Our podcast: Digital News Report 2025. Episode 2: AI and personalised news

From article summaries and translations to recommendations and text-to-video, we look at what audiences think of various uses of AI in news

In this episode of our Digital News Report 2025 series we look at how audiences think about the various ways that AI is being rolled out across newsrooms. We look at various uses of AI in news, including summaries, translations and customised homepages, and what newsrooms plan to put more resources into. We gauge comfort levels among audiences and explore the reasons behind any differences across countries and age groups. We also discuss the potential for AI to engage those who are avoiding news.

Spotify | Apple

Speakers

Amy Ross Arguedas is a Postdoctoral Researcher Fellow at the Reuters Institute for the Study of Journalism and works on the Digital News Project including as a co-author of the Digital News Report.

Host Mitali Mukherjee is the Director of the Reuters Institute and is a political economy journalist with more than two decades of experience in TV, print and digital journalism.

Transcript

Comfort with online personalisation | Regional differences in comfort with personalisationAppetite for news personalisation | How AI is being used in newsrooms | Comfort with AI-driven news personalisation | Differences by country | Differences by age | Using AI to re-engage audiences | Takeaways for news publishers

Comfort with online personalisation 

Mitali: Can you first explain Amy, on a comfort level, how comfortable people are with online personalisation in general. I mean, are there specific areas where they're more accepting of that personalisation? 

Amy: Yeah. So to help contextualise people's comfort with personalisation when it comes to news, specifically, we first asked a broader question looking at how comfortable people are with content being automatically selected for them or personalised for them across different types of websites or apps. And what we find here is that close to half of our respondents say that they're comfortable with news personalisation, but that comfort tends to be lower than what we see in other domains. So for example, people are most comfortable, and this is 63% of our respondents with personalised selection when it comes to weather. And of course, we tend to want to know the weather of places where we are or where we're going to be, rather than, you know, randomly.

Majorities are also comfortable with the automated selection of things like music and online television and movies. And of course, a lot of us are used to using Netflix and Spotify and having things recommended to us based on our preferences. And people see real benefits of having genres that they like recommended, you know, rather than having to sift through a whole catalogue of music, for example, or television.

Comfort is relatively lower, so just under half, when it comes to news. And, you know, stories of the day can be about almost any topic. And this is, you know, one aspect that makes news personalisation kind of different in a way. And actually, comfort is lowest when it comes to social media and video feeds, so things like YouTube and Tiktok. And it's possible that this is shaped by things like negative experiences that people have had with recommendations on these platforms, or, you know, just having encountered public debate about these kinds of issues.

Regional differences in comfort with personalisation 

Mitali: What makes this data really rich Amy, is the fact that we have such a large sort of international canvas, if you will, to compare and contrast. And it was quite fascinating to see variances both between countries, in fact, I should say, sort of continents almost, and ages as well. Unpack that for us a little.

Amy: Yeah, that's right. So in terms of countries, what we see is that across all of these areas that we asked about, comfort tends to be much lower in parts of Europe, and this is particularly Western and Northern Europe, whereas we see that comfort tends to be much higher in much of Latin America, Africa and Asia. So kind of regional differences like you were alluding to, we also see age differences for some of these domains, but not for all of them. So we don't actually see age differences when it comes to news, but we do see age differences when it comes to the personalisation of music, online, TV and movies, and the biggest gap we see is when it comes to social media and video networks, and in all of these cases, it's younger people saying that they feel more comfortable than older people. And you know, when we're looking at social media and video networks, over half of the people under 35 say that they're comfortable, compared to only 38% among those who are 35 and older. So of course, we know that younger people tend to be heavier users of platforms like Tiktok, where algorithmic recommendation is really central or integral to the user experience. So perhaps it’s kind of unsurprising that we find this greater openness to this kind of recommendation.

Appetite for news personalisation 

Mitali: Let's stick with personalised news selection for a little bit. Amy, across the regions, did you see similarities in terms of what people see as the main benefits and the main concerns, I suppose, around personalisation of news.

