Artificial intelligence (AI) is transforming the media industry, establishing itself as a critical tool this year. But what exactly are media houses using AI for? In this blog, we explore diverse AI applications from around the globe to inspire innovation in journalism.
Automating Back-End Tasks
AI is increasingly used to streamline everyday tasks, such as idea generation, translation, graphic design, and proofreading. While these tasks are manageable for humans, they take up valuable time that could be better spent on more complex work. According to a Reuters survey, Editors-in-Chief identified automation of these tasks as one of the most beneficial uses of AI.
For example, Le Monde uses AI to translate its articles for its English-language edition, expanding its reach to a global audience. The AI has been trained to follow Le Monde’s stylistic nuances and the initial translation by AI goes through two rounds of proofreading to ensure quality. This initiative has contributed to a 10% increase in overall subscriptions.
Similarly, Helsingin Sanomat leverages AI-driven tools like "Hennibot" to enhance content, craft engaging titles, and structure articles based on performance metrics like reading rates, read times, and shares.
Article Summaries
The NRK (Norwegian Broadcasting Corporation) has been using AI-generated article summaries at the top of their articles to help readers decide if they wish to read the full article or not. The initiative's purpose is to make the articles more accessible and to attract younger audiences. The initiative has shown positive results, with 19% of readers expanding and reading the summary. The readers who used the summary also spent twice the amount of time on the article compared to the people who did not. Younger readers were also more likely to use the feature.
Screenshot from nrk.no. Translated with Google Translate.
In Argentina, Clarín has developed an AI tool to create article summaries, offer alternative article formats for readers, and find additional context on the topic of the article. UalterAI offers six formats to enhance news readability, including a concise summary, chronological ordering, key highlights, a data table, an alphabetical index of names, and an FAQ format. Approximately 30% of readers use UalterAI, with summaries and data tables being the most popular formats.
Screenshots from clarin.com. Translated with Google Translate.
Content Creation
While some media outlets like CNET, G/O Media, and Sports Illustrated have faced backlash for publishing AI-generated content riddled with inaccuracies and plagiarism, others have found success by balancing AI with human oversight. For example, a German publishing company Kölner Stadt-Anzeiger Medien, uses a virtual journalist to produce 11% of its stories, with human journalists deciding which stories the AI covers, and proofreading all AI-generated content. The AI journalist, called Klara, is responsible for 8-12% of all traffic. The AI also curates and selects the headlines and links that the readers see on the site, which has led to a major boost in CTR (click-through rate)(50-80% increase). It’s clear that while AI can assist in content creation, thorough fact-checking and editing remain essential to ensure quality and reliability.
Screenshot from express.de. Translated with Google Translate.
Chatbots
Innovative media organisations are also using AI-driven chatbots to engage audiences in unique ways. A Paraguayan newsroom, El Surtidor, built a chatbot, called Eva, to help tell the stories of women affected by drug trafficking. The chatbot was built based on years’ worth of reporting done by Juliana Quintana. Users can choose what questions to ask, leading the conversation down different paths to reveal different parts of the story. This immersive experience aims to increase awareness of drug trafficking’s impact on women. The chatbot was launched in September 2024, and by November 2024, over 10,000 interactions with Eva had been recorded.
A journalist in London built a chatbot that helps rework news articles into scripts for viral videos. After years of experience in social media, journalist Sophia Smith Galer trained her own chatbot to rewrite content into scripts for viral short-form videos. Journalists whom she has worked with in the past have often told her that they neither had the time nor the money to learn how to write scripts for social media. Smith Galer developed this chatbot in response, aiming to support journalists in their work while safeguarding her own content from being used to train other AI tools.
Image courtesy of Sophia Smith Galer. Source.
Newsletter Personalisation
JAMES is an AI-tool used by various media houses, including the Finnish Keskisuomalainen, to personalise newsletters for each reader based on their reading history and interests. The newsletters include only articles the reader has not yet read. Keskisuomalainen saw an increase of 4.5% in click rates after adopting JAMES, and other clients have seen even better results. After implementing JAMES, media houses have also seen improvements in habitualisation rates, re-activation of "zombie" subscribers, and lower subscriber churn rates compared to control groups.
