Immigration

Many debates focused on immigration again. Consequently, so did many fact-checks.

• In recent years, more people have died than have been born in the Netherlands. Natural population growth has stagnated, and the population is only increasing due to immigration. Immigration, and to a lesser extent, the aging population are important themes in the party platforms, but Nieuwscheckers’ overview of demographic developments shows that the solutions are not simple.
• According to Jimmy Dijk (SP), the immigration problem is actually a class issue. On the program WNL Op Zondag, he stated that the reception of asylum seekers is unevenly distributed: the twenty poorest municipalities reportedly host twenty times as many asylum seekers and refugees as the richest municipalities. That is incorrect: the poorest municipalities actually host roughly twice as many asylum seekers. This is also stated on the SP website, so Jimmy Dijk probably misspoke. [Nieuwscheckers, DVHN] 
• During the RTL partly leaders’ debate, Joost Eerdmans (JA21) said that a thousand asylum seekers come to the Netherlands every week. Although this is correct for October, it’s not representative for the whole year. It usually happens that more asylum seekers arrive in the fall, and in 2025, the number of asylum applications in the Netherlands was lower than in previous years. [AD]  
• Does Frans Timmermans really live near an asylum seekers’ centre as he claims? The AD newspaper looked at the neighbourhood in Maastricht  where he lives and concluded that the nearest asylum centres are three to six kilometres away, on the other side of the Meuse River.
• According to Timmermans, deals Spain made with other countries to take back rejected asylum seekers have been successful. However, the figures he refers to cover all deportations, not only people deported through those deals specifically. Furthermore, the figures for the Netherlands and Spain are difficult to compare due to differences in definitions. [Nieuwsuur]
• Are 60 percent of the homeless people in the Netherlands former immigrant workers, as Henri Bontenbal claimed? This claim is unfounded: it is an estimate made by the Salvation Army but it only concerns homeless people who are living on the streets. [AD] 
• Abuses at employment agencies for immigrant workers must be eliminated, argued BBB Minister Mona Keizer on Eva Jinek’s talk show. However, when a law aimed at tackling these agencies was voted on in April, the BBB was one of the two parties to vote against it. [de Volkskrant]
• Last week, the municipalities of Valkenswaard, Eersel, and Breskens filed complaints that people posing as municipal employees were spreading the lie that new asylum centres would be established in those municipalities. There have been no new reports since then. It is unknown who is behind the fake news. [NU.nl]

How many social housing units are allocated to refugees?

During the SBS6 debate on Thursday, the party leaders clashed over whether 7, 17, or even 37 percent of social housing units will go to refugees. The initial figures come from Statistics Netherlands (CBS), the latter from mathematician and anthropologist Jan van de Beek, a popular figure with the PVV. After a fact-check, the AD newspaper concluded that “at least one in five social housing units that became available for first-time buyers in 2023 went to refugees.” The NOS, less definitively stated: “At its core, the debate about figures revolves around the question: are you looking at all social housing units going to refugees or only at the ‘starter homes’ […]?”
Van de Beek has repeatedly mentioned a different percentage of housing units allocated to refugees compared to first-time buyers in recent years. [Feit of Fake]
LUBACH also addressed issues related to refugees, social housing, and the law proposed by Housing Minister Mona Keijzer (BBB) ​​that aims to prohibit giving refugees priority.

Healthcare: Basic health insurance package not strictly frozen after all

D66 and the VVD want to cut healthcare costs by not reimbursing new medicines or treatments, for example for cancer, within the basic health insurance package. Frans Timmermans criticised the party leaders of this in the election debates. Indeed, the CPB (Netherlands Bureau for Economic Policy Analysis), which analysed the election manifestos, noted the request of D66 and VVD to “freeze” the basic health insurance package: no new medicines would be added. The CDA wants to allow them to be included to a limited extent, and GroenLinks-PvdA wants to keep the package open.

However, the “freeze” scenario that D66 and VVD recommended will likely not actually occur after the elections: both Jetten and Yeşilgöz denied that this scenario was their actual intention. [Volkskrant, AD, DVHN, RTL Nieuws]

VVD exaggerates consequences of abolishing mortgage interest deduction

Plans by the CDA and GroenLinks-PvdA parties to abolish the mortgage interest deduction are a breaking point for the VVD, Yeşilgöz said. She claims this would cost households 400 to 500 euros per month. This is incorrect. Only homeowners with a new, above-average mortgage would lose such an amount if the deductions were to be abolished immediately. Parties in favour of this abolition, only want to do so in the longer term (the CDA even wants to do so over a period of 30 years). Moreover, they intend to use the savings to reduce other costs, for example by lowering income tax. [Pointer, Radar, Volkskrant]

However, tax experts warn that wealthy homeowners will be able to deduct their mortgage interest in Box 3 starting in 2028, which could increase income inequality. [Follow the Money, FD]  

Political AI images: seduction more important than deception

AI tools can generate increasingly realistic images. Based on just two photos, AI was able to make a VRT journalist gallop on a camel while being chased by lions. AI is also making an appearance in the fashion world: Zalando advertises using AI models. This week, the British Channel 4 aired a programme  (“Is AI Taking Over My Job?”)presented by AI journalist “Aisha Gaban”. Omroep Brabant had newsreader Nina van den Broek cloned last year.

Abuse

Abuse of AI tools is an obvious concern. AI-generated pornographic images of real people are easy to find via social media and search engines. Fake ads featuring photos of Dutch celebrities were already a serious problem, but now perpetrators are also trying to mislead victims with deepfake videos. This week, André van Duin warned the public about ads in which he appears to be promoting medications for high blood pressure. YouTube has introduced a tool that allows popular YouTubers to detect and remove deepfakes of themselves.

