Four different responses to public skepticism around AI in news

The value of news is based on its relation with the public

The value of any kind of news is in large part in the eye of the beholder. It's a relationship between what journalists and news publishers do and the people who engage with it, members of the public.

That relationship is generally important, and specifically crucial for journalists and news publishers trying to navigate widespread public skepticism around the use of AI in news.

This relationship has always been important, but will, I expect, grow more important in the context of an ever-more competitive marketplace for attention.

This context is one in which generative AI represents an additional disruption, layered on top of an already radically disruptive move from a relatively low-choice offline media environment dominated by competing legacy media organizations controlling the main channels of communication, to a high-choice environment where many of the commanding heights are occupied by competing technology companies who own and operate the platforms most of us rely on to find, access, and engage with all sorts of content, including news.

News publishers who have lost much of their control over distribution, discovery, data, and advertising already now face the prospect of also losing ground when it comes to content. True, some of what AI can generate at scale – cheaply, near-instantaneously and in personalized fashions – today falls short of the best journalism has to offer. It is entangled in problematic political economics, and some of it is generic low-information generalities, misleading, or outright false and fabricated. But it is also demonstrably already the case that many generative AI users find a fair share of synthetic content useful and valuable. Generative AI use is currently growing roughly twice as fast as internet use did in the 1990s and early 2000s.

In this evolving environment, many kinds of content are likely to grow ever more commoditized. But relations are not necessarily so.

That is why I expect that the relationship between news and journalism on the one hand and the public it aims and claims to serve on the other hand will only grow more important going forward.

That relationship is in many ways challenged by external forces – by AI, by already-dominant platforms, in many countries by political attacks and much more.

But it is also challenged by an inability or unwillingness by the industry to be mindful of and care for relations with the public.

At worst, there is a tendency to take the relationship for granted, assume people will accept journalism as valuable because journalists think it is, and that they will defer to the authority it aspires to. Even absent such outright arrogance, the profession and the industry does not always put in the work to maintain the relationship with the public, understand what trust does or does not look like from different people’s point of view, let alone ensure that journalists and news publishers earn and retain it on the public’s terms.

In the rest of this piece, I will, first, briefly summarize key points from recent research I have been involved in documenting how the rapid development of AI and public perception of AI when it comes to news provides one such external challenge. Then, I will outline four different responses journalists and news media publishers might embrace when it comes to public skepticism around AI in news as part of efforts to be more active stewards of their relationship with the public, before ending with the question ‘what do you do next?’

The public is skeptical of AI in news

Together with colleagues, I have for several years been conducting research on how people use and think about generative AI and news. A few key findings provide important evidence of how skeptical the public is. The results I summarize below are based on online surveys of nationally representative samples of adults conducted in 2025 across a range of mostly high-income countries in Europe, North America, and Asia. They will not be representative of every country or every community, but still capture important aspects of the context many journalists and news publishers navigate. (Those interested in details on the research can read more here.)

First, when asked whether news produced mostly by artificial intelligence with some human oversight is likely to be more or less of various qualities we might associate with news, compared to news produced entirely by human journalists, people tend to be pretty pessimistic. While there are more respondents who say they think such news will be more up to date, easier to understand, and informative than say the opposite, there are significantly more pessimists than optimists when it comes to how transparent and, crucially, trustworthy people think such news will be. The quality most people expect to characterize news mostly produced by artificial intelligence? Cheaper to make. So, in short, less costly to churn out, but less transparent and trustworthy, all in a context where journalists and news publishers already struggle with suspicious minds and low trust or even active distrust.

Second, some journalists and news publishers might protest “we do not produce news mostly by artificial intelligence!” and assume that this public skepticism might provide a dark background against which their human journalism can shine. Perhaps that is so for some of them, and for some parts of the public, but not for most. Because a majority of respondents, when asked how frequently, if at all, they think generative AI is being used by news media, answer “always” or “often”, at a rate behind only search engine and social media companies. So, irrespective of whether or not a given journalist or news publisher is in fact using AI, much of the public is likely to assume they are, and their (fairly skeptical) beliefs about what that will mean for them (less transparent, less trustworthy, cheaper) is likely to be “priced in” whether or not AI is actually used.

