken their policies.

The EDMO research also said that, while many of the items reported stayed active for days, TikTok had the fastest response time, taking about six hours on average to act on reported content.

According to TikTok, 97.5% of the content they removed for violating policies on civic integrity, misinformation or AI generated content was done before it had been reported to them. This was based on stats from 22 September to 19 October.

They also said that their in-app election information centre was viewed more than 110,000 times. 

At the time of writing, YouTube and Meta have not responded to requests for comment.

The EDMO research also noted that there was “a lack of access to political advertisements surrounding the 2025 Irish Presidential Election and the period leading up to it”, with both Google and Meta removing the researchers’ access to campaign ads.

All 49 examples of illegal content, such as images of spoiled votes, were acted upon after EDMO reported them. However, the report says that the platforms themselves made little effort to prepare for such posts, even after EDMO forewarned them it was a risk.

“Not only did none of the platforms have policies in place for this eventuality,” the report notes, “only one platform articulated its policy position on content relating to the ‘spoil the vote’ campaign.

“Unfortunately, this was provided on 29 October, five days after the election.”

Close
JournalTv
News in 60 seconds