AI-driven digital manipulation ‘tested’ Dutch election integrity, researchers warn

There was significant digital interference during the October 2025 Dutch parliamentary elections, including new generative AI threats, targeted manipulation, and moderation failures, according to a new report by the Dutch Hybrid Election Integrity Observatory (HEIO). 

The election "took place under unprecedented digital pressure that tested the resilience of democratic institutions and exposed vulnerabilities in the digital information ecosystem," it said.

HEIO is a collective of digital organisations, including The University of Amsterdam, AI Forensics, and Post-X Society, who observed and tracked political accounts and posts during the October election period. 

Using digital means to interfere in elections, such as posting false information on social media, is becoming increasingly common.

For example, in 2024, Romania's top court annulled the first round of votes in its presidential election due to suspected interference from a Russian social media campaign, promoting a pro-Russian candidate. 

Digital Interference 

Back in the Netherlands, HEIO's report found an increased use of AI to push political boundaries and spread false information, through AI-generated posts, songs, and deepfakes. 

While the total number of AI-generated posts about the Dutch election remained small, the technology proved very effective at generating engagement, depending on the platform.

Non-AI posts performed better on TikTok than generated posts; but on Facebook, AI political posts had about 23 times more median engagement than non-AI posts.

The HEIO survey pointed to one Facebook page which generated AI content targeting minorities, immigrants, and political opponents of the far-right Dutch Party for Freedom (PVV).

Geert Wilders is the leader of the PVV party, and the Facebook page was entitled: "We will not press charges against Geert Wilders." 

It is unclear which charges the title referred to, but the page received 74 million views between June and October 2025.

Much of the generated content had common narratives, including Dutch citizens fighting against immigrants for resources and economic anxiety. 

"Parties with fewer normative constraints, frequently positioned on the far-right, as well as satirical and parody accounts, showed elevated adoption rates", the report said.

But the Dutch Green politician and former EU commissioner, Frans Timmermans, also had thousands of Vietnam-based Facebook accounts liking his posts.

"GenAI content produced by political actors gained 74m views, which is quite a lot. It changed the playing field," Pieter van Boheemen, director of Post X Society, told EUobserver. 

The quick speed and low effort required to generate this type of content is particularly of concern.

"And we are not sure whether we will be able to detect GenAI in the near future," van Boheemen added.

Beyond AI, the researchers found more traditional disinformation posts, along with accounts run in foreign countries interacting with Dutch election content. 

They found 550 accounts engaging on X from Nigeria, Ghana, and the Ivory Coast, and also 23,000 accounts engaging from Vietnam on Facebook.

Despite all the interference they catalogued, the report concluded that the elections were tested, but fundamentally free and fair.

Fixes?

The researchers say that social media platforms rarely remove political content that could violate policy unless it receives media attention. 

The HEIO researchers noted that, to get a moderation response from the platforms, they were forced to to amplify the harmful content they were researching.

"Platforms get away with soft promises on election protection," said van Boheemen.

"Enforcement years later is not sufficient when it comes to critical moments like elections," he added.

HEIO said platform moderation should be independently certified, while also seeing a need for European regulation to become more agile, with real-time enforcement and monitoring. 

Election interference is on the European Commission's mind.

The executive announced, in December 2024, that it was investigating TikTok under the Digital Services Act for election integrity issues stemming from the Romanian vote, for instance.

In November 2025, the European Commission also announced the European Democracy Shield initiative, which includes measures to combat online misinformation and foreign election interference.  

Become a subscriber and support EUobserver's journalism in 2026.