Third, maybe a “sinking tide” of trust driven by public skepticism might lower all boats in the marketplace of attention? Declining trust may not be great, but, after all, trust is not only relational, but also relative, so if citizens are growing more and more skeptical about everything in what is sometimes assumed to be a general crisis of trust, perhaps journalism and news media is losing ground “only” in an absolute sense and not in comparison to other organizations? That is unlikely for two reasons. First, when it comes to trust in general, while trust in news media is in overall long-term decline in much of the world, and the same applies for some of the institutions most frequently featured in news coverage, including governments and political parties, this is not a general crisis of trust. Trust in other institutions, including the civil service, the legal system, and for that matter science, has generally held up or even increased in the same period in many countries. (As shown here.) Second, when asked how much better or worse the use of generative AI will make their experience of interacting with various sectors of society, the optimists currently actually outnumber the pessimists for most sectors – except news media, governments, and political parties. The contrast between the preponderance of optimists when it comes to search engine companies and even (!) social media companies versus the preponderance of pessimists when it comes to news media is particularly striking, and particularly worrying for journalists and publishers who care about their relation with the public.

So people are skeptical of what the use of artificial intelligence in news will mean for them, often assume news media are already using generative AI often or even all the time, and are more pessimistic about the implications of AI use in news media for them than they are when it comes to AI use by technology companies. Journalists and news publishers will not like this, but we have to navigate the world as it is, not as we wish it was.

Four different responses to public skepticism around AI in news

So what are they to do? One way to think of the options available here is to imagine two stylized axes, one of which concerns whether a given journalist or news publisher is a less or more intense user of generative AI, and the other of which concerns whether they are leaving their choice of what to do with AI implicit or explicitly telling the public a story about what they are doing and why. These two axes can then help us identify four ideal type responses. No real situation will ever be so clear and neat, but hopefully the typology is useful to think with.

The first is what we might call “resigned journalism”. This involves (a) limited use of generative AI and (b) abstaining from explicit communication about how it is used, if at all, and why. Some news publishers will end up here by default because they are not changing their editorial practices, and communicating about that probably seems superfluous. Others may decide not to change (perhaps hoping for a fast-follower option based on learning from others’ mistakes, and successes) and, perhaps understandably, not interested in drawing attention to their holding pattern. In both cases, the position is complicated by the tendency for people to assume that journalists and news media are already intense users of AI.

The second is what we might think of as “furtive AI journalism”. This involves (a) increasingly intense use of generative AI and (b) abstaining from explicit communication about how it is used and why. Some news publishers will end up here because they may feel that the work should speak for itself, and while they are keen to use generative AI, they are not more keen to commit scarce resources to explicitly communicating how and why they use it than they are to commit scarce resources to explicitly communicating any other part of their editorial practices, workflow, or internal operations. Click deep enough, and you might find something, but it may well be about as detailed and publicized as the organization’s policy on, say, the use of anonymous sources or press releases and other handouts, or the inner operations of their dynamic paywall. Other news publishers may end up here because they want to realize the “cheaper to make”-part of what much of the public thinks generative AI has to offer when it comes to news, but may privately think that the public might be right about the “less transparent”- and “less trustworthy”-bits too, and are therefore reluctant to draw attention to the issue. Given our research findings, in the absence of explicit communication, much of the public will assume news publishers are engaged in furtive AI journalism.

The third then represents what some have called “romantic journalism”. This involves (a) limiting the use of generative AI, at least on the editorial side and (b) explicitly communicating about this decision and what it is based on. An example of this, and explicitly named that way, would be the “romantic journalism manifesto” launched at the 2026 International Journalism Festival in Perugia by Lea Korsgaard from Zetland in Denmark and Joshi Hermann from Mill Media in the UK. It is a nuanced and thoughtful view that I encourage everyone to consider. It is also a clear example of explicit communication to differentiate one’s position from a competitive set many of whom are at least experimenting with generative AI use, also in editorial, and all of whom much of the public, as said, may assume is already using generative AI quite intensely. It is easy to understand the attraction here, and it is effectively the mirror twin of the response above – if much of the public thinks using generative AI will lead to news that is cheaper, less transparent, and less trustworthy, and you fear that they may in many cases be right, it is important to really clearly differentiate what you do from the rest of the profession and industry, to stand apart from being judged by the lowest common denominator.

