AI A

AI AND THE FUTURE OF NEWS

MAY 2024

What Does the Public in Six Countries

Think of Generative AI in News?
______________

Richard Fletcher and Rasmus Kleis Nielsen

Contents
About the Authors 2
Acknowledgements 2

Executive Summary and Key Findings 3

Introduction 6
Methodology 8
1. Public Awareness and Use of Generative AI 10
2. Expectations for Generative AI’s Impact
on News and Beyond 17
3. How People Think Generative AI Is
Being Used by Journalists Right Now 25
4. What Does the Public Think About How
Journalists Should Use Generative AI? 28
Conclusion 34

References 37

Report published by the Reuters Institute for the

Study of Journalism (2024) as part of our work
on AI and the Future of News, supported by seed
funding from Reuters News and made possible
by core funding from the Thomson Reuters
Foundation.

DOI: 10.60625/risj-4zb8-cg87

1
 THE REUTERS INSTITUTE FOR THE STUDY OF JOURNALISM

About the Authors

Dr Richard Fletcher is Director of Research at the Reuters Institute for the Study of
Journalism. He is primarily interested in global trends in digital news consumption, the use
of social media by journalists and news organisations, and, more broadly, the relationship
between computer-based technologies and journalism.

Professor Rasmus Kleis Nielsen is Director of the Reuters Institute for the Study of
Journalism, Professor of Political Communication at the University of Oxford, and served
as Editor-in-Chief of the International Journal of Press/Politics from 2015 to 2018. His work
focuses on changes in the news media, political communication, and the role of digital
technologies in both.

Acknowledgements

We would like to thank Caryhs Innes, Xhoana Beqiri, and the rest of the team at YouGov for
their work on fielding the survey. We would also like to thank Felix Simon for his help with the
data analysis. We are grateful to the other members of the research team at RISJ for their input
on the questionnaire and interpretation of the results, and to Kate Hanneford-Smith, Alex
Reid, and Rebecca Edwards for helping to move this project forward and keeping us on track.

2
 WHAT DOES THE PUBLIC IN SIX COUNTRIES THINK OF GENERATIVE AI IN NEWS?

Executive Summary

Based on an online survey focused on understanding if and how people use generative
artificial intelligence (AI), and what they think about its application in journalism and other
areas of work and life across six countries (Argentina, Denmark, France, Japan, the UK, and
the USA), we present the following findings.

Findings on the public’s use of generative AI

ChatGPT is by far the most widely recognised generative AI product – around 50% of the
online population in the six countries surveyed have heard of it. It is also by far the most
widely used generative AI tool in the six countries surveyed. That being said, frequent use of
ChatGPT is rare, with just 1% using it on a daily basis in Japan, rising to 2% in France and the
UK, and 7% in the USA. Many of those who say they have used generative AI have used it just
once or twice, and it is yet to become part of people’s routine internet use.

In more detail, we find:

• While there is widespread awareness of generative AI overall, a sizable minority of the

public – between 20% and 30% of the online population in the six countries surveyed –
have not heard of any of the most popular AI tools.

• In terms of use, ChatGPT is by far the most widely used generative AI tool in the six
countries surveyed, two or three times more widespread than the next most widely used
products, Google Gemini and Microsoft Copilot.

• Younger people are much more likely to use generative AI products on a regular basis.
Averaging across all six countries, 56% of 18–24s say they have used ChatGPT at least
once, compared to 16% of those aged 55 and over.

• Roughly equal proportions across six countries say that they have used generative AI
for getting information (24%) as creating various kinds of media, including text but also
audio, code, images, and video (28%).

• Just 5% across the six countries covered say that they have used generative AI to get the
latest news.

3
 THE REUTERS INSTITUTE FOR THE STUDY OF JOURNALISM

Findings on public opinion about the use of generative AI in

different sectors
Most of the public expect generative AI to have a large impact on virtually every sector of
society in the next five years, ranging from 51% expecting a large impact on political parties to
66% for news media and 66% for science. But there is significant variation in whether people
expect different sectors to use AI responsibly – ranging from around half trusting scientists
and healthcare professionals to do so, to less than one-third trusting social media companies,
politicians, and news media to use generative AI responsibly.

In more detail, we find:

• Expectations around the impact of generative AI in the coming years are broadly similar
across age, gender, and education, except for expectations around what impact generative
AI will have for ordinary people – younger respondents are much more likely to expect a
large impact in their own lives than older people are.

• Asked if they think that generative AI will make their life better or worse, a plurality in
four of the six countries covered answered ‘better’, but many have no strong views, and a
significant minority believe it will make their life worse. People’s expectations when asked
whether generative AI will make society better or worse are generally more pessimistic.

• Asked whether generative AI will make different sectors better or worse, there is
considerable optimism around science, healthcare, and many daily routine activities,
including in the media space and entertainment (where there are 17 percentage points
more optimists than pessimists), and considerable pessimism for issues including cost of
living, job security, and news (8 percentage points more pessimists than optimists).

• When asked their views on the impact of generative AI, between one-third and half of our
respondents opted for middle options or answered ‘don’t know’. While some have clear
and strong views, many have not made up their mind.

Findings on public opinion about the use of generative AI in journalism

Asked to assess what they think news produced mostly by AI with some human oversight
might mean for the quality of news, people tend to expect it to be less trustworthy and less
transparent, but more up to date and (by a large margin) cheaper for publishers to produce.
Very few people (8%) think that news produced by AI will be more worth paying for compared to
news produced by humans.

In more detail, we find:

• Much of the public think that journalists are currently using generative AI to complete
certain tasks, with 43% thinking that they always or often use it for editing spelling and
grammar, 29% for writing headlines, and 27% for writing the text of an article.

4
 WHAT DOES THE PUBLIC IN SIX COUNTRIES THINK OF GENERATIVE AI IN NEWS?

• Around one-third (32%) of respondents think that human editors check AI outputs to
make sure they are correct or of a high standard before publishing them.

• People are generally more comfortable with news produced by human journalists than
by AI.

• Although people are generally wary, there is somewhat more comfort with using news
produced mostly by AI with some human oversight when it comes to soft news topics
like fashion (+7 percentage point difference between comfortable and uncomfortable)
and sport (+5) than with ‘hard’ news topics, including international affairs (-21) and,
especially, politics (-33).

• Asked whether news that has been produced mostly by AI with some human oversight
should be labelled as such, the vast majority of respondents want at least some disclosure
or labelling. Only 5% of our respondents say none of the use cases we listed need to
be disclosed.

• There is less consensus on what uses should be disclosed or labelled. Around one-third
think ‘editing the spelling and grammar of an article’ (32%) and ‘writing a headline’ (35%)
should be disclosed, rising to around half for ‘writing the text of an article’ (47%) and
‘data analysis’ (47%).

• Again, when asked their views on generative AI in journalism, between a third and half of
our respondents opted for neutral middle options or answered ‘don’t know’, reflecting a
large degree of uncertainty and/or recognition of complexity.

5
 THE REUTERS INSTITUTE FOR THE STUDY OF JOURNALISM

Introduction

The public launch of OpenAI’s ChatGPT in November 2022 and subsequent developments have
spawned huge interest in generative AI. Both the underlying technologies and the range of
applications and products involving at least some generative AI have developed rapidly (though
unevenly), especially since the publication in 2017 of the breakthrough ‘transformers’ paper
(Vaswani et al. 2017) that helped spur new advances in what foundation models and Large
Language Models (LLMs) can do.

These developments have attracted much important scholarly attention, ranging from
computer scientists and engineers trying to improve the tools involved, to scholars testing
their performance against quantitative or qualitative benchmarks, to lawyers considering their
legal implications. Wider work has drawn attention to built-in limitations, issues around the
sourcing and quality of training data, and the tendency of these technologies to reproduce
and even exacerbate stereotypes and thus reinforce wider social inequalities, as well as the
implications of their environmental impact and political economy.

One important area of scholarship has focused on public use and perceptions of AI in general,
and generative AI in particular (see, for example, Ada Lovelace Institute 2023; Pew 2023). In
this report, we build on this line of work by using online survey data from six countries to
document and analyse public attitudes towards generative AI, its application across a range of
different sectors in society, and, in greater detail, in journalism and the news media specifically.