Using AI to Advance Watchdog Journalism
A Filipino journalist, Jaemark Tordecilla, created a custom GPT that scans audit reports and flags issues related to government spending of public funds. This tool helps journalists by quickly identifying documents that merit closer examination, significantly reducing the time spent sifting through extensive paperwork. Anyone with a paid ChatGPT subscription can train their own custom GPT to perform various tasks. Even if this particular use case isn’t directly relevant to you, Tordecilla’s experiment offers valuable insights for training your own GPT model.
In another case, AI was used to sift through an enormous collection of technical documents (200,000 in total) related to the Mauritius Leaks, identifying patterns and highlighting documents of interest to investigators. By employing AI, processing the leaked documents became not only more efficient but also feasible—something that would have taken a team of dedicated reporters years to accomplish manually.
These examples show that, when properly trained, AI can effectively comb through vast datasets to spot patterns and anomalies, providing invaluable support to investigative reporters tasked with analyzing extensive documents. However, while AI serves as a powerful tool, human oversight remains essential to catch any potential errors or inaccuracies.
Aggregating Public Data
Realtime is a site that uses AI to aggregate information from public data sources, detecting trends and shifts in predictions, and presenting them in concise “data stories” with supporting graphs. Subscribers can sign up for the newsletter to receive regular updates.
AI can be very helpful in aggregating large amounts of information from different sources; however, it’s crucial to verify those sources and ensure the program is not fabricating information.
Screenshot from realtime.org
Image Recognition Assists in Collecting Evidence for Investigative Journalism
Image recognition technology has been effectively utilised by The Wall Street Journal and The New York Times to support investigative journalism. The Wall Street Journal employed it to locate hazardous lead cabling across the United States, while The New York Times used satellite imagery and AI to track the number of 2,000-pound bombs dropped in Gaza in areas designated as safe for civilians. Both publications ensured the accuracy of their findings through rigorous human oversight.
AI Use at Radio Stations
Swiss and Czech radio stations have been experimenting with AI in content creation and audio production. The Swiss station Couleur 3 replaced human hosts with AI clones of their voices, played AI-generated music, and used AI-generated scripts. This trial received mixed feedback, leading the broadcaster to conclude that human hosts are still superior. In the Czech Republic, a radio station launched a podcast featuring short stories written by AI, which also received mixed reviews but demonstrated AI’s potential to assist in creating diverse content.
In Slovakia, Radio Express uses a cloned version of their popular presenter’s voice to cover the night shift, with the AI providing commentary on music and news stories sourced from the station’s website.
These experiments underscore AI’s evolving role in content production while revealing its limitations in delivering the depth and relatability that audiences often expect from human presenters.
AI-Generated News Presenters
Grupo Formula, a major Mexican broadcast group, has introduced AI-generated human avatars, NAT and SOFI, to deliver soft news tailored for younger audiences. The experiment has seen some success, with NAT gaining 13,000 followers on Instagram and SOFI producing several viral videos. These avatars operate with human support, as news stories are written, overseen, and refined in post-production by human staff. Focused on delivering “soft news” and lighter stories, NAT and SOFI showcase AI's potential to engage younger viewers, though they rely on human oversight to maintain content quality and relevance.
Screenshot from NAT's Instagram @nat_tvoai
Conclusion
AI holds significant promise as an assistant to enhance journalistic processes, streamline research, and improve content management. It can help with tasks such as analysing vast amounts of data, rewriting articles, and even generating content. However, AI is most effective when used as a tool with human oversight. While AI-driven solutions can greatly aid media professionals, they cannot yet replace human creativity, critical thinking, and ethical judgment. For now, AI should support human efforts, with humans playing a key role in ensuring that content remains accurate, trustworthy, and engaging.
You May Also Like
These Related Stories