The Fraught Relationship between Journalism and AI

I’m chairing the next session at this AoIR 2024 conference, which is on the intersections (or collision) between journalism and AI. We start with Sangeet Kumar, who notes the long history of complex interactions between digital media platforms and news publishers; news is just a type of content for platforms, while for news producers it is a mission and vocation. There is a substantial amount of traffic coming from digital search and social media platforms to journalistic sites, and therefore a substantial level of dependency.

This has led to some controversial government interventions like the Australian News Media Bargaining Code or the Canadian bill C-18 to try and address declining revenue streams towards journalistic publishers; this has also led to Meta’s growing disconnect from news content, however, while Google remains a major source of traffic to new sites.

The arrival of generative AI complicates this struggle even further: what happens when people receive AI-generated news summaries of news stories rather than links that point to (and direct traffic to) news sites? Concerns about the impact on traffic and advertising revenue for news sites are increasingly acute; this is also because it is not at all clear what datasets the latest generative AI models are trained on (earlier models were more transparent in this regard).

Companies like OpenAI have also made deals with some news publishers, however – potentially preempting lawsuits for scraping news sites without permission and against the terms set in robots.txt files on Websites. Such deals are unevenly distributed and tend to favour major US and Anglo sources only, however; similarly, current lawsuits about content scraping have mainly been brought by major US and Anglo news publishers.

All of these developments move us towards a closed information ecosystem and increase the asymmetrical relationship between platforms and news publishers, while making information sourcing even less transparent. Entire news passages may be regurgitated by AI tools, without appropriate attribution or acknowledgment. Even if links are include in AI summaries this now no longer directs much traffic to news sites: rather, it just lends an air of authenticity to AI-generated news summaries.

This leads to a less inclusive Web, and to a move by news publishers to opt out of Open AI scraping. This narrowing of the input funnel also reduces the quality of AI-generated content, of course, which further degrades the reliability of AI summaries and affects the overall information environment; AI hallucinations and misrepresentations may also spread misinformation. While OpenAI has a market capitalisation of hundreds of billions of US dollars, more and more journalistic organisations are closing.