We go beyond already published work on countries including the USA (Pew 2023; 2024),
Switzerland (Vogler et al. 2023), and Chile (Mellado et al. 2024), both in terms of the questions
we cover and specifically in providing a cross-national comparative analysis of six countries
that are all relatively privileged, affluent, free, and highly connected, but have very different
media systems (Humprecht et al. 2022) and degrees of platformisation of their news media
system in particular (Nielsen and Fletcher 2023).

The report focuses on the public because we believe that – in addition to economic, political,
and technological factors – public uptake and understanding of generative AI will be among the
key factors shaping how these technologies are being developed and are used, and what they,
over time, will come to mean for different groups and different societies (Nielsen 2024). There
are many powerful interests at play around AI, and much hype – often positive salesmanship,
but sometimes wildly pessimistic warnings about possible future risks that might even distract
us from already present issues. But there is also a fundamental question of whether and how
the public at large will react to the development of this family of products. Will it be like
blockchain, virtual reality, and Web3? All promoted with much bombast but little popular
uptake so far. Or will it be more like the internet, search, and social media – hyped, yes, but also
quickly becoming part of billions of people’s everyday media use.

6
 WHAT DOES THE PUBLIC IN SIX COUNTRIES THINK OF GENERATIVE AI IN NEWS?

To advance our understanding of these issues, we rely on data from an online survey focused on
understanding if and how people use generative AI, and what they think about its application
in journalism and other areas of work and life. In the first part of the report, we present the
methodology, then we go on to cover public awareness and use of generative AI, expectations
for generative AI’s impact on news and beyond, how people think AI is being used by journalists
right now, and how people think about how journalists should use generative AI, before offering
a concluding discussion.

As with all survey-based work, we are reliant on people’s own understanding and recall. This
means that many responses here will draw on broad conceptions of what AI is and might mean,
and that, when it comes to generative AI in particular, people are likely to answer based on their
experience of using free-standing products explicitly marketed as being based on generative
AI, like ChatGPT. Most respondents will be less likely to be thinking about incidents where
they may have come across functionalities that rely in part on generative AI, but do not draw as
much attention to it – a version of what is sometimes called ‘invisible AI’ (see, for example, Alm
et al. 2020). We are also aware that these data reflect a snapshot of public opinion, which can
fluctuate over time.

We hope the analysis and data published here will help advance scholarly analysis by
complementing the important work done on the use of AI in news organisations (for example,
Beckett and Yaseen 2023; Caswell 2024; Diakopoulos 2019; Diakopoulos et al 2024; Newman
2024; Simon 2024), including its limitations and inequities (see, for example, Broussard
2018, 2023; Bender et al. 2021), and help centre the public as a key part of how generative AI
will develop and, over time, potentially impact many different sectors of society, including
journalism and the news media.

7
 THE REUTERS INSTITUTE FOR THE STUDY OF JOURNALISM

Methodology

The report is based on a survey conducted by YouGov on behalf of the Reuters Institute for the
Study of Journalism (RISJ) at the University of Oxford. The main purpose is to understand if and
how people use generative AI, and what they think about its application in journalism and other
areas of work and life.

The data were collected by YouGov using an online questionnaire fielded between 28 March and
30 April 2024 in six countries: Argentina, Denmark, France, Japan, the UK, and the USA.

YouGov was responsible for the fieldwork and provision of weighted data and tables only, and
RISJ was responsible for the design of the questionnaire and the reporting and interpretation of
the results.

Samples in each country were assembled using nationally representative quotas for age group,
gender, region, and political leaning. The data were weighted to targets based on census or
industry-accepted data for the same variables.

Sample sizes are approximately 2,000 in each country. The use of a non-probability sampling
approach means that it is not possible to compute a conventional ‘margin of error’ for
individual data points. However, differences of +/- 2 percentage points (pp) or less are very
unlikely to be statistically significant and should be interpreted with a very high degree of
caution. We typically do not regard differences of +/- 2pp as meaningful, and as a general rule
we do not refer to them in the text.
Table 1. Nationally representative sample sizes
Table 1. Nationally representative sample sizes

Country Sample Size Fieldwork Dates

Argentina 2,018 9th to 23rd April 2024

Denmark 2,011 9th to 22nd April 2024

France 2,056 9th to 22nd April 2024

Japan 2,007 16th to 30th April 2024

UK 2,113 28th March to 5th April 2024

USA 2,012 28th March to 5th April 2024

It is important to note that online samples tend to under-represent the opinions and
behaviours of people who are not online (typically those who are older, less affluent, and have
limited formal education). Moreover, because people usually opt in to online survey panels,
they tend to over-represent people who are well educated and socially and politically active.

8
 WHAT DOES THE PUBLIC IN SIX COUNTRIES THINK OF GENERATIVE AI IN NEWS?

Some parts of the survey require respondents to recall their past behaviour, which can be
flawed or influenced by various biases. Additionally, respondents’ beliefs and attitudes related
to generative AI may be influenced by social desirability bias, and when asked about complex
socio-technical issues, people will not always be familiar with the terminology experts rely on
or understand the terms the same way. We have taken steps to mitigate these potential biases
and sources of error by implementing careful questionnaire design and testing.

Some figures in this report do not display all of the percentages. All percentages can be viewed
in the interactive figures at: https://reutersinstitute.politics.ox.ac.uk/what-does-public-six-
countries-think-generative-ai-news.

9
 THE REUTERS INSTITUTE FOR THE STUDY OF JOURNALISM

1. Public Awareness and Use of Generative AI

Most of our respondents have, by now, heard of at least some of the most popular generative AI
tools. ChatGPT is by far the most widely recognised of these, with between 41% (Argentina) and
61% (Denmark) saying they’d heard of it.

Other tools, typically those built by incumbent technology companies – such as Google Gemini,
Microsoft Copilot, and Snapchat My AI – are some way behind ChatGPT, even with the boost
that comes from being associated with a well-known brand. They are, with the exception of Grok
from X, each recognised by roughly 15–25% of the public.

Tools built by specialised AI companies, such as Midjourney and Perplexity, currently have little
to no brand recognition among the public at large. And there’s little national variation here,
even when it comes to brands like Mistral in France; although it is seen by some commentators
as a national champion, it clearly hasn’t yet registered with the wider French population.

We should also remember that a sizable minority of the public – between 19% of the online
population in Japan and 30% in the UK – have not heard of any of the most popular AI tools
(including ChatGPT) despite nearly two years of hype, policy conversations, and extensive
media coverage.
Figure 1. Proportion who have heard of each generative AI tool
Figure 1. Proportion that have heard of each generative AI tool
In every country, awareness of ChatGPT is much higher than for all other tools. Next are tools from large technology
In every country, awareness of ChatGPT is much higher than for all other tools. Next are tools from large
companies, followed by specialised AI products.
technology companies, followed by specialised AI products.

Argentina Denmark France Japan UK USA

ChatGPT 41% 61% 55% 56% 58% 53%

Google Gemini (formerly Bard) 15% 15% 13% 17% 15% 24%

Snapchat My AI 17% 29% 13% 4% 14% 21%

Microsoft Copilot 15% 13% 13% 14% 17% 22%

Meta AI (LLaMA) 12% 7% 15% 13% 12% 27%

Bing AI 11% 12% 8% 11% 17% 24%

YouChat 15% 5% 10% 5% 7% 16%

Midjourney 4% 6% 8% 2% 8% 7%

Rakuten AI 4% 1% 5% 6% 3% 7%

Replika 3% 2% 3% 1% 3% 7%

Claude 3% 2% 3% 2% 3% 5%

Grok 1% 2% 2% 1% 4% 6%

Mistral (Mixtral) 2% 2% 3% 2% 2% 3%

Perplexity.ai 2% 1% 2% 1% 2% 3%

None of these 22% 21% 24% 19% 30% 19%

AI_brandheard. Have you heard of any of the following generative AI chatbots or tools? (Please select all that
AI_brandheard.
apply). Have you
Base: Total sample heard
in each of any
country of the following generative AI chatbots or tools? (Please select all that apply). Base: Total
≈ 2000.
sample in each country ≈ 2000.

