Original title: The Media Career of Artificial Intelligence in Europe’s Elite Press, 2015–2026
Viktor Suter · September 2026
News outlets talk about artificial intelligence (AI) constantly, but how has that news changed over time? Researchers tracked European newspaper articles about AI from 2015 to mid-2026 to find out. They used a team of large language models to read and sort thousands of articles, checking the computer choices against human judgment. First, news coverage exploded. Coverage first peaked in the late 2010s, but it surged far more after ChatGPT arrived in late 2022, reaching more than seven times its earlier average level by early 2026. Second, the tone turned sour. Net valenceA measure of how positive or negative the articles are overall, calculated by subtracting negative from positive scores., which measures positive versus negative tone, dropped sharply into negative territory across all ten countries studied. Finally, the topics changed. Instead of focusing on science and research, journalists shifted toward conflicts over jobs, copyright disputes, and chatbot harms as AI moved into everyday use.
Imagine you are following a new band from their first small gigs to stadium tours. At first, reviews focus on their talent and potential. Then they release a hit song, and suddenly coverage explodes. But now the stories are about lawsuits over songwriting credits, arguments with fans, and the pressure of fame. In this paper, the band is AI, the hit song is ChatGPT, and the shift in coverage from potential to conflict mirrors how the press changed its story about AI after 2022.
If you read European newspapers, the story about AI changed completely after 2022. Coverage exploded and turned negative, focusing on copyright fights, job losses, and chatbot harms instead of scientific breakthroughs. This matters because media narratives influence how the public and policymakers think about regulating AI.
The abstract doesn't explicitly list limitations, but the study only covers center-left and center-right newspapers in ten European countries, so it does not represent tabloids, far-left or far-right outlets, or non-European media. The analysis ends in mid-2026, so it cannot capture later developments. The coding relies on LLMs, which may have biases not fully corrected by human validation. The valence measure is a simple net score, not a nuanced sentiment analysis. The study does not examine the accuracy or impact of the coverage, only its volume, tone, and topics.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
The Media Career of Artificial Intelligence in Europe’s Elite Press, 2015–2026 — Viktor Suter — SocArXiv — doi:10.31235/osf.io/vz4wj_v1
Abstract licence: CC BY
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.