Amy: Yeah, we wanted to get a better sense of why it is that people say that they feel comfortable or uncomfortable with personalised news selection specifically so kind of the beliefs or the rationales that are underpinning these attitudes. So we included a follow up question in our survey in many of the countries, asking people to tell us a bit more about why they felt this way. So on the comfortable side, people typically brought up four key benefits that they tended to see. So the first of them was this idea of you know news personalisation, just making news, bringing more relevant news to them, so kind of helping them out in that way.

Another subset of users talked about personalisation, helping them save time and just kind of avoiding the effort again of having to sift through a broader universe of news stories, another subset of respondents expressed kind of a greater sense of trust in news selection performed by algorithms compared to humans, thinking essentially that it's less biased than humans. And there was an interesting quote in there from a person you know, saying, you know these technologies are programmed to make selections based on data rather than personal opinions or preferences. And so feeling like that was a benefit. And then lastly, there was a small number of people who felt like, you know, depending or relying on algorithms ultimately brought them a greater diversity of topics or a greater diversity of viewpoints than they would otherwise come across in their everyday use.

Now, in all of these cases, people are telling us that they think that these technologies work well, and that as a result, that they benefit from them. When it comes to why people feel uncomfortable, on the other hand, we see kind of a broader range of rationales here. So we see some people telling us that they think that these technologies are basically bad at what they do, you know, so not delivering what they want. Or on the flip side, we see some people who are uncomfortable precisely because they think that they're effective but that they can lead to negative outcomes.

And then there's also another group of people that raise concerns that kind of go beyond the news content itself. So I'll say a little bit more about these. So first of all, some people felt like, you know, personalised selection of news would essentially bring them, you know, irrelevant or low quality news content. And some people would kind of raise examples here based on their experience of social media, maybe being served kind of clickbait type headlines and things like that, and that kind of feeding into this perception. Another group of people was concerned about just missing out so kind of a FOMO of, you know, overlooking or developing blind spots, even on important topics or stories if you're only being served things, they're kind of based on what you've consumed in the past. Other people, contrary to what the other group thought, felt that this news would be kind of more biased or worse yet manipulated to make them think in a certain kind of way.

And then there was this last set of concerns that, as I mentioned, went beyond the news itself. And so this was people who were essentially concerned about their privacy and feeling like personalisation technologies are invasive or kind of creepy in some way. Quite a number of responses mentioned ‘Big Brother’ here. And then there was also a group of people who really foregrounded this idea of, you know, self determination, of not wanting to lose control over the news that they were choosing to see.

How AI is being used in newsrooms 

Mitali: I know there's lots more to sort of delve into with audiences and AI personalisation, but let's step to the news side of it for a little bit. You know, one of the things we produce, also at the Reuters Institute, is [our] Trends and Predictions [report], which is how news organisations are thinking about priorities. And for two years running, now, AI has really been on top of that list, along with, you know, engaging better with young audiences in terms of news outlets rolling out AI initiatives, can you point us to sort of key examples or interesting examples of what's happening there across the industry?

Amy: I think here, it might be helpful to differentiate between different kinds of personalisation. So you know, so far, we've been talking about personalised selection, so kind of the actual stories that are being chosen and served to people, which, of course, isn't fundamentally new to us. You know, even though AI can be used to kind of enhance or extend personalised selection, generative AI now makes it technically possible to personalise news formats and to do so at scale. So not the story itself, but basically how it's being presented to people based on their personal needs or personal preferences, and we see some news organisations experimenting with these kinds of things already. So for example, the BBC has been testing an OpenAI tool that's called Whisper that can transform audio into text, and so this can add basically subtitles and transcripts. And they've been doing this to some of the audio pieces published on BBC Sounds.

Other news outlets have been trying out tools that do the opposite. So they allow people to turn text into audio using an AI generated voice. Another example that we see are AI generated summaries, which have also been rolled out by a number of newsrooms, including, for example, Swedish news newspaper, Aftonbladet, excuse my pronunciation. But these are, you know, essentially quick versions of stories that people can generate at the top of full articles. And it's kind of a synthesised version if they don't want to read the full text. And I'll also mentioned another really interesting example that's being tested by the Argentinian newspaper Clarin, which has a tool that offers a range of additional analysis that people can have the tool produce for it, so it has summary bullet points, but it also can do other things like pull out important quotes or key figures from the piece. And it also provides a glossary of like technical terms, for example, and allows people to even produce, like a Frequently Asked Questions section and kind of the articles presented through that lens.