10
 WHAT DOES THE PUBLIC IN SIX COUNTRIES THINK OF GENERATIVE AI IN NEWS?

While our Digital News Report (Newman et al. 2023) shows that in most countries the news
market is dominated by domestic brands that focus on national news, in contrast, the search
and social platform space across countries tends to feature the same products from large
technology companies such as Google, Meta, and Microsoft. At least for now, it seems like the
generative AI space will follow the pattern from the technology sector, rather than the more
nationally oriented one of news providers serving distinct markets defined in part by culture,
history, and language.

The pattern we see for awareness in Figure 1 extends to use, with ChatGPT by far the most
widely used generative AI tool in the six countries surveyed. Use of ChatGPT is roughly two or
three times more widespread than the next products, Google Gemini and Microsoft Copilot.
What’s also clear from Figure 2 is that, even when it comes to ChatGPT, frequent use is rare,
with just 1% using it on a daily basis in Japan, rising to 2% in France and the UK, and 7% in the
USA. Many of those who say they have used generative AI have only used it once or twice, and it
is yet to become part of people’s routine internet use.
How frequently people use ChatGPT, Gemini and Copilot
Figure 2. How frequently people use ChatGPT, Gemini, and Copilot
ChatGPT is the most widely used generative AI product, but few use it frequently.
ChatGPT is the most widely used generative AI product, but few use it frequently.

Daily Weekly Monthly Once or twice Never Don't know Not heard of

ChatGPT 0% 20% 40% 60% 80% 100%

Denmark – 35% 8% 9% 15% 25% 39%

USA – 32% 7% 11% 10% 20% 47%

UK – 29% 7% 15% 27% 42%

Argentina – 28% 7% 11% 12% 59%

France – 27% 7% 13% 27% 45%

Japan – 22% 12% 33% 44%

Google Gemini 20% 40% 60% 80% 100%

USA – 14% 10% 76%

Argentina – 11% 85%

Japan – 9% 8% 83%

UK – 7% 8% 85%

France – 6% 7% 87%

Denmark – 6% 9% 85%

Micosoft Copilot 20% 40% 60% 80% 100%

USA – 12% 10% 78%

Argentina – 10% 85%

France – 8% 87%

UK – 7% 9% 83%

Japan – 7% 7% 86%

Denmark 6% 6% 87%

AI_branduse. How often, if at all, do you typically use each of the following generative AI chatbots or tools for any purpose?
Base: Total sample in each country ≈ 2000.

11generative AI chatbots or tools for

AI_branduse. How often, if at all, do you typically use each of the following
any purpose? Base: Total sample in each country ≈ 2000.
Source: Data from 'What does the public in six countries think of generative AI in news?,' published in May 2024
 THE REUTERS INSTITUTE FOR THE STUDY OF JOURNALISM

Use of ChatGPT is slightly more common among men and those with higher levels of formal
education, but the biggest differences are by age group, with younger people much more likely
to have ever used it, and to use it on a regular basis (Figure 3). Averaging across all six countries,
16% of those aged 55 and over say they have used ChatGPT at least once, compared to 56% of
18–24s. But even among this age group infrequent use is the norm, with just over half of users
saying they use it monthly or less.

Proportion that that

Figure 3. Proportion have ever
have everused ChatGPT
used ChatGPT by age group
Averaging across all six countries, younger people are much more likely to say they have ever used ChatGPT, but
Averaging
even amongacross all six countries,
younger younger people
people frequent use isare much more likely to say they have ever used
rare.
ChatGPT, but even among younger people frequent use is rare.

56%

50%
17%
43%

40
15% 33%
12%
30 28%
14%
8%
20 13%
18% 16%
7%
13% Once or twice
5% 9%
10
8% Monthly
6% 2%
9%
6% 4% 3% Weekly
3% 1%
18–24 25–34 35–44 45–54 55+ Daily

AI_branduse. How often, if at all, do you typically use each of the following generative AI chatbots or tools for
AI_branduse.
any How
purpose? Base: often, if at all, do you typically
18–24/25–34/35–44/45–54/55+ use each
across of the
Argentina, following
Denmark, generative
France, AI chatbots
Japan, UK, USA = or tools for any purpose?
Base: 18–24/25–34/35–44/45–54/55+ across Argentina, Denmark, France, Japan, UK, USA = 1272/2038/1935/2020/4952.
1272/2038/1935/2020/4952.
Source: Data from 'What does the public in six countries think of generative AI in news?,' published in May 2024

Although people working in many different industries – including news and journalism – are
looking for ways of deploying generative AI, people in every country apart from Argentina are
slightly more likely to say they are using it in their private life rather than at work or school
(Figure 4). If providers of AI products convince more companies and organisations that these
tools can deliver great efficiencies and new opportunities this may change, with professional
use becoming more widespread and potentially spilling over to people’s personal lives – a
dynamic that was part of how the use of personal computers, and later the internet, spread.
However, at this stage private use is more widespread.

12
Figure 4. Proportion that say they have used generative AI
WHAT DOES THE PUBLIC IN SIX COUNTRIES THINK OF GENERATIVE AI IN NEWS?

in each
Figure context that say they have used generative AI in each context
4. Proportion
Inmost
In mostcountries,
countries, people
people are slightly
are slightly moretolikely
more likely to have
say they say they
used have usedAIgenerative
generative AI in their personal rather than
in their personal
their professional lives.
rather than their professional lives.

In my private life At work/school

Six-country 27%
average 21%

35%
USA
28%

30%
Denmark
22%

25%
France
19%

25%
UK
20%

23%
Argentina
26%

23%
Japan
12%

AI_place. You said you have used a generative AI chatbot (e.g. ChatGPT, Microsoft Copilot, etc.) or tool … Which,
ifAI_place.
any, of the You said have
following you have used
you tried to ause
generative
it for (evenAI chatbot
if it (e.g. ChatGPT,
didn't work)? Base: TotalMicrosoft Copilot,
sample in each etc.)≈ or tool … Which, if any, of the
country
following have you tried to use it for (even if it didn’t work)? Base: Total sample in each country ≈ 2000.
2000.

Averaging across six countries, roughly equal proportions say that they have used generative
AI for getting information (24%) as creating media (28%), which as a category includes creating
images (9%), audio (3%), video (4%), code (5%), and generating text (Figure 5). When it comes
to creating text more specifically, people report using generative AI to write emails (9%) and
essays (8%), and for creative writing (e.g. stories and poems) (7%). But it’s also clear that many
people who say they have used generative AI for creating media have just been playing around
or experimenting (11%) rather than looking to complete a specific real-world task. This is also
true when it comes to using generative AI to get information (9%), but people also say they
have used it for answering factual questions (11%), advice (10%), generating ideas (9%), and
summarisation (8%).

13
 THE REUTERS INSTITUTE FOR THE STUDY OF JOURNALISM

Proportion that have used generative AI for each task

Figure 5. Proportion that have used generative AI for each task
Averaging across six countries, roughly equal proportions of people have used generative AI for getting
Averaging across six countries, roughly equal proportions of people have used generative AI for getting information
information as creating media, but using generative AI for news is rare.
as creating media, but using generative AI for news is rare.

