Elections under pressure
The Dutch elections were free and fair, but faced considerable pressure online. Election campaigns made extensive use of AI-generated content. We also observed coordinated manipulation campaigns, failures of platform moderation, and the exploitation of new techniques to spread disinformation and hate speech.
What we observed
And more. Read the full report.
In short
The Netherlands held general elections on October 29th 2025. Our consortium of researchers and civil society organisations tracked digital threats to election integrity.
We exposed troll armies, malicious use of Generative AI (GenAI) content, dubious political advertising, recommendations of extremism, disinformation on election fraud, platform moderation failure and other threats to election integrity.
Our alerts already led to immediate action from political actors, platforms, government and parliamentarians; more work is needed to improve accountability and transparency.
Frequently asked questions
Were the elections conducted fairly?
Yes. The HEIO consortium concludes that the Dutch parliamentary elections were fundamentally free and fair. In our assessment, the outcome reflects the genuine preferences of voters. At the same time, this conclusion must be immediately nuanced: the elections took place under digital pressure, marked by a surge in AI-generated content, coordinated manipulation campaigns, and demonstrable failures of platform moderation. Free and fair does not mean free from threat.
Did foreign interference affect the outcome?
That is difficult to establish with certainty, but we did document concrete attempts at interference. Thousands of accounts from Vietnam were identified placing likes on the Facebook page of GL-PvdA leader Frans Timmermans. On X, hundreds of accounts from Nigeria, Ghana, and Ivory Coast were found retweeting polarizing content and posts from far-right parties PVV and FvD. The strategy was not aimed at fabricating disinformation, but at amplifying existing societal divisions. Whether this directly influenced the election outcome cannot be proven.
What role did artificial intelligence play in these elections?
A significant role. An AI-generated protest song reached number two on Spotify’s Netherlands Top 50 and inspired thousands of TikTok videos. A coordinated Facebook page operation, presumably linked to the PVV, became the most popular political page in the Netherlands, accumulating 74 million views between June and October 2025. AI-generated images grew progressively more extreme throughout the campaign, including violent and hateful content targeting politicians and minorities. Particularly concerning is that AI-generated content will become increasingly difficult to detect as the technology improves.
Why were platforms unable to address these problems more effectively?
Platforms were structurally too slow and too limited in their response. Not a single piece of content reported through normal reporting mechanisms was removed. Even when it clearly violated the platforms’ own terms of service. Only when findings were made public through media coverage did platforms take action. Livestreams on TikTok became venues for death threats, antisemitism, and racism without intervention. HEIO concludes that platforms offer no meaningful protection to users or the democratic process.