We use cookies to further personalise and enhance the user experience, conduct analytical research (for example, counting visits and traffic sources), place advertisements and contact third parties. Users can manage their cookie settings by clicking the "Choose your preferences" link.

Cookie policy
Published 2023-06-09

Seven AI hacks that could improve Aftonbladet

A fact checker that tells you what is true and not in tv-debates, an AI-generated podcast featuring the most popular articles and the Aftonbladet AI chatbot. These are only three out of seven ideas from an AI hackathon arranged by Aftonbladet.

“It was amazing to see how many innovative ideas a group of people with different competencies and backgrounds can come up with and make a first demo of in just two days,” says John Collins, Engineering Director at Aftonbladet.

Understanding how to use AI in the newsroom is a question the media is now working with all over the world, following the fast developments going on.

“AI will give us many options – but also put us at risk. Experimenting is a good way of learning,” says Moa Gårdh, Product Director at Aftonbladet.

“There are many initiatives going on across News Media in Schibsted – hackathons are, and will be, an important way to get inspiration and insights,” she explains.

More than 30 people participated in the hackathon – coders, journalists, UX designers, analytics-, product- and business developers from the newsroom at Aftonbladet worked together in teams. At this stage, the hacks are just prototypes and we will continue to explore and experiment with how we can further develop them.

Most innovative and best collaboration

The Aftonbladet fact checker was chosen as the most innovative and useful hack. The team identified a recurring problem in televised debates – fact-checking politicians’ statements. The team took on this challenge by building a tool that transcribes audio from a debate, isolates statements that could be fact-checked, and uses real-time information from the internet to verify these statements.


“The pursuit of truth is a journalist’s most important task. Now that we have seen lies become increasingly efficient political weapons, we want to seek help in new technology, and we have actually created a tool with tremendous potential,” says Jakob Andersson, Live TV editor and a member of the winning team.

The AI-generated daily podcast

An AI-generated podcast featuring the most popular articles of the day on Aftonbladet was awarded the best cross-team collaboration. Since users today want more options on how to consume content and it’s expensive to produce content in multiple formats, this is an interesting idea.

The Aftonbladet AI chat bot

One of the teams built a chatbot, by combining GPT-4, LangChain and Aftonbladet’s content API. This would enable end users to chat and ask questions about a chosen news story with the bot.

Better support on Slack

A bot that could make Slack support channels more efficient would be helpful internally. The team built a bot that uses data from an existing Slack channel to be able to answer future questions. The bot can learn new facts on the fly, meaning it always keeps improving with new information!

Open source translator

Another idea was to use an open-source translation network to empower users who don’t speak Swedish, to translate articles without relying on Google Translate or similar third-party resources.

Article image generation

By using the power of image generation models like DALL-E and Midjourney, costs for stock photos could be reduced. One of the teams showed a concept where an article is first segmented into topics, and a suitable article image is generated. The team also proposed a watermarking solution to point out to end users when an article is illustrated with an AI-generated illustration.

Evolving article summaries

Several brands in Schibsted including Aftonbladet have recently launched an article summary feature powered by GPT-4. One hack suggested allowing editors to choose from different “tones” of summaries that use different prompts to tweak the response. Some of the proposed tones were: Serious, playful, promotional-friendly and (jokingly) emoji-only.