How newsrooms around the world use AI: a JournalismAI 2023 global survey

Over the last few months, newsrooms and media organisations from around the world have been completing the second iteration of our JournalismAI survey. We conducted the first one in 2019 and published our findings in a report; New powers, new responsibilities. A global survey of journalism and artificial intelligence. So much has happened in the world of AI since then and we’re excited to share with you some of our preliminary findings in this new and updated research.

This year, we made it a point to reach a more diverse group of participants in terms of size, we invited small and large newsrooms; experience, participants include emerging and legacy organisations; and region; contributions came in from Latin America, Africa, the Middle East and North Africa, Asia Pacific, Europe, and North America. We’re immensely grateful to each and every one of the respondents.

More than 60 newsrooms and media organisations have shared their insights with us so far, with more coming in as we publish this update. The survey includes 35 questions ranging from the technical, to the ethical, the region-specific, and of course, we dedicated a section to generative AI technologies! The survey has been supplemented with great conversations we’ve had with many that have enriched this research.

This article will provide a glimpse of our findings thus far. A more thorough analysis of all the input we receive will follow in the comprehensive report we are publishing this September.

We can tell you that most newsrooms we surveyed have already experimented with generative AI technologies like ChatGPT, but not necessarily to create content. The use cases we’ve learned about are quite diverse: code writing, summaries, enhancing headlines and SEO. One respondent said they were using ChatGPT as a ‘banter buddy’, “Imagine having a trusted companion in ChatGPT, ready to engage in lively banter and brainstorming sessions,” they said.

Most respondents agree that generative AI technologies present a new set of opportunities other AI technologies have not provided, but they are more divided as to whether they also bring a unique set of challenges.

When we asked, “why do you use AI technologies in your newsroom? What do you hope to achieve by using these technologies?” More than half the respondents said they hoped to automate mundane tasks and simplify workflows to free up journalists to engage in “more creative, relevant, and innovative work.”

Limited resources and technical expertise are still the most significant challenges to AI adoption in the newsroom, compared to what respondents said in 2019, but this does not mean hiring more technical people. Many respondents told us that achieving interoperability and synchronisation with other departments was a challenge. It’s also about bridging the knowledge gap between technologists and journalists, in terms of tech skills and journalistic skills as well. This requires a nuanced understanding by the newsroom leadership of the types of training needed for each department. Similarly, the ethical aspect continues to be a central concern. Setting ethical guidelines and debiasing techniques seemed to be the most difficult area for many newsrooms.

For non-English speaking journalists, the challenges are more pronounced. Respondents highlighted language limitations of AI tools in other languages like transcription tools and algorithmic bias is experienced at seemingly higher margins than in English. This has pushed some to develop their own tools in-house, which takes considerable time and resources. One respondent lamented: “Applications used in [news] gathering failed, sometimes while trying to collect Arabic language data, this is why we developed our own internal tool.” In addition, many environments struggle with internet infrastructure and penetration, which makes AI adoption in the newsroom a luxury. Despite those challenges, the enthusiasm –  as well as the scepticism  – for AI in the newsroom remains high!

The full report will be launched at a yet to be confirmed date in September 2023. Sign up to the JournalismAI newsletter to be the first to know when this report will be launched.

This article was written by Mira Yaseen, Lead Researcher of JournalismAI, and Professor Charlie Beckett, who is leading the project.

Previous
Previous

BiasBlocker: We asked a language model to identify racism and it tried to erase baby Hitler

Next
Next

Will there soon be real AI media and AI journalists? Reflecting and acting in response to new technological gaps