Political propaganda

This technology is also used for political propaganda, particularly by the radical right. However, it is not primarily used to deceive people with quasi-real images, as we expected. According to Teresa Weikman (University of Amsterdam), AI images are primarily used to ridicule opponents, provide visuals for the party’s ideas, entertain, and provoke. The intention is more to entice voters to go along with the party’s narrative than to mislead them with fake news.

Examples of this use of AI can be seen in the approximately one thousand AI images of party leaders collected by Pointer. Yeşilgöz is depicted as Gargamel, Timmermans as the Ayatollah, and Wilders as a clown. The most frequently depicted politicians were Timmermans and Wilders: Timmermans was almost always depicted negatively, while Wilders was depicted positively in more than half of the cases. Yeşilgöz was exclusively depicted in a negative manner, often with a Pinocchio nose.

On TikTok, the PVV is being targeted in AI videos where older people shout at young people of colour: “I vote for the PVV!” AFP Fact-check uses these videos to explain how to distinguish AI from real life.

Chatbots blunder on voting advice and news

Chatbots may seem like oracles you can ask anything, but they are not suitable for voting advice, warned the Dutch Data Protection Agency this week. The privacy watchdog compared four chatbots with the results from Kieskompas and Stemwijzer. In more than half of the answers, the bots only suggested two parties: the PVV and GroenLinks-PvdA. “Chatbots present themselves as neutral, but they are not,” they noted. Last week, RTL Z published a similar study, in which the bots primarily gave left-leaning voting advice. This likely has to do with the way the research was set up: a chatbot’s answer heavily depends on what questions you ask and how you ask them exactly.

Google’s AI answers to election-related questions are also unreliable, according to research from the University of Amsterdam. Google drew relatively heavily on party websites, particularly D66 and GroenLinks-PvdA, and often cited sources that were irrelevant to the questions.

When it comes to following the news, chatbots perform no better. A significant portion of the news information they provide is incorrect, according to a study by the European Broadcasting Union (EBU) and the BBC. The study tested free versions of four chatbots (ChatGPT, Copilot, Perplexity, and Gemini). The bots were asked 30 questions in 19 different languages, such as: “Who is the Pope?” or “Can Trump become president for a third term?” The answers were assessed for accuracy and sourcing by journalists from 22 European news organizations. Approximately 45 percent of the answers had flaws in accuracy, sources, and context. About 20 percent contained obvious errors, such as fabricated details, non-existent quotes, and outdated information.

Research study: Fact-checking can be especially harmful to unknown politicians

Last week, we discussed research showing that politicians whose falsehoods are fact-checked live on TV become less popular, particularly among voters who do not have a clear preference for a politician or political party. Does this effect also occur on social media? One American study suggests that it does. Only voters without a strong political preference participated, ensuring the results were not influenced by prior biases.

The researchers recruited over four hundred Americans to participate in a simulated election campaign on social media. The participants were asked to imagine they had moved to Malta and had to vote for a Maltese politician. Malta is a real country, but one that is unfamiliar to the participants, which allowed the researchers to observe reactions to new, unfamiliar politicians, without prior preferences.

The participants were instructed to familiarise themselves with the election by viewing a timeline that mimicked a real social media platform. On this timeline, they saw posts by and about two fictional candidates: one slightly left-leaning and one slightly right-leaning. The timeline included neutral, positive, and misleading posts (for example: “When I was Minister for Employment, unemployment fell dramatically”). In some cases, an independent, also fictional, fact-checking platform responded to these posts by rating them as true or false and briefly explaining why.

The researchers then asked participants to rate the candidates in terms of competence, honesty, and likeability, and to indicate who they would vote for.

The study found that uncorrected misinformation led participants to think more positively of, and vote more often for, the candidate spreading the falsehoods – especially if that candidate was right-wing. However, once the misleading statements were fact-checked, the participants’ appreciation and preference for the politician in question dropped sharply.

This study, therefore, highlights that fact-checks are particularly impactful when voters do not yet have a fixed view on a politician. Established figures like Wilders or Timmermans are unlikely to lose support if caught in a falsehood, but for new or lesser-known candidates, being corrected by a fact-check can be detrimental.

More election news

• The Rathenau Instituut will, at the request of the Dutch House of Representatives, investigate which technological possibilities social media platforms offer to favour political candidates. The research results of the study will be published in the run-up to the 2026 municipal elections.
• Follow the Money interviewed Bits of Freedom director Evelyn Austin, who is concerned about the spread of disinformation and hate speech on social media. She believes the algorithms of social media platforms like TikTok facilitate threats and hate speech because tech companies have a financial interest in doing so.
• Research by EenVandaag shows that TikTok directs users interested in politics mainly towards right-wing parties like the PVV and Forum voor Democratie. Five accounts were created across the political spectrum: the conservative-right and centrist accounts mostly received right-wing content, while the left-wing and undecided voters’ accounts were shown few political videos.
• Despite growing concerns on disinformation surrounding the parliamentary elections, municipalities have not yet reported this to the Netherlands Authority for Consumers and Markets (ACM). The ACM is involved in the implementation of the European Digital Services Act [Binnenlands Bestuur]
• The podcast Over Mediadoctoren speaks with political scientist Armèn Hakhverdian (University of Amsterdam) about election myths. For example, is it true that workers who formerly voted for the PvdA now vote for the PVV? Are young people more left-leaning than previous generations? And was the Netherlands ever a left-wing country?
• Zombie news: although it has been debunked multiple times, a fake article claiming that Jesse Klaver wants to abolish private property is circulating again during this election. [AFP]

Header picture: European Parliament, CC BY 2.0 <https://creativecommons.org/licenses/by/2.0>, via Wikimedia Commons