Case Study: Sweden’s Aftonbladet Built AI-Driven Editorial Tools and an Election Chatbot


Published October 8, 2024

AI in the Newsroom is a series of case studies from the Online News Association (ONA) that highlight specific ways journalists are building and using AI tools.

In 2023, Aftonbladet, Scandinavia’s largest newspaper, launched an AI Hub to experiment with artificial intelligence. The hub’s most notable project was the development of an election chatbot for the EU elections, designed to engage their audience and answer their questions in real time.

Opportunity

Sweden’s Aftonbladet recognized the growing influence of AI but also saw hesitation in their newsroom regarding its role, particularly in editorial tasks. Many editorial staff were cautious about AI, concerned about how the technology might affect creativity and the core principles of journalism.

To address this skepticism, Aftonbladet set out to create an AI Hub that would serve as a space for experimentation, rapid prototyping, and education that ensures AI adoption benefits both journalists and readers.

Solution

The AI Hub was formed with an eight-person team, including journalists, developers, and a UX designer. The decision to have editorial staff lead the initiative helped build trust within the newsroom and align the tools with journalistic needs. Here are some of the their key initiatives.

Aftonbladet’s AI-powered chatbot “Valkompisen” engaged readers in conversations about the EU elections, answering over 150,000 questions.

AI Literacy: The hub launched a newsroom-wide masterclass called “How to Be a Prompt Queen,” which focused on teaching journalists effective AI prompting skills. The team followed up with workshops and individual guidance, ensuring AI adoption went beyond just the small team of experts.

Editorial Tools Development:

  • AI-generated article summaries were integrated into the CMS. These summaries appeared below article leads and offered expandable content, that has shown a high click-through rate of 43%, and 53% among readers aged 19 to 36. Despite initial resistance from reporters, high engagement rates demonstrated the feature’s value.

  • AI Buffet is a toolkit of editorial tools. The lineup includes Spånaren, which suggests follow-up stories; Buddy Reader, used for proofreading and feedback; SEO Super Optimiser, for evaluating article previews; and Youth Assistant, which generates simplified fact boxes and timelines for younger audiences.

Using AI to Track Diversity: The team used AI to analyze over 100,000 articles, along with thousands of social media posts, videos, and podcasts, revealing a significant gender gap. This led to the development of a custom GPT to help journalists find female experts and a bot to monitor gender balance on the front page.

Building an EU Election Chatbot: For the 2023 EU elections, the AI Hub developed “Valkompisen,” a chatbot designed to answer readers’ questions about the elections. The bot, wearing a virtual EU bandana, responded to over 150,000 questions, with 60% of them being user-generated. Despite the EU elections traditionally being a low-engagement event, the bot fielded 18,000 questions on its first day alone, surpassing all expectations.

Results so far

The AI Hub’s projects have seen considerable success, both in newsroom efficiency and audience interactions. Key results include:

  • Increased audience engagement: Audience logins are a significant challenge for Scandinavian media, where campaigns typically achieve a 5% login rate. The election chatbot helped drive a tenfold increase in logins, as users were required to log in to continue interacting with the bot, making it a key success metric.

  • Positive reader feedback: Despite initial concerns over AI-generated errors, the chatbot maintained high accuracy and neutrality. Feedback, especially from younger audiences, was largely positive.

  • Editorial tool usage: Tools like AI-generated article summaries saw strong engagement—40% of readers who saw the summaries expanded them, and those who used the feature spent more time reading the full articles.

Lessons learned

Aftonbladet’s experience with AI has revealed important lessons about AI’s strengths and limitations within the newsroom:

  • Editorial AI Skepticism: Despite early expectations, tools like AI-generated headlines did not perform as well as anticipated. Journalists often found that human creativity outperformed AI-generated headlines, leading to reluctance in fully adopting AI for editorial tasks.

  • Integration Is Key: Tools integrated into the CMS, like article summaries, were far more successful than those requiring journalists to switch between platforms.

  • AI as a Learning Tool: The AI Hub’s focus on education—through prompting classes and newsroom workshops—helped shift the mindset toward embracing AI responsibly.

  • Shifting Focus to Audience Tools: As the AI Hub evolved, its focus shifted from editorial tools to audience-facing innovations. This included projects like text-to-speech services, podcast translations, and future chatbot initiatives aimed at enhancing the audience experience.

Go deeper

To learn more about AI innovation at Aftonbladet, listen to this Newsroom Robots episode with Martin Schori, Deputy Managing Editor and Associate Publisher of Aftonbladet and the Program Lead for their AI Hub.

We also encourage you to connect with other newsroom AI innovators through the ONA Community Slack.

This resource is part of the AI in the Newsroom series. Read other case studies you might have missed:

Related Topics
Author
Nikita Roy
ICFJ Knight Fellow and Newsroom Robots Podcast Host