For getting information 24% For creating media 28%

Answering factual 11% Playing around or 11%

questions experimenting

Asking advice 10% Writing an email or letter 9%

Generating ideas 9% Making an image 9%

Playing around or 9% 8%
Writing an essay or report
experimenting

Summarising text 8% Creative writing 7%

Seeking support 7% A job application/interview 5%

Recommendations 6% Programming or coding 5%

Translations 6% Making a video 4%

Getting the latest news 5% Making audio 3%

Data analysis 5% Creating test data 3%

Other 1% Other 2%

AI_outputs. You said you have used a generative AI chatbot (e.g. ChatGPT, Microsoft Copilot, etc.) or tool …
AI_outputs.
Which, if any, ofYou
the said you have
following used
have you a generative
tried AI chatbot
to use it for (even (e.g.
if it didn't ChatGPT,
work)? Base: Microsoft Copilot,
Total sample across etc.) or tool … Which, if any, of the
following Denmark,
Argentina, have youFrance,
tried toJapan,
use itUK,
forUSA
(even if it didn’t work)? Base: Total sample across Argentina, Denmark, France, Japan, UK,
= 12,217.
USA = 12,217.
Source: Data from 'What does the public in six countries think of generative AI in news?,' published in May 2024

An average of 5% across the six countries say that they have used generative AI to get the latest
news, making it less widespread than most of the other uses that were mentioned previously.
One reason for this is that the free version of the most widely used generative AI product –
ChatGPT – is not yet connected to the web, meaning that it cannot be used for the latest news.
Furthermore, our previous research has shown that around half of the most widely used news
websites are blocking ChatGPT (Fletcher 2024), and partly as a result, it is rarely able to deliver
the latest news from specific outlets (Fletcher et al. 2024).

The figures for using generative AI for news vary by country, from just 2% in the UK and
Denmark to 10% in the USA (Figure 6). The 10% figure in the USA is probably partly due to
the fact that Google has been trialling Search Generative Experiences (SGE) there for the last
year, meaning that people who use Google to search for a news-related topic – something
that 23% of Americans do each week (Newman et al. 2023) – may see some generative AI text
that attempts to provide an answer. However, given the documented limitations of generative
AI when it comes to factual precision, companies like Google may well approach news more
cautiously than other types of content and information, and the higher figure in the USA may
also simply be because generative AI is more widely used there generally.

14
Figure 6. Proportion that
WHAT DOES THE PUBLIC say
IN SIX they
COUNTRIES have
THINK usedAIgenerative
OF GENERATIVE IN NEWS? AI
to try and get the latest news
Figure 6. Proportion that say they have used generative AI to try to get the latest news
Using generative
Using generativeAIAItotoget
getthe
thelatest news
latest is most
news common
is most in the
common in USA, which
the USA, may be
which maypartly because
be partly people are
because
seeing generative AI search results in Google.
people are seeing generative AI search results in Google.

Six-country
5%
average

USA 10%

Argentina 6%

Japan 5%

France 3%

Denmark 2%

UK 2%

AI_tasks_information.
AI_tasks_information. YouYou saidsaid
youyou
havehave
usedused a generative
a generative AI chatbot
AI chatbot (e.g. ChatGPT,
(e.g. ChatGPT, MicrosoftMicrosoft Copilot,
Copilot, etc.) or tooletc.) or
for getting
tool for getting
information information
... Which, if any, of...
theWhich, if any,
following haveofyou
thetried
following
to use have you tried
it for (even to use
if it didn’t it for Base:
work)? (evenTotal
if it didn't
samplework)?
in each
Base:
countryTotal sample in each country ≈ 2000.
≈ 2000.

Numerous examples have been documented of generative AI giving incorrect answers when
asked factual questions, as well as other forms of so-called ‘hallucination’ that result in poor-
quality outputs (e.g. Angwin et al. 2024). Although some are quick to point out that it is wrong
to expect generative AI to be good at information-based tasks – at least at its current state of
development – some parts of the public are experimenting with doing exactly that.

Given the known problems when it comes to reliability and veracity, it is perhaps concerning
that our data also show that users seem reasonably content with the performance – most of
those (albeit a rather small slice of the online population) who have tried to use generative AI
for information-based tasks generally say they trusted the outputs (Figure 7).

In interpreting this, it is important to keep in mind two important caveats.

First, the vast majority of the public has not used generative AI for information-based tasks, so
we do not know about their level of trust. Other evidence suggests that trust among the large
part of the public that has not used generative AI is low, meaning overall trust levels are likely
to be low (Pew 2024).

Second, people are more likely to say that they ‘somewhat trust’ the outputs rather than
‘strongly trust’, which indicates a degree of scepticism – their trust is far from unconditional.
However, this may also mean that from the point of view of members of the public who have
used the tools, information from generative AI while clearly not perfect is already good enough
for many purposes, especially tasks like generating ideas.

15
 THE REUTERS INSTITUTE FOR THE STUDY OF JOURNALISM

Figure 7. Proportion that say they trusted the generative AI outputs for each task
Averaging across six countries, people who have used generative AI to get information mostly trust the outputs, but
most people have not tried to use generative AI.

Strongly trust Somewhat trust Neither trust nor distrust Don't know
Somewhat distrust Strongly distrust Have not used
Strongly trust Somewhat trust Neither trust nor distrust Don't know
Somewhat distrust Strongly
Answering factual questions distrust Have not used

Answering factual questions

Generating ideas

Generating ideas
Getting the latest news

Getting the latest news

Averaging across six countries, people who have used generative AI to create media mostly think it performed well,
but most people have not tried to use generative AI.
Very well Somewhat well Don't know Somewhat badly Very badly Have not used

Very well Somewhat well Don't know Somewhat badly Very badly Have not used
Writing an email or letter

Writing an email or letter

Making an image

Making an image
Programming or coding

Programming or coding

AI_tasktrust. You said you have used a generative AI chatbot (e.g. ChatGPT, Microsoft Copilot, etc.) or tool for getting
information ... Generally speaking, do you trust or distrust the outputs when you use it for each of the following?
AI_taskperformance. You said you have used a generative AI chatbot (e.g. ChatGPT, Microsoft Copilot, etc.) or tool for creating
media (e.g. text, images, video, audio, code, data) ... Generally speaking, do you think it performs well or badly when you use it for
each of the following? Base: Total sample in each country ≈ 2000. Note: See website for percentages.

When we ask people who have used generative AI to create media whether they think the
product they used did it well or badly, we see a very similar picture. Most of those who have
tried to use generative AI to create media think that it did it ‘very’ or ‘somewhat’ well, but again,
we can only use this data to know what users of the technology think.

The general population’s views on the media outputs may look very different, and while early
adopters seem to have some trust in generative AI, and feel these technologies do a somewhat
good job for many tasks, it is not certain that everyone will feel the same, even if or when they
start using generative AI tools.

16
 WHAT DOES THE PUBLIC IN SIX COUNTRIES THINK OF GENERATIVE AI IN NEWS?

2. Expectations for Generative AI’s Impact on News

and Beyond

We now move from people’s awareness and use of generative AI products to their
expectations around what the development of these technologies will mean. First, we find
that most of the public expect generative AI to have a large impact on virtually every sector
of society in the next five years (Figure 8). For every sector, there is a smaller number who
expect low impact (compared to a large impact), and a significant number of people (roughly
between 15% and 20%) who answer ‘don’t know’.

Averaging across six countries, we find that around three-quarters of respondents think
generative AI will have a large impact on search and social media companies (72%), while
two-thirds (66%) think that it will have a large impact on the news media – strikingly, the
same proportion who think it will have a large impact upon the work of scientists (66%).
Around half think that generative AI will have a large impact upon national governments
(53%) and politicians and political parties (51%).

Interestingly, there are generally fewer people who expect it will have a large impact on
ordinary people (48%). Much of the public clearly thinks the impact of generative AI will be
mediated by various existing social institutions.

Bearing in mind how different the countries we cover are in many respects, including in
terms of how people use and think about news and media (see, for example, Newman et al.
2023), it is striking that we find few cross-country differences in public expectations around
the impact of generative AI. There are a few minor exceptions. For example, expectations
around impact for politicians and political parties are a bit higher than average in the USA
(60% vs 51%) and a bit lower in Japan (44% vs 51%) – but, for the most part, views across
countries are broadly similar.

17
Proportion that think generative AI will have a large
THE REUTERS INSTITUTE FOR THE STUDY OF JOURNALISM

impact upon each

Figure 8. Proportion that think generative AI will have a large impact upon each
Averaging across six countries, most of the public expect generative AI to have a large impact on virtually every
Averaging across six countries, most of the public expect generative AI to have a large impact on virtually
sector of society in the next five years, including the news media.