So some interesting experiments are happening, and also some other publishers have created entirely new products. The Independent here in the UK has launched a new digital service that's called Bulletin, which uses Google AI tools to to kind of create article summaries overseen by journalists. It's kind of a separate, freestanding product that's being advertised, essentially as a news product for busy people. So all of it in there is kind of summarised versions of articles. And then lastly, there's others like, you know, the Washington Post or the Financial Times that have been trying out AI tools that can answer questions based on their corpus of articles. So instead of modifying the story, kind of the format of the story here, these are more advanced search functions that can basically answer complex queries based, again, on the news corpus.

Comfort with AI-driven news personalisation 

Mitali: So let's fold that into the chapter that you've worked on for the Digital News Report. Amy, what did you find in terms of overall news appetite for this AI-driven news personalisation. And you know, did you sort of come across use cases that people felt more enthused about based on all the examples that you just pointed to?

Amy: Yeah, so in our survey, we asked people about eight specific AI personalisation options across these different categories that we've been talking about, so some that involve, you know, personalised selection of stories, other others that allow kind of personalising formats, and then also an AI chat bot option. And what we see here is that interest in any individual option is relatively low right now, so under 30% of respondents say that they're interested in any single one of these which, to be fair, might be shaped somewhat by kind of a low familiarity with these tools, but in terms of kind of how they rank.  

We find more of an appetite when it comes to options that make news consumption essentially more efficient to consume or more relevant for users. So article summaries, for example, are at the top of the list, with 27% of respondents saying that they're interested. And these are followed in second place by article translation, where we have around 24% interest. And then in third and fourth place, we have AI-driven recommendations or alerts and customised news homepages, and these are personalisation selection options where we have around one in five of our survey respondents saying that they're interested. On the flip side, we found interest to be lowest for the options that change the modality of news. So for example, these text-to-audio or these video-to-text options, those were the ones that ranked lowest among the options that we offered people.

Mitali: That's interesting. And if we were to do a match the following exercise, you know, how do these levels of interest compare with what we have available in terms of what the news industry is currently focusing on?

Amy: Yeah, that's an interesting point, because what we see among industry leaders in a survey that we published earlier this year is that the text-to-audio option, which is close to the bottom of the audience list, is actually at the top of the list of AI initiatives that newsrooms said that they were planning for 2025 and of course, there are many considerations driving these decisions, and you know that the text-to-audio option is probably seen as something that's pretty simple and cheap and uncontroversial to implement, but there is also this kind of mismatch in some of these categories.

Differences by country 

Mitali Mukherjee: We talked a little bit at the start about sort of regional variations in terms of appetite for this. How does audience support for different formats vary across countries, because we do have examples in Asia, for example, of, you know, an AI-generated news anchor, and, you know, a lot more sort of experimentation on that front. 

Amy: We do see some pretty interesting cross national differences in the relative interest that people have for different options. And broadly speaking, article summarisation tends to be high across the board in most places. But for example, the translation option is at the top of the list in some of the linguistically unique European countries that have smaller populations, so places like Finland and Hungary, which suggests that people in some of these countries have a stronger interest in being able to kind of expand the universe of news that they can consume beyond their language.

Another one of the options that we provided was the use of AI to adapt news to different reading levels. And this can be done, for example, by simplifying vocabulary or sentence structure. So this kind of thing, and what we found is that this option tends to rank higher in countries that have lower literacy rates or lower reading proficiency levels. So for example, in places like India, Kenya, Nigeria, this was actually one of the top three options, and in India, it was the most popular option of the eight. So those were some of the interesting differences.

And then, more broadly speaking, we tend to see kind of more enthusiasm across the board, in countries where comfort with the use of AI in journalism is higher. So again, places like India and Thailand, whereas we see much lower interest in countries where AI comfort is lower, such as the UK. So actually here in the UK, this is where we see the percentage of people who say that they're interested in none of these options is the highest of the 48 markets in the survey.

Differences by age 

Mitali: Let's talk about younger audiences for a little bit. We're referring to digital natives, online natives, a community that's quite used to accessing online news, getting information through automation. Is there a perceptible difference in terms of how they see AI driven news formats compared to older generations?

