AI's news coup: quick fix or flop
In light of the recent content deal between OpenAI and News Corp, we are yet again thrust back into the issue of financing journalism in the digital age. Only a short while ago, this issue reared its head when Meta decided to forgo renewing its deals with news organisations. Will content deals with AI companies be the saviour of news organisations? Some think not. Tim Dunlop writes that the deal with News Corp is “a triumph of short-termism and a failure of media companies to think strategically.”
One argument against these deals is that media companies are holding out in vain for clicks on links back to their websites, where visitors may then take out subscriptions or hang around long enough to generate some ad revenue. The argument follows that as people consume more content through AI, if the AI holds itself out as reputable (due to access to reliable sources or not), how many people will feel the need to verify the information?
It's hard to say. For one, we are not sure exactly how content from news organisations will be presented to end users. The News Corp announcement mentions only that OpenAI ‘has permission’ to display content. Without further detail, these deals could be nothing more than payments to stop news organisations from suing for copyright infringement. At the same time, if (or when) content consumption becomes AI-channel-dominated, how many people will search for news links via traditional or other means? As Michael highlighted in the last edition, Google claims that links in AI Overviews resulted in more clicks compared to traditional search. Yet we’ve heard little detail about how sources in the AI Overviews are returned and, so far, have seen plenty of examples of outright dangerous information being showcased.
Presumably, Google will still be returning links to indexed web content (in addition to paid-for links). What is returned will likely still be determined by the SEO algorithm (what is most relevant to the search or query), but now, Google is actively selecting just a handful of sources to elevate through its AI offerings. It seems unlikely that this will see an uplift in link click-throughs across the entire news industry.
With this in mind, even if potentially detrimental to an open information ecosystem, the safe bet for those offered the opportunity may be to take the money, with the potential that organic growth may decrease, priced in (which may already be the case with these deals). AI companies get access to valuable human-written content, and news organisations get a new revenue source when AI may result in organic traffic decreasing regardless.
But are more ‘strategic’ business models available in the face of AI?
Professor of Journalism at Columbia University Jeff Jarvis suggests news companies could offer an API (a method for AI companies to request and receive content in real-time from news providers). This seems reasonable, but the choice between getting a set amount of dollars versus a potentially uncertain and fluctuating income based on usage over time may dissuade media companies from taking this approach, at least on its own (but we’ll circle back to this).
A more interesting prospect is the idea of news organisations creating their own large language models (LLMs). For example, imagine conversing with the ABC’s NewsAgent, an AI model trained on past content and which has real-time access to up-to-date reporting. This may provide a nice middle ground. The public can still get the flavour of news they are accustomed to and in an AI-packaged format, and news publishers keep traffic on their websites and potential subscriber revenue. As with most public AI models, news organisations could then also introduce paid API access to their LLMs (and underlying content if desired) for other companies, including the AI companies, to build on top of it.
This is not without its own issues, particularly barriers due to financial constraints, access to technical expertise, and a lack of high-quality non-English pre-trained language models. It also doesn’t solve the issue of commercial interests' impacts on the journalism industry's civic role.
As Monica has previously stated, “[f]inding a sustainable model is the holy grail of the information ecosystem – again, and this time, with a renewed sense of urgency!”
Kieran Lindsay, Research Officer
