Putting the AI cart before the news horse

This week saw the release of the annual Digital News Report: Australia (DNR) from the team at the News & Media Research Centre at the University of Canberra. With the report now in its 11th issue, longer term trends are becoming clearer even as they are disrupted by emerging developments.

One of these developments – and a focus of the CMT – is AI. We have been conducting research on the impacts of this technology since early 2023 and will soon be publishing a report on the second phase of our study. While our research focuses on industry perspectives, the DNR provides important insight into the thinking of the Australian public. In our research we have found that newsrooms are particularly concerned about maintaining audience trust; indeed, this appears to be a strong driver of the relatively limited implementation of AI in Australian newsrooms compared to many other parts of the world.

It is therefore interesting to see that public comfort with newsroom AI use is relatively high in Australia, and growing year on year. 21 percent of Australians (compared to 18% globally, and up from 17% in 2024) indicated they were comfortable or very comfortable with news produced mostly by AI with some human oversight. Almost half of Australians (45%, up from 43%) were comfortable with news produced mostly by journalists with some help from AI. Newsrooms may be heartened by this latter figure, as according to our research, assistive uses of AI are the focus of experimentation in newsrooms, certainly far more than AI-produced content.

The young and highly educated are the most comfortable with AI use in news, as are news subscribers. These groups also tend to be the most trusting of and engaged with news and the most politically engaged, but also more likely to be concerned about misinformation. General trust in news in Australia is slightly higher (43%) than the global average (39%). But those countries where trust is highest, such as strong democracies in Europe, tend to be the least comfortable with AI news. This is not a straight correlation, however. A notable exception is South Africa, which has the third highest trust in news of surveyed countries at 55 percent and the second-highest comfort with AI news at 34 percent.

Interestingly, while 47 percent saw AI news as cheaper to make and 31 percent as more up to date than human news, 26 percent saw it as less biased, and 22 percent saw it as more accurate. However, only 17 percent saw it as more trustworthy. As Hal Crawford observes in his commentary on AI in the report, these figures should not necessarily be taken at face value. AI – to the extent that it is not hallucinating – relies entirely on published news for its content. Of course, any one news report may be inaccurate, but the idea that AI output could be generally more accurate (not to say more up to date) than published news is to put the cart before the horse.

The question of bias also needs unpacking. Bias in AI output is a well-known defect of the technology, and recent qualitative research from RMIT and others found AI bias to be a significant concern amongst audiences. It may be that those who perceive AI news as less biased than human-produced news have particularly strong concerns about bias in the latter.

Turning again to our research at CMT, we have found that newsrooms are particularly concerned about the long-term effects of AI on the news industry. This, unlike problems of accuracy and bias, is something over which they have limited control. Although audience use of AI chatbots for news remains very low, the increasing integration of AI into search engines and other platforms may reduce click-through rates to news sites, affecting subscriber and advertising revenue. Perceptions of AI as more accurate and less biased may drive greater use of AI for news, with the ability of these tools to quickly summarise multiple sources a potential boon for those seeking to verify or diversify their news.

It is likely to be some time before these trends are clarified.

References 

Digital News Report (University of Canberra): https://www.canberra.edu.au/research/centres/nmrc/digital-news-report-australia 

RMIT report (Analysis & Policy Observatory): https://apo.org.au/node/329601