The fourth and final is what I would call “modernist journalism”, which then involves investing scarce resources in (a) increasingly intense use of generative AI and (b) explicitly communicating about this decision and what it is based on. This would be a modernist view in the sense that the term “modernist” has in some forms of architecture, art, and literature, a broad ethos of “make it new”, a belief that our best selves might be ahead of us and can be enabled by new approaches, new forms, new technologies. Leaving aside some AI-enthusiastic start-ups, it is hard to identify examples of this among established news publishers, though there are clearly journalists and other news media professionals who are effectively pursuing this kind of response. A record number of Pulitzer Prize awardees disclosed AI use in 2026, the New York Times has published a human-computer interaction paper on the development and use of the Epstein Files Engine as an inhouse A.I. agent using Google BigQuery SQL queries across various corpora, Bloomberg’s AI researchers contributed 8 papers to the 2026 Association for Computational Linguistics conference on issues ranging from agentic systems to responsible AI to financial NLP, and journalists at Die Zeit and Der Spiegel have taken time to explain how they used AI to build a searchable database of Nazi Party members. These are not examples only of using AI. They are also steps of explicitly communicating how it was used, what for, and why. While in large part oriented at fellow professionals, they contain at least the kernel of a more explicit communication about what generative AI in news can also mean – beyond making it cheaper to produce.

What do you do next?

So what do you do next? I strive to be impartial as a scientist, and stay clear of professorial preaching. But I also care about journalism and news publishing, and about helping journalists and news publishers think through the context they are in, some of the options they have, and what the implications might be.

Perhaps for some, this piece reads like a call for modernist journalism. It is not meant to be. Each of the positions outlined above could be named differently, resigned journalism might be called “stoic”, furtive AI journalism could be called “stealthy”, romantic journalism “retro”, and modernist journalism “techno-solutionist”. Each comes with risks and possible rewards.

My concrete proposal here is not that journalists and news publishers should or ought to commit to one specific response out of these four. Instead, my concrete proposal is that they decide which one comes closest to what they want to do, and consider whether and how they communicate with members of the public about what they do, how, and why.

Some paths require little change. Others, especially those committed to greater use of generative artificial intelligence and/or more explicit communication require more.

Here it is useful to step away from AI and a changing competitive situation as the proximate cause for caring about the relationship between journalism and news publishers on the one hand and on the other hand the public that their work is supposed to serve, and who provide the basis for everything they are trying to achieve – whether political impact, social significance, or for that matter financial sustainability.

Because anything that involves more explicit communication probably requires a culture shift in how much of journalism is often practiced and how many newsrooms often work. To be blunt, research suggests that, journalists and news publishers still, even after all the profession and the industry has been through over the past few decades, often have a pretty partial picture of how the public sees their work and do not prioritize investing in their relationship with the public as much as they prioritize investing in many other things.

Perhaps this is the cultural heritage of a profession and an industry that came of age in a context where it could be assumed that the work would speak for itself, and that if it did not, ultimately it did not matter very much as members of the public did not have that many alternatives to journalism and news media to choose from. But that was then, this is now. As I see it, empirical research suggests that journalism today does not speak for itself, and reminds us that people have an abundance of alternatives.

However one chooses to approach this situation, and however one chooses to communicate about it, generative AI is yet another radical disruption of a profession and an industry that has taken more than its fair share of knocks in recent years. That disruption will no doubt be painful for many journalists and news publishers, even as some also see various opportunities in it.

What you do with AI in this disruption is the immediate question. The wider question is one that transcends AI and centers journalism and news publishers’ relation with the parts of the public they want to engage with and asks who you are doing it for, how they see it, and whether you can explain that it is valuable not just for you, but also, ultimately more importantly, for them.

This is not (only) about whether or not you use AI, for what, or how much. It is about much more basic questions of ensuring you are clear about what you do, who you do it for, why, and that what you do and what you say you do is aligned, explainable, and intelligible to those parts of the public you want to engage with your work. Communicating your choices about AI is but one component of this; a small part of trying to invest more care and effort into the relationship between news and the public. A simple starting point would be to ask yourself three basic questions: What does the part of the public we serve assume we do with AI? What do we do to explain what we do with AI? Would people see the principles we preach and what we actually practice as aligned - or not - if they knew precisely what we do with AI?

Truly AI re-wiring journalism and news publishers, for those who wish to, will take years, if not decades.

But the wider work on journalism and news publishers’ relation with the public is not only or even primarily about AI. It is cultural, in a way more foundational. It could take longer. It could also begin tomorrow.