Very/somewhat large impact Don't know Very/somewhat small impact

Social media companies 72% 16% 12%

Search engine companies 71% 16% 12%

Scientists 66% 18% 17%

News media 66% 17% 17%

Healthcare professionals 59% 18% 23%

59% 20% 22%

Military 56% 21% 23%

The national government 53% 21% 25%

Politicians and political parties 51% 21% 28%

Law enforcement 50% 21% 29%

Ordinary people 48% 17% 35%

Retailers 47% 21% 32%

AI_actorsimpact. How much impact, if any, do you think generative AI will have on the actions of each of the
AI_actorsimpact.
following in the next 5How
yearsmuch impact,
(i.e. April if any,
2029)? Base:do yousample
Total think generative AI will
across Argentina, have onFrance,
Denmark, the actions
Japan,ofUK,
each of the following in the next
5 years
USA (i.e. April 2029)? Base: Total sample across Argentina, Denmark, France, Japan, UK, USA = 12,217.
= 12,217.
Source: Data from 'What does the public in six countries think of generative AI in news?,' published in May 2024

For almost all these sectors, there is little variation across age and gender, and the main
difference when it comes to different levels of education is that respondents with lower levels
of formal education are more likely to respond with ‘don’t know’, and those with higher levels
of education are more likely to expect a large impact. The number who expect a small impact
remains broadly stable across levels of education.

The only exception to this relative lack of variation by demographic factors is expectations
around what impact generative AI will have for ordinary people. Younger respondents, who, as
we have shown in earlier sections, are much more likely to have used generative AI tools, are
also much more likely to expect a large impact within the next five years than older people, who
often have little or no personal experience of using generative AI (Figure 9).

18
 WHAT DOES THE PUBLIC IN SIX COUNTRIES THINK OF GENERATIVE AI IN NEWS?

Figure 9. Proportion that think generative AI will have a large impact on ordinary people
Younger people in every country are more likely to think that generative AI will have a large impact on ordinary
people in the next five years.

60%

Argentina
50%
USA
Six-country average
40% Japan
Denmark
UK
30% France

20%
18–24 25–34 35–44 45–54 55+

AI_actorsimpact. How much impact, if any, do you think generative AI will have on the actions of each of the following in the
10
next 5 years (i.e. April 2029)? Base: 18–24/25–34/35–44/45–54/55+ across Argentina, Denmark, France, Japan, UK, USA =
1272/2038/1935/2020/4952.

Expectations around the impact of generative AI, whether large or small, in themselves say
nothing about how people think about whether this impact will, on balance, be for better or
for worse.

Because generative AI use is highly mediated by institutions, and our data document that
much of the public clearly recognise this, a useful additional way to think about expectations
is to consider whether members of the public trust different sectors to make responsible use
of generative AI.

We find that public trust in different institutions to make responsible use of generative AI is
generally quite low (Figure 10). While around half in most of the six counties trust scientists
and healthcare professionals to use generative AI responsibly, the figures drop below 40% for
most other sectors in most countries. Figures for social media companies are lower than many
other sectors, as are those for news media, ranging from 12% in the UK to 30% in Argentina
and the USA.

There is more cross-country variation in public trust and distrust in different institutions’
potential use of generative AI, partly in line with broader differences from country to country
in terms of trust in institutions.

19
Figure 10. Proportion that strongly/somewhat trusts each
THE REUTERS INSTITUTE FOR THE STUDY OF JOURNALISM

to use10.
Figure generative AIstrongly/somewhat
Proportion that responsibly trusts each to use generative AI responsibly
While around half in most counties trust scientists and healthcare professionals to use generative AI responsibly,
While around half in most counties trust scientists and healthcare professionals to use generative AI
the figures for news media range from 12% in the UK to 30% in Argentina and the USA.

Argentina Denmark France Japan UK USA

Healthcare professionals 53% 45% 47% 51% 51% 53%

Scientists 54% 47% 44% 44% 52% 50%

Military 33% 35% 37% 25% 33% 42%

Law enforcement 32% 39% 34% 30% 28% 40%

31% 26% 22% 40% 24% 37%

Search engine companies 38% 22% 28% 34% 20% 36%

Retailers 33% 25% 21% 29% 20% 33%

Ordinary people 29% 24% 20% 20% 21% 32%

News media 30% 21% 18% 23% 12% 30%

The national government 21% 30% 19% 18% 13% 28%

Social media companies 30% 14% 18% 23% 9% 27%

Politicians and political parties 15% 16% 13% 12% 7% 21%

AI_actorstrust. How much do you trust or distrust each of the following to make responsible use of generative
AI_actorstrust.
AI? Howinmuch
Base: Total sample do you≈trust
each country 2000. or distrust each of the following to make responsible use of generative AI? Base: Total
sample in each country ≈ 2000.

But there are also some overarching patterns.

First, younger people, while still often sceptical, are for many sectors more likely to say they
trust a given institution to use generative AI responsibly, and less likely to express distrust. This
tendency is most pronounced in the sectors viewed with greatest scepticism by the public at
large, including the government, politicians, and ordinary people, as well as news media, social
media, and search engines.

Second, a significant part of the public does not have a firm view on whether they trust or
distrust different institutions to make responsible use of generative AI. Varying from sector
to sector and from country to country, between roughly one-quarter and half of respondents
answer ‘neither trust nor distrust’ or ‘don’t know’ when asked. There is much uncertainty and
often limited personal experience; in that sense, the jury is still out.

Leaving aside country differences for a moment and looking at the aggregate across all six
countries, we can combine our data on public expectations around the size of the impact
that generative AI will have with expectations around whether various sectors will use these
technologies responsibly. This will provide an overall picture of how people think about these
issues across different social institutions (Figure 11).

If we compare public perceptions relative to the average percentage of respondents who expect
a large impact across all sectors (58%, marked by the vertical dashed line in Figure 11) and the
average percentage of respondents who distrust actors in a given sector to make responsible
use of generative AI (33%, marked by the horizontal dashed line), we can group expectations
from sector to sector into four quadrants.

20
 WHAT DOES THE PUBLIC IN SIX COUNTRIES THINK OF GENERATIVE AI IN NEWS?

• First, there are those sectors where people expect generative AI to have a relatively large
impact, but relatively few expect it will be used irresponsibly (e.g. healthcare and science).

• Second, there are sectors where people expect the impact may not be as great, and
relatively fewer fear irresponsible use (e.g. ordinary people and retailers).

• Third, there are sectors where relatively few people expect a large impact, and relatively
more people are worried about irresponsible use (e.g. government and political parties).

• Finally, there are sectors where more people expect large impact, and more people fear
irresponsible use by the actors involved (e.g. social media and the news media, who are
Figure 11. Proportion that distrust each to use
viewed very similarly by the public in this respect).
generative AI responsibly plotted against
proportion who think
Figure 11. Proportion it willeach
that distrust have a large
to use impact
generative AI responsibly plotted against proportion
that think it will have a large impact
On average across six countries, people think that generative AI will have an above
On average across six countries, people think that generative AI will have an above average impact on the news
average impact on the news media, but there is above average distrust in them to use it
media, but there is above average distrust in them to use it responsibly.
responsibly.

100%
Distrust to use Average impact
across all sectors
generative AI
responsibly

80

Politicians
60 and political
parties

News
40 media
Average distrust
across all sectors

Retailers
20
Scientists

Generative AI will
have a large
impact
0
0 20 40 60 80 100%

AI_actorsimpact. How much impact, if any, do you think generative AI will have on the
AI_actorsimpact.
actions of each of theHow much
following impact,
in the next 5ifyears
any, do
(i.e.you
Aprilthink generative
2029)? AI willHow
AI_actorstrust. have on the actions of each of the following in the
next 5doyears
much (i.e.or
you trust April
distrust eachAI_actorstrust.
2029)? of the following toHowmakemuch do youuse
responsible trust or distrust
of generative each of the following to make responsible use of
AI?
generative
Base: AI? Base:
Total sample Total
across sampleDenmark,
Argentina, across Argentina,
France, Japan, Denmark,
UK, USAFrance, Japan, UK, USA = 12,217.
= 12,217.

It is important to keep this quite nuanced and differentiated set of expectations in mind in
interpreting people’s general expectations around what impact they think generative AI will
have for them personally, as well as for society at large.