Amy: Yes, that's right. We do see age differences here as well and younger groups, these are people who tend to be more comfortable with AI. They often use ChatGPT in their everyday life more than older groups, they tend to show more interest in the use of AI, specifically for the personalisation of formats, as well as news chatbots relative to older people. So you know, news organisations may well want to focus some of these kinds of innovation on younger audiences.

Using AI to re-engage audiences 

Mitali: A running thread in the Digital News Report Amy and a lot of our findings has been disengagement amongst the public towards news, a lack of interest, turning away from it. What do your findings suggest about how AI personalisation can actually be used in order to re-engage these these news avoiders.

Amy: Yeah, and this is an important point, especially since many newsrooms have been thinking about AI personalisation as a tool to tackle some of these challenges, however, our data suggests that interest in AI personalisation actually tends to be considerably lower among people who are less interested in news, and also lower among those who tend to avoid news more frequently. That said, there are potentially certain pockets of news avoiders for whom things may look a bit different. So for example, there's a subset of people who avoid news because they find it hard to understand. And these people tend to express higher interest across the board, relative to the general group of news avoiders. And we see the largest gap when it comes to the use of AI to adapt news for different reading levels. So this is one group that might potentially be interested in seeing the use of AI to modify the language of news stories. So I think that there are potentially subsets here, basically that they're going to be more interested in some of these applications, and that are, you know, particularly relevant to to the reasons why they're avoiding news.

Takeaways for news publishers 

Mitali: So many of your examples, Amy, point to the fact that there is no one size fits all in terms of countries or cultures, and you know what works and what doesn't, and but is there sort of a broad takeaway on what news publishers make of this? Because at this point it looks like there's very high enthusiasm and interest amongst news organisations, and it's not quite matched by what news audiences have said they feel when it comes to AI,

Amy: I think it's important to make a caveat here that you know the reported interest in our survey, which is what we're capturing with these questions, it doesn't necessarily mean people will either use the options in practice, just as lack of interest doesn't mean that they're not going to use these options. So, you know, it is possible that some of the respondents didn't really understand what this looks like in practice. What will this mean for, you know, their consumption.

But I do think AI has become such a hot topic in the news industry, and I do think that there is a real risk, you know, of overestimating how excited or how open the public is to some of these public facing applications, and the kinds of tools that they're most interested in, as we talked about a few minutes ago. You know, I think that there is this, this mismatch around the industry focus and audience interest. And more generally speaking, I think it's important to keep in mind that audiences tend to be kind of more skeptical of the personalisation of news than they are in other areas of digital life.

Another point that I would make is that I think it's likely that we'll see this play out quite differently across countries, and we saw some country specific interest in particular applications, for example. But beyond this, we would expect more enthusiasm, and in markets like Thailand and India, where attitudes towards the use of AI tend to be much more favorable than the Northern and Western European countries, you know, where our audiences tend to be more skeptical. So it's going to be important for news organisations to really think about the strategies and also the possible trade-offs of rolling out some of these tools for audiences that have an appetite for them, like younger people, without scaring away more hesitant users.

And the last point that I would emphasise is that AI can power really different kinds of personalisation, and so it's really important for news organisations to be mindful in how they're communicating the use of these technologies to audiences, particularly in light of the concerns we heard from people around personalised selection. So for example, you'll remember that some people expressed concern about personalisation, leading them to miss out on important stories of the day or develop blind spots. But news personalisation can be, and it often is in the context of news websites, designed to retain a common core of important stories, and then personalisation happens around that. And these are, you know, important design decisions that I think are often thought about very carefully in newsrooms.

And so I think clear communication can really help offer audiences some reassurance. And then lastly, I think it's also worth keeping in mind how important the idea of self determination is to some people, we saw it in these open responses. And so I think that offering audiences the ability to exercise a degree of control, at least over personalisation, it might also help placate some of these concerns, even though it's likely that, in practice, they're only going to be taken up by a small number of people, but having the sense of control, I think it might help as well.


Listen to the whole series

Join our free newsletter on the future of journalism

In every email we send you'll find original reporting, evidence-based insights, online seminars and readings curated from 100s of sources - all in 5 minutes.

  • Twice a week
  • More than 20,000 people receive it
  • Unsubscribe any time

signup block