Is generative AI a game-changer for journalism?

I’m a bit more sceptical than six months ago. It’s a huge transformation in terms of the potential for automation and better personalization. But where it moves the dial and how quickly I’m not sure. In the short term it won’t change that much.  

What made you more sceptical? 

Experience tells me: it’s one of those things that at first looks like magic. But when you really think about what journalism does, generative AI will only do part of the job. When you look at what it can do now, it’s great at speed and scale, aggregating things, summarizing things, but a lot of the output is mediocre. Journalism aims to be better than that and connect people as well as inform them. 

You interact with media leaders a lot. What do you sense as the general mood?

It depends on who you talk to. Most journalists are not that engaged but, on the business and innovation sides, people are hugely engaged. Many hope for more efficiency. But they are also worried about what this is going to mean for jobs. On the business side some look at this as an existential threat. Will it further destroy business models? Then there are disinformation concerns. 

You tend to be fascinated by new tech. What fascinates you most with generative AI?

It offers the opportunity to make journalism more relevant. How can we use AI tools to battle news avoidance? For example, to create content that will engage young people by turning a text story into a video story. Then again, we’ll see a lot more of everything, lots more content, more marketing messages too. In that sense an overload of content could make news avoidance a lot worse because it will be even harder to find journalism that is relevant. My other concern is that the promised efficiency gains may not be that great. And publishers might not invest the savings in journalism but just take the money.

What has surprised you most?

The speed. The change from AI-created photographs that looked a bit weird into photorealistic images. That happened within a year. This is unbelievably exciting but also worrying. Anyone can now use AI to create/fabricate a picture of a child sitting amidst the rubble in Gaza with a sad look in their eyes, and it’s hard to distinguish from the real thing.

Some say journalism could evolve from a push activity to a pull activity. What do you think?

People have talked about this for a long time: Oh, with AI you could change the ending of a Netflix drama to your liking. In reality, high-quality linear drama has become even more important in a world that becomes fragmented and confusing. So, the outcomes here are not binary and there’ll still be both push and pull in news. 

What about custom-made journalism?

We clearly haven’t made customization in news work. Most people don’t want the news agenda to be customized, because they ‘want to know what they don’t know’. But the personalization of formats is a different matter and more promising, but how this will happen is not obvious because the user interfaces are complicated. Any friction makes it harder. 

We see an ever-increasing dependency on tech. Will there be anything in this for publishers?

We would hope that publishers have learned something from previous changes and especially the importance of the interface, which is likely to become an even bigger factor with AI. Publishers should have learned how to develop direct relationships with their customers including building great interfaces and platforms of their own. There will also be a significant amount of money from licensing in the new world of AI, but who will get it? Probably the big companies, including news agencies. Probably not smaller or local players. 

You are the lead author of the Digital News Report, the world’s largest ongoing digital survey about news consumption. Will you do research around AI in this year’s edition?

We’ve conducted qualitative and quantitative research this year which uncovers high levels of public scepticism about AI and the news – much more in Europe compared with the United States. But we also show that the public is much more comfortable with back-end tasks where journalists remain in control. There’s much less comfort in general with use of AI around politics or other hard news topics compared with entertainment or sports – and especially where there aren’t enough human checks. Interestingly, transparency is important but the public does not want AI labels everywhere, only when it is materially important.

Your audience research very much centres around trust. Will AI destroy even more of it?

In our recent Trends and Predictions Report, 70% of news executives said AI will most likely lower trust. There’s a lot of media coverage of deep fakes, for example around Joe Biden or Taylor Swift. That makes people more sceptical and more worried. They might develop the perception that you cannot trust anything. But conversely a flood of unreliable synthetic content may make people want to seek out someone they can trust. Trust in some brands might increase. 

You’re talking about the Covid effect, when trust in media spiked?

There are lots of unknowns here, but it is possible. Much will depend on what the platforms will be doing. It is actually in their interest to promote trustworthy content and keep their platforms as clean as possible during this shift. 

What is the particular task for public service media? They have a big reach and enjoy trust. Will they be winners of all this?

Their big challenge is how to get young audiences consuming them with visibility and attribution, given their preference for platforms. This could be done with some kind of regulation, for example to prioritize PSB [public service broadcasting] content on certain platforms. They could be the big winners – or big losers if their content is harder to find or AI further flattens content. 

What is missing from current conversations?

The audience’s perspective on this. So far, we’ve mainly had a debate about technology or about the business perspective, about licensing. We also need a longer-term view. Currently, everyone is experimenting, but we need to figure out what we really want from AI strategically. Public service media should really be at the heart of these debates because they are looking to act in the interest of all audiences, not just the privileged few.