Asked if they think that generative AI will make their life better or worse, more than half of our
respondents answer ‘neither better nor worse’ or ‘don’t know’, with a plurality in four of the six
countries covered answering ‘better’, and a significant minority ‘worse’ (Figure 12). The large

21
 THE REUTERS INSTITUTE FOR THE STUDY OF JOURNALISM

number of people with no strong expectations either way is consistent across countries, but the
balance between more optimistic responses and more pessimistic ones varies.
Figure 12. Proportion that think generative AI will make
each
Figure better
12. Proportion that think generative AI will make each better
People are slightly more pessimistic about the impact of generative AI on society compared to the impact on their
People are slightly more pessimistic about the impact of generative AI on society compared to the
own lives, but many people are uncertain.
impact on their own lives, but many people are uncertain.

Much/somewhat better Neither Don't know Much/somewhat worse

Six-country average
Society 30% 28% 11% 31%
My life 28% 41% 12% 18%

Argentina
Society 44% 21% 12% 23%
My life 41% 34% 14% 11%

Denmark
Society 27% 34% 12% 27%
My life 23% 49% 12% 15%

France
Society 18% 30% 10% 42%
My life 20% 43% 11% 26%

Japan
Society 34% 42% 8% 16%
My life 27% 51% 11% 11%

UK
Society 22% 23% 14% 41%
My life 22% 40% 14% 24%

USA
Society 36% 21% 9% 35%
My life 37% 31% 9% 22%

AI_bettersociety. Overall, do you think that generative AI will make society better or worse?
AI_bettersociety.Overall,
AI_betterpersonal. Overall,dodo you
you think
think thatthat generative
generative AI willAI will your
make make lifesociety better
better or orBase:
worse? worse? AI_betterpersonal. Overall, do you
Total
think that
sample generative
in each country ≈AI will make your life better or worse? Base: Total sample in each country ≈ 2000.
2000.

People’s expectations when asked whether generative AI will make society better or worse are
more pessimistic on average. There are about the same number of optimists, but significantly
more pessimists who believe generative AI will make society worse. Expectations around what
generative AI might mean for society are more varied across the six countries we cover. In
two (France and the UK), there are more who expect it will make society worse than better. In
another two (Denmark and the USA), there are as many pessimists as optimists. And in the
remaining two (Argentina and Japan) more respondents expect generative AI products will
make society better than expect them to make society worse.

Looking more closely at people’s expectations, both in terms of their own life and in terms
of society, younger people and people with more formal education also often opt for ‘neither

22
 WHAT DOES THE PUBLIC IN SIX COUNTRIES THINK OF GENERATIVE AI IN NEWS?

better nor worse’ or ‘don’t know’, but in most countries – Argentina being the exception – they
are more likely to answer ‘better’ (Figure 13).
Figure 13. Proportion that think generative AI will make
their 13.
Figure lives much/somewhat
Proportion better
that think generative AI will make their lives much/somewhat better
Younger people in most countries are more likely to think that generative AI will make their lives better.
Younger people in most countries are more likely to think that generative AI will make their lives better.

60%

50

Argentina
40

30
Japan
Six-country average
20
USA
Denmark
10 UK
France

0
18–24 25–34 35–44 45–54 55+

Younger people in most countries are more likely to think that generative AI will make society better.
Younger people in most countries are more likely to think that generative AI will make society better.

60%

50
Argentina

40

30 Japan
Denmark
Six-country average
20
USA
UK
10 France

0
18–24 25–34 35–44 45–54 55+

AI_bettersociety. Overall, do you think that generative AI will make society better or worse?
AI_bettersociety.Overall,
AI_betterpersonal. Overall,dodo you
you think
think thatthat generative
generative AI willAI will your
make make lifesociety better
better or orBase:
worse? worse?
18-24/25-
AI_betterpersonal.across
34/35-44/45-54/55+ Overall, do youDenmark,
Argentina, think that generative
France, Japan,AIUK,will
USAmake your life better or worse? Base: 18–24/25–34/35–44/45–
= 1272/2038/1935/2020/4952.
54/55+ across Argentina, Denmark, France, Japan, UK, USA = 1272/2038/1935/2020/4952.

Asked whether they think the use of generative AI will make different areas of life better or
worse, again, much of the public is undecided, either opting for ‘neither better nor worse’ or
answering ‘don’t know’, underlining that it is still early days.

We now look specifically at the percentage point difference between optimists who expect AI
to make things better and pessimists who expect it to make them worse gives a sense of public
expectations across different areas (Figure 14). Large parts of the public think generative AI

23
 THE REUTERS INSTITUTE FOR THE STUDY OF JOURNALISM

will make science (net ‘better’ of +44 percentage points), healthcare (+36), and many daily
routine activities, including transportation (+26), shopping (+22), and entertainment (+17),
better, even though there is much less optimism when it comes to core areas of the rule of
law, including criminal justice (+1) and more broadly legal rights and due process (-3), and
Figure 14. pessimism
considerable Net difference
for somebetween proportionissues,
very bread-and-butter that including
think cost of living (-6),
equality (-6), and job security (-18).
generative AI will make each better or worse
Figure 14.
Averaging Net
across sixdifference
countries, largebetween
parts of theproportion that think
public think generative generative
AI will AIhealthcare,
make science, will make and each better
or worse
many daily routine activities better, but more people think that generative AI will make news worse.
Averaging across six countries, large parts of the public think generative AI will make science, healthcare, and
many daily routine activities better, but more people think that generative AI will make news worse.

44

Freedom, legal rights, due

40 36

News and journalism

26
22

Cost of living

Job security
20 17 16 16
13

Equality
process
1
Healthcare

Transportation

Shopping

Entertainment

Education

Food and nutrition

Climate change and

sustainability

Crime and justice

−3
−6 −6 −8

−18

AI_betterfields. Do you think that the use of generative AI in each of the following areas will make them better or worse? Base:
Total sample across Argentina, Denmark, France, Japan, UK, USA = 12,217. Note: Figures are percentage point difference between
much/somewhat better and much/somewhat worse.
Do you think that the use of generative AI in each of the following areas will make them better or
worse? Base: Total sample across Argentina, Denmark, France, Japan, UK, USA = 12,217. Note: Figures are percentage
News
point and journalism
difference is alsobetter
between much/somewhat an and
area where, onworse.
much/somewhat balance, there is more pessimism
than optimism
(-8) – a striking contrast to another area involving the media, namely entertainment (+17).
But there is a lot of national variation here. In countries that are more optimistic about the
potential effects of generative AI, namely Argentina (+19) and Japan (+8), the proportion
that think it will make news and journalism better is larger than the proportion that think it
will become worse. The UK public are particularly negative about the effect of generative AI
on journalism, with a net score of -35. There is a similar lack of consensus across different
countries on whether crime and justice, legal rights and due process, cost of living, equality,
and job security will be made better or worse.

24
 WHAT DOES THE PUBLIC IN SIX COUNTRIES THINK OF GENERATIVE AI IN NEWS?

3. How People Think Generative AI Is Being Used by

Journalists Right Now

Many of the conversations around generative AI and journalism are about what might happen
in the future – speculation about what the technology may or may not be able to do one day,
and how this will shape the profession as we know it. But it is important to remember that some
journalists and news organisations are using generative AI right now, and they have been using
some form of AI in the newsroom for several years.

We now focus on how much the public knows about this, what they think journalists currently
use generative AI for, and what processes they think news media have in place to ensure quality.

In the survey, we showed respondents a list of journalistic tasks and asked them how often they
think journalists perform them ‘using artificial intelligence with some human oversight’. The
tasks ranged from behind-the-scenes work like ‘editing the spelling and grammar of an article’
and ‘data analysis’ through to much more audience-facing outputs like ‘writing the text of an
article’ and ‘creating a generic image/illustration to accompany the text of an article’.

