Dutch Prime Minister Scrutinized for Using AI to Write Social Media Posts
The Hague, Thursday, 10 September 2026.
Dutch Prime Minister Rob Jetten faces scrutiny after revelations he used artificial intelligence to draft highly sensitive social media posts, including messages commemorating the MH17 plane crash disaster.
The Scale of Automated Communications
An investigation by the Dutch newspaper de Volkskrant has revealed that Dutch Prime Minister Rob Jetten and his political party, Democrats 66 (D66), utilized generative artificial intelligence models to draft hundreds of social media posts between 2021 and 2026 [1][2]. Using the advanced AI-detection tool “Pangram,” which de Volkskrant adopted in July 2026 after extensive validation showing zero false positives on pre-2022 texts, researchers analyzed 1,812 unique messages [2]. The investigation uncovered at least 232 unique AI-generated posts distributed across 434 channels, including X, Instagram, Facebook, and LinkedIn [1][2][3]. Of these 232 automated messages, 68 were published directly on Prime Minister Jetten’s personal profiles, while 164 originated from D66 party accounts [1][2].
The Scale of Automated Communications
The reliance on tools like ChatGPT and Claude escalated dramatically following the collapse of the Schoof I Cabinet on June 3, 2025 [1]. Starting June 4, 2025, approximately 30% of all subsequent social media posts published by Jetten and D66 were produced using artificial intelligence [1][2][3]. This calculation shows that the party’s output became heavily automated: of the 232 AI-generated posts, Jetten’s personal accounts were responsible for 29.31% of the total automated content, while D66 accounted for 70.69% [1][2]. Furthermore, the investigation revealed that AI was not confined to social media; it was also used to draft 1.5 pages of Jetten’s book, Hoe het wél kan, and to formulate lines in three D66 campaign videos [1][2][3].
Rhetorical Patterns and Technical Failures
Linguistic experts and researchers have identified distinct technical markers that expose the automated nature of Jetten’s communications. Analysis of the 232 unique AI-generated posts revealed a recurring rhetorical device known as the “drieslag” (three-part list), which appeared in over 80% of the texts, totaling 384 instances [2]. Additionally, these automated posts demonstrated a frequency of contrasts more than double that of human-written material [3]. While typical AI-generated texts exhibit a 30% higher frequency of em-dashes (gedachtestreepjes) compared to human writing, Jetten and D66’s AI posts showed a marked decrease in em-dash usage starting June 2025, a shift that experts suggest was a deliberate effort to mask the automated origin of the texts [2].
Rhetorical Patterns and Technical Failures
Despite the sophistication of these language models, the automated process introduced notable factual errors and logical inconsistencies. In June 2025, an AI-generated LinkedIn message regarding the Dutch housing market contained a fundamental error regarding the sequence of the moving chain [2]. Later, in June 2026, an Instagram post proposed a highly impractical solution to prison cell shortages by suggesting that prisoners could perform labor outside while simultaneously remaining confined in their cells [2][3]. Suzan Verberne, a Professor of Natural Language Processing at Leiden University, described these posts as “middle-of-the-road texts” that lack genuine spontaneity and substance, pointing out that AI models merely prioritize the statistically most probable next word rather than actual, meaningful communication [2][3].
The Ethics of Automated Empathy
The controversy deepens when examining the sensitive nature of the topics delegated to machine generation. The AI-generated content addressed major national and international milestones, including the war in Ukraine, the commemoration of the MH17 aviation disaster, and the government’s formal apology to the Moluccan community [1][2]. In June 2026, Jetten visited Kamp Westerbork—a historic site that served as a Holocaust transit camp and later housed thousands of Moluccan families—and subsequently posted AI-generated reflections on Facebook and Instagram to his 500,000 followers [2]. Out of the 68 AI-generated posts on Jetten’s personal accounts, 41 were written in the first person using the pronoun “I” (“ik”), creating a false impression of direct, personal reflection [2].
The Ethics of Automated Empathy
This automated approach to deeply personal and emotional topics has drawn sharp criticism from communication and linguistic experts. Marc van Oostendorp, a Professor of Dutch at Radboud University, argued that using AI to express profound emotions at historical sites like Kamp Westerbork is “the opposite of transparency,” leaving citizens to wonder if the Prime Minister’s expressed feelings are genuine or merely statistically desirable emotions [2][3]. Claes de Vreese, a Professor of AI and Society at the University of Amsterdam, warned that having a computer generate personal messages on sensitive topics dilutes the Prime Minister’s authenticity and severely damages public trust in official government communication channels [1][2].
Political Accountability and Regulatory Gaps
In response to the revelations, D66 party representatives defended their workflow, stating that every post begins and ends with the oversight, review, and refinement of human staff [1][3]. They asserted that the AI-generated texts are based on personal input from Jetten himself, who subsequently reviews and approves them, though the party admitted that some posts are occasionally approved without any modifications [2]. This defense stands in stark contrast to the political positions of other coalition leaders; a comparative analysis of the social media accounts of Dilan Yeşilgöz (VVD) and Henri Bontenbal (CDA) revealed virtually zero AI-generated content [1][2].
Political Accountability and Regulatory Gaps
The controversy also highlights a significant political contradiction regarding transparency regulations. D66’s 2024 European election platform explicitly advocated for mandatory “AI labels for all content that has been generated with the help of AI” [2][3]. However, the party has failed to apply any such labels to their own social media accounts [2][3]. While the European Union’s AI Act, which officially entered into force on August 2, 2026, mandates strict transparency for AI-generated content, it provides exemptions for texts that undergo human editorial review—a loophole D66 has utilized to justify their lack of disclosure [2][3]. Claes de Vreese noted that while this practice may formally comply with the letter of the law, it completely violates its spirit, leaving a major credibility gap for the Prime Minister [2].