Polls dominate election media: what are the consequences?
amsterdam, vrijdag, 24 oktober 2025.
Kim van Keken examined the influence of polls on the election campaign and journalistic practices. Her research shows that polls are becoming increasingly important in the media and how they shape the agenda of both politicians and journalists. This raises questions about journalistic self-restraint and the responsibility of the media to use polls responsibly.
Polls as a news driver: the observation
Polls increasingly determine which topics are considered newsworthy: commentators, journalists and politicians repeatedly refer to measurement moments to shape and legitimise political stories, according to a recent journalistic study into the role of polls in the election campaign [1]. This dominance changes editorial choices — from headlines to who is interviewed — and places a greater responsibility on news organisations to interpret polling news carefully [1].
A real change in seat numbers — one calculation example
Polls for the 29 October 2025 election show an estimate where the PVV is expected to win 34 seats compared with 37 seats in the 2023 elections; this is a decline that can be expressed in relative terms as -8.108 percent change in seat numbers [4].
Why framing matters: polls influence what politicians and voters do
Political scientists warn that not only the figures themselves can have an effect, but especially the way polls are framed in reporting — for example highlighting minimal differences or constructing narratives around ‘winning’ or ‘declining’ parties — which can trigger self-reinforcing responses from politicians and voters [2]. Tom Louwerse (Peilingwijzer) emphasises that polls are useful, but that misuse — such as exaggerating small seat differences — can lead to undesired media and political dynamics [2].
Specific use of AI in journalism: automated poll analysis and production
A concrete application of artificial intelligence in newsrooms is the automatic retrieval, aggregation and generation of poll analysis: systems combine multiple polls, calculate sample-weighted averages or seat estimates and produce initial drafts of news items or graphics that editorial teams then edit [5][6]. European policy initiatives and strategies encourage organisations to use AI for scalable data processing and content workflow optimisation, including — explicitly — applications that monitor and visualise polls and public opinion [5].
How those AI systems work technically
Under the hood many of these tools use combined workflows: data-collection scripts pull in poll results, statistical modules (sometimes based on classical regression or Bayesian aggregation) harmonise definitions, and language models or template generators turn the outcomes into text proposals and graphics that can be placed directly into editorial CMSs [6][5]. Studiohonderd22 describes contemporary AI strategies and taxonomies that technically and organisationally support editorial use cases like these [6][5].
Benefits for news production and consumption
AI-driven poll products speed up news production (faster publication of summaries and visuals), make complex datasets more accessible to newsrooms with limited statistical capacity and help maintain consistency in repeated reporting — benefits that can increase newsroom efficiency and audience reach [5][6]. Automation can also enable 24/7 monitoring, allowing newsrooms to spot trends earlier that the public might otherwise notice only later [5].
Risks and drawbacks: misuse and epistemic dangers
At the same time automation and AI bring new risks: exaggerating small fluctuations, presenting unwarranted precision, and normalising a poll-driven agenda in which editorial nuance and checks on methodology fade into the background — exactly the concerns also raised in qualitative studies on poll dominance [1][2]. Moreover, public debate about digital wellbeing and technology misuse warns that careless use of automated content can contribute to polarisation and disruption of the democratic information environment [3].
Specific ethical considerations for AI-driven poll reporting
Key ethical points that newsrooms should heed include transparency about models and sources used, explicit mention of uncertainty margins and methodological limitations, and mandatory human editorial oversight of automatically generated texts to prevent framing bias [1][2][5]. Regulatory and societal discussions about AI and digital services also stress that technical solutions should not be separate from policy and journalistic codes — especially when deepfakes, fake photos or automated deception are already matters of concern in party manifestos and public debate [3][6].
Practical recommendations for newsrooms (operationalisable)
Journalistic self-restraint can take concrete form by: a) fixed formats and templates that always display uncertainty margins and sample information; b) requiring human final editing for AI-generated poll texts; c) periodically assessing poll methodologies; and d) supplying readable metadata with automated visuals — measures that increase both reliability and public accountability [1][2][5].
Timeliness and context of the observations
The examples and warnings discussed occur in the midst of the current election campaign, shortly before the 29 October 2025 election, where polls visibly colour the coverage and prompt researchers, journalists and editorial teams to reflect on practices and responsibilities [4][1][2].