In an era of intense competition for reader attention and subscription revenue, Denmark’s media group, JP/Politikens, proved that artificial intelligence can be a powerful partner in journalism. Well ahead of the recent AI boom, the publisher launched a project with its national news brand, Ekstra Bladet, to explore how AI could not only improve the news experience but also strengthen the bottom line. The result was a remarkable 35% increase in subscription conversions, a testament to a thoughtful, editorially-aligned approach to technology.
The project, titled Platform Intelligence in News (PIN), was built on a core vision: to use AI to create a wider, deeper, and richer news experience that would ultimately boost both advertising revenue and subscriptions. The team focused on implementing AI systems that aligned with the newsroom’s editorial values, while also promoting healthy norms for the wider media industry by sharing their experiences.
To create a wider news experience, the team implemented AI-driven recommender systems to help readers discover relevant news. Recognizing that news articles have a short shelf life compared to entertainment content, they engineered a sophisticated architecture using a mix of AI models. This approach helped to surface popular and trending stories while also enabling readers to discover niche content based on their individual reading history. This multi-model approach paid off significantly, leading to a 110% increase in free article consumption and a 38% increase in paid article readership.
For a deeper news experience, the PIN project leveraged Natural Language Processing (NLP) to dynamically surface related stories. The system automatically categorized articles and identified key entities like people and organizations, which made it easier to index content and provide relevant suggestions. This powerful demonstration of NLP’s value resulted in a 120% increase in traffic from related article suggestions, making Ekstra Bladet’s news archives more accessible and improving the longevity of its content.
The final piece of the puzzle was using Natural Language Generation (NLG) to create a richer news experience, particularly for expanding local news coverage. The team developed a toolset called MAGNA, which handled routine tasks like generating summaries. This freed up journalists to focus on in-depth reporting and investigation. A core principle of this work was a “human-in-the-loop” review, where editors were always in control. A key finding was that by grounding generative AI in factual, external knowledge bases, they were able to reduce factual errors from 41% to under 10%.
Ekstra Bladet’s journey revealed several critical lessons for the news industry. First, AI systems are not neutral, and their design choices greatly influence outcomes. A mix of systems is often needed to align with editorial values and serve a diverse readership. Second, generative AI is most reliable when anchored in external, factual knowledge bases, which significantly reduces the risk of hallucinations. By building a comprehensive AI system in-house and openly sharing their learnings, JP/Politikens not only revitalized Ekstra Bladet but also demonstrated a blueprint for a more responsible, reader-centric future for journalism.