We specifically asked about doing these ‘using artificial intelligence with some human
oversight’ because we know that some newsrooms are already performing at least some tasks
in this way, while few are currently doing them entirely using AI without a human in the loop.
Even tasks that may seem fanciful to some, like ‘creating an artificial presenter or author’, are
not without precedent. In Germany, for example, the popular regional newspaper Express has
created a profile for an artificial author called Klara Indernach,1 which it uses as the byline for
its articles created with the help of AI, and several news organisations across the world already
use AI-generated artificial presenters for various kinds of video and audio.

Figure 15 shows that a substantial minority of the public believe that journalists already always
or often use generative AI to complete a wide range of different tasks. Around 40% believe
that journalists often or always use AI for translation (43%), checking spelling and grammar
(43%), and data analysis (40%). Around 30% think that journalists often or always use AI for
re-versioning – whether it’s rewriting the same article for different people (28%) or turning text
into audio or video (30%) – writing headlines (29%), or creating stock images (30%).

1
https://www.express.de/autor/klara-indernach-594809

25
 THE REUTERS INSTITUTE FOR THE STUDY OF JOURNALISM
Figure 15. How often people think journalists use
generative
Figure 15. HowAI forpeople
often eachthink
of the following
journalists use generative AI for each of the following
On average across six countries, much of the public think that journalists are currently completing certain tasks
On average
‘mostly across
using six countries,
artificial much of with
intelligence the public
somethink that journalists
human oversight’.are currently completing

Always/Often Sometimes/Rarely Never Don't know

Editing the spelling and grammar of an

43% 30% 22%
article

Translation into different languages 43% 30% 23%

Data analysis 40% 31% 24%

Making charts and infographics 38% 33% 24%

Turning a written article into audio or

30% 37% 27%
video (or vice versa)

Creating a generic image/illustration to

30% 40% 25%
accompany the text of an article

Writing a headline 29% 40% 25%

Creating an image if a real photograph is

28% 40% 25%
not available

Rewriting the same article for different

28% 39% 28%
people

Writing the text of an article 27% 42% 25%

17% 42% 12% 30%

AI_news_prevalence. Thinking about news right now… How often, if at all, do you think the news media do each
AI_news_prevalence. Thinking about news right now … How often, if Base: at all,Total
do you think
sample the news media do each of the following
across
mostly using
Argentina, artificial
Denmark, intelligence
France, Japan, UK,with
USAsome human oversight? Base: Total sample across Argentina, Denmark, France, Japan, UK,
= 12,217.
USA = 12,217.

In general, the order of the tasks in Figure 15 reflects the fact that people – perhaps correctly
– believe that journalists are more likely to employ AI for behind-the-scenes work like
spellchecking and translation than they are for more audience-facing outputs. This may be
because people understand that some tasks carry a greater reputational risk for journalists, and/
or that the technology is simply better at some things than others.

The results may also reveal a degree of cynicism about journalism from some parts of the
public. The fact that around a quarter think that journalists always or often use AI to create
an image if a real photograph is not available (28%) and 17% think they create an artificial
presenter or author may say more about their attitudes towards journalism as an institution
than about how they think generative AI is actually being used. However unwelcome they might
be – and however wrong they are about how many news media use AI – these perceptions are a
social reality, shaping how parts of the public think about the intersection between journalism
and AI.

Public perceptions of what journalists and news media already use AI for are quite consistent
across different genders and age groups, but there are some differences by country, with
respondents in Argentina and the USA a little more likely to believe that AI is used for each of
these tasks, and respondents in Denmark and the UK less likely.

26
 WHAT DOES THE PUBLIC IN SIX COUNTRIES THINK OF GENERATIVE AI IN NEWS?

Among those news organisations that have decided to implement generative AI for certain
tasks, the importance of ‘having a human in the loop’ to oversee processes and check errors
is often stressed. Human oversight is nearly always mentioned in public-facing guidelines on
the use of AI for editorial work, and journalists themselves mention it frequently (Becker et al.
2024).

Large parts of the public, however, do not think this is happening (Figure 16). Averaging across
the six countries, around one-third think that human editors ‘always’ or ‘often’ check AI
outputs to make sure they are correct or of a high standard before publishing them. Nearly half
think that journalists ‘sometimes’, ‘rarely’, or ‘never’ do this – again, perhaps, reflecting a level
of cynicism about the profession among the public, or a tendency to judge the whole profession
Figure 16. How often people think human editors check
and industry on the basis of how some parts of it act.
generative AI outputs before publishing
Figure 16. How often people think human editors check generative AI outputs before publishing
On average
On average across
acrosssix
sixcountries,
countries,around one-third
around think
one third thatthat
think human editors
human always
editors or often
always checkcheck
or often generative AI
outputs to make sure they are correct or of a high standard before publishing them.
generative AI outputs to make sure they are correct or of a high standard before publishing them.

Always Often Sometimes Rarely Never Don't know

Six-country
11% 21% 29% 14% 21%
average

Japan 17% 22% 26% 8% 7% 20%

Argentina 16% 24% 25% 13% 17%

USA 11% 19% 33% 16% 17%

France 8% 20% 26% 16% 25%

Denmark 7% 21% 31% 15% 24%

UK 7% 18% 30% 16% 25%

AI_news_checking.
AI_news_checking. How How often,
often, if atifall,
at do
all,you
do think
you think
humanhuman
editorseditors check
check AI AI outputs
outputs to make
to make sure sure
they are they are
correct or of a high
correct
standardorbefore
of a high standard
publishing before
them? Base:publishing
Total samplethem? Base:
in each Total
country sample in each country ≈ 2000.
≈ 2000.

The proportion that think checking is commonplace is lowest in the UK, where only one-third
of the population say they ‘trust most news most of the time’ (Newman et al. 2023), but we also
see similarly low figures in Denmark, where trust in the news is much higher. The results may,
therefore, also partly reflect more than just people’s attitudes towards journalism and the
news media.

27
 THE REUTERS INSTITUTE FOR THE STUDY OF JOURNALISM

4. What Does the Public Think About How Journalists

Should Use Generative AI?

Various forms of AI have long been used to produce news stories by publishers including, for
example, Associated Press, Bloomberg, and Reuters. And content produced with newer forms
of generative AI has, with mixed results, been published by titles including BuzzFeed, the Los
Angeles Times, the Miami Herald, USA Today, and others.

Publishers may be more or less comfortable with how they are using these technologies to
produce various kinds of content, but our data suggest that much of the public is not – at least
not yet. As we explore in greater detail in our forthcoming 2024 Reuters Institute Digital News
Report (Newman et al. 2024), people are generally more comfortable with news produced by
human journalists than by AI.

However, averaging across six countries, younger people are significantly more likely to say they
are comfortable with using news produced in whole or in part by AI (Figure 17). The USA and
Argentina have somewhat higher levels of comfort with news made by generative AI, but there
too, much of the public remains sceptical.
Figure 17. Proportion that say they are comfortable with
news made
Figure 17. in each
Proportion thatway
say they are comfortable with news made in each way
Averaging across six countries, younger people are significantly more likely to say they are comfortable with using
news produced in whole or in part by artificial intelligence.

All 18–24

60%
61%
58%

49%

40
41%

30%

20 22% 21%

14%

Entirely by AI Mostly by AI with Mostly by a human Entirely by a human

some human journalist with some journalist
oversight help from AI

AI_news_comfort. In general, how comfortable or uncomfortable are you with using news produced in each of
AI_news_comfort.
the In general,
following ways? Base: how comfortable
Total sample/18-24 or uncomfortable
across Argentina, are youJapan,
Denmark, France, with using news
UK, USA = produced in each of the following
ways? Base: Total sample/18–24 across Argentina, Denmark, France, Japan, UK, USA = 12,217/2113.
12,217/2113.

We also asked respondents whether they are comfortable or uncomfortable using news
produced mostly by AI with some human oversight on a range of different topics. Figure 18
shows the net percentage point difference between those that selected ‘very’ or ‘somewhat’

28
 WHAT DOES THE PUBLIC IN SIX COUNTRIES THINK OF GENERATIVE AI IN NEWS?

comfortable and those that selected ‘very’ or ‘somewhat’ uncomfortable (though, as ever, a
significant minority selected the ‘neither’ or ‘don’t know’ options). Looking across different
topics, there is somewhat more comfort with using news produced mostly by AI with some
human oversight when it comes to ‘softer’ news topics, like fashion (+7) and sports (+5), than
‘hard’ news topics including politics (-33) and international affairs (-21).

But in every area, at this point in time, only for a very small number of topics are there more
people uncomfortable with relying on AI-generated news than comfortable. As with overall
comfort, there is somewhat greater acceptance of the use of AI for generating various kinds of
news with at least some human oversight in the USA and Argentina.

Putting aside country differences, there is again a marked difference between our respondents
overall and younger respondents. Among respondents overall, there are only three topic areas
out of ten where slightly more respondents are comfortable with news made mostly by AI with
some human oversight than are uncomfortable with this. Among respondents aged 18 to 24,
Figure
this rises18. Net
to six outdifference between proportion comfortable
of ten topic areas.
and uncomfortable with news on each topic being made
Figure 18. Net difference between proportion comfortable and uncomfortable with news on each
using generative
topic being made usingAIgenerative AI
Averaging
Averaging across
across six countries,
six countries, much much of theare
of the public public are uncomfortable
uncomfortable withproduced
with news being news being produced mostly by artificial
mostly
intelligence with some human oversight, but younger people are more comfortable.

All 18-24
| |
Politics −33 −20
Crime −25 −16
International news −21 −8
Local news −13 −2
Business −11 3
Celebrity or entertainment −2 8
Science and technology 0 5
Arts and culture 2 8
Sports 5 17
Fashion and beauty 7 15
−30 −20 −10 0 10 20

AI_news_comfort_topic. In general, how comfortable or uncomfortable are you with using news on each of the
AI_news_comfort_topic. In general, how comfortable or uncomfortable are Base: you with
Totalusing news on each of the following topics
sample/18-24
produced
across mostly
Argentina, by artificial
Denmark, intelligence
France, Japan, UK,with
USA =some human oversight?
12,217/2113. Base:
Note: Figures Total sample/18–24
are percentage across Argentina, Denmark,
point difference
France, very/somewhat
between Japan, UK, USA = 12,217/2113.
comfortable Note: Figures
and very/somewhat are percentage point difference between very/somewhat comfortable and
uncomfortable.
very/somewhat uncomfortable.

It is important to remember that much of the public does not have strong views either way, at
least at this stage. Between one-quarter and one-third of respondents answer either ‘neither
comfortable nor uncomfortable’ or ‘don’t know’ when asked the general questions about
comfort with different degrees of reliance on generative AI versus human journalists, and
between one-third and half of respondents do the same when asked about generative AI news
for specific topics. It is an open question as to how these less clearly formed views will evolve.

One way to assess what the public expects it will mean if and when AI comes to play a greater
role in news production is to gauge people’s views on how it will change news, compared to a
baseline of news produced entirely by human journalists.

29
 THE REUTERS INSTITUTE FOR THE STUDY OF JOURNALISM

We map this by asking respondents if they think that news produced mostly by AI with some
human oversight will differ from what most are used to across a range of different qualities
and attributes.

Between one-third and half of our respondents do not have a strong view either way. Focusing
on those respondents who do have a view, we can look at the net percentage point difference
between how many respondents think AI will make the news somewhat more or much more
(e.g. more ‘up to date’ or more ‘transparent’), versus somewhat less or much less, of each,
helping to provide an overarching picture of public expectations.

On balance, more respondents expect news produced mostly by AI with some human oversight
to be less trustworthy (-17) and less transparent (-8), but more up to date (+22) and – by a large
margin – cheaper to make (+33) (Figure 19). There is considerable national variation here, but
with the exception of Argentina, the balance of public opinion (net positive or negative) is
usually the same for these four attributes. For the others, the balance often varies.
Figure 19. Net difference between proportion that think
Figure 19. Net difference between proportion that think generative AI will make news more or
generative
less of each AI will make news more or less of each
On average across six countries the public think that the use of artificial intelligence in news production will
help publishers by cutting costs.
will help publishers by cutting costs.

33

22
Relevant to my life

20
Transparent

Trustworthy
Distinctive

8
Accurate

7
2
0
Cheaper to make

Up to date

Easier to understand

Informative

Entertaining

Unbiased

−1
−3 −3
−8

−17

AI_news_qualities. In general, do you think that news produced mostly by artificial intelligence with some human oversight is
likely to be more or less of each of the following compared to news produced entirely by a human journalist? Base: Total sample
AI_news_qualities.
acrossoversight
human Argentina, Denmark,
is likely France,
to be more Japan,
or less UK,
of each of USA = 12,217.
the following Note: Figures
compared to newsare percentage
produced point
entirely by a difference between much/
somewhat
human more Base:
journalist? and much/somewhat
Total sample acrossless.
Argentina, Denmark, France, Japan, UK, USA = 12,217. Note: Figures
are percentage point difference between much/somewhat more and much/somewhat less.

Essentially our data suggest that the public, at this stage, primarily think that the use of AI in
news production will help publishers by cutting costs, but identify few, if any, ways in which
they expect it to help them – and several key areas where many expect news made with AI to
be worse.

In light of this, it makes sense that, when asked if news produced mostly by AI with some
human oversight is more or less worth paying for than news produced entirely by a human
journalist, an average of 41% across six countries say less worth paying for (Figure 20). Just 8%
say they think that news made in this way will be more valuable.

30
 WHAT DOES THE PUBLIC IN SIX COUNTRIES THINK OF GENERATIVE AI IN NEWS?

There is some variation here by country and by age, but even among the generally more AI-
positive younger respondents aged 18–24, most say either less worth paying for (33%) or about
the same (38%). The implications of the spread of generative AI and how it is used by publishers
for people’s willingness to pay for news will be interesting to follow going forward, as tensions
may well mount between the ‘pivot to pay’ we have seen from many news media in recent years
and the views we map here.
Figure 20. Proportion that think news made mostly by AI
will
Figurebe
20.more worth
Proportion thatpaying formade mostly by AI will be more worth paying for
think news
Few people think that news produced mostly by artificial intelligence with some human oversight is more worth
paying for than news produced entirely by a human journalist.
worth paying for than news produced entirely by a human journalist.

More worth paying for About the same Don't know Less worth paying for

Six-country
8% 32% 19% 41%
average

Argentina 15% 37% 18% 31%

USA 14% 32% 14% 40%

Japan 9% 42% 21% 28%

France 24% 24% 46%

Denmark 34% 17% 46%

UK 25% 17% 56%

AI_news_pay.
AI_news_pay.
oversight is more In general,
or less worthdo you for
paying think that
than news
news produced
produced mostly
entirely by artificial
by a human intelligence
journalist? with
Base: Total some human oversight is more or
sample
less
in worth
each paying
country for than news produced entirely by a human journalist? Base: Total sample in each country ≈ 2000.
≈ 2000.

Looking across a range of different tasks that journalists and news media might use generative
AI for, and in many cases already are using generative AI for, we can again gauge how
comfortable the public is by looking at the balance between how many are comfortable with a
particular use case and how many are uncomfortable.

As with several of the questions above, about a third have no strong view either way at this
stage – but many others do. Across six countries, the balance of public opinion ranges from
relatively high levels of comfort with back-end tasks, including editing spelling and grammar
(+38), translation (+35), and the making of charts (+28), to widespread net discomfort with
synthetic content, including creating an image if a real photo is not available (-13) and artificial
presenters and authors (-24) (Figure 21).

31
 Figure 21. Net difference between
THE REUTERS proportion
INSTITUTE FOR comfortable
THE STUDY OF JOURNALISM

and uncomfortable with journalists using AI for the

Figure 21. Net difference between proportion comfortable and uncomfortable with journalists using
AI following
for the following
Averaging across
Averaging acrosssix
six countries, there
countries, there areare relatively
relatively high levels
high levels of comfort
of comfort with back-end
with back-end tasks beingtasks
done being done by AI with
some human
by AI oversight,
with some human but discomfort
oversight, with AIwith
but discomfort being usedused
AI being for synthetic media.
for synthetic media.

Creating an image if a real

Writing the text of an

presenter or author
Rewriting the same
article for different

photