Dutch privacy watchdog warns against using AI chatbots for voting advice
Source: Autoriteit Persoonsgegevens; via Anadolu Agency · 2025-10-21
The Dutch Data Protection Authority has warned voters not to rely on artificial intelligence (AI) chatbots for voting advice, saying the tools are biased and could undermine democratic integrity, the local broadcaster NOS reported on Tuesday.
According to the data protection authority's research, AI chatbots frequently recommended only two political parties, the far-right Party for Freedom (PVV) and the left-wing GroenLinks–PvdA alliance, when asked which party best matched the users' views.
The study compared four well-known AI chatbots with official Dutch voting tools such as Kieskompas and StemWijzer. It found that most political parties were not represented in the chatbots' recommendations.
"Chatbots seem like clever tools, but as a voting aid, they consistently fail," Monique Verdier, the vice-chair of the privacy watchdog, said. "As a result, voters may unknowingly be advised to vote for a party that doesn't best align with their preferences. This directly affects a cornerstone of democracy."
In more than 30% of the tested cases, PVV was listed as the top recommendation, while GroenLinks–PvdA appeared first in nearly 25% of cases. Centrist parties such as the Christian Democratic Appeal and the Christian Union were barely suggested, at 1.3% and 0.3%, respectively.
Researchers described this as a "vacuum cleaner effect," where left-leaning users were drawn toward GroenLinks–PvdA and right-leaning ones toward PVV, leading to what they called a "distortion and polarization" of the political landscape.
Verdier warned that the trend could threaten democratic processes, saying: "Because if, regardless of what you enter, you end up with two parties, you might think, 'I have to vote for that party,' even though that party doesn't align at all with your preferences and the things you consider important."
The Dutch Data Protection Authority urged developers to ensure chatbots refuse to provide voting advice and instead redirect users to official election guidance websites.
Lines of inquiry this paper opens 16
Research framings built by reading the notes related to this paper — the questions it feeds into.
How can AI systems reliably guide voters without introducing political bias?- Does designing chatbots to satisfy partisan users prevent them from reducing polarization?
- Do chatbots actually give consistent voting recommendations regardless of user input?
- Why do older voters show less willingness to consult AI for politics?
- What political information quality issues arise when AI tools guide voting?
- What accuracy do AI chatbots actually provide on election topics?
- Should election regulators require chatbots to refuse voting advice entirely?
- Do other AI assistants perform similarly on voter election questions?
- How many Dutch voters actually plan to use AI for voting advice?
- Why do chatbots trained on internet data show consistent political bias?
- What makes voting-advice tools like Kieskompas more reliable than chatbots?
- Can voters distinguish between confident chatbot answers and accurate ones?
- Does ChatGPT displace search engines or question-and-answer platforms?
- Can AI systems stay current with real-time candidate filings across states?
- How much does stated election information intent predict actual behavior?
- How often do chatbot news users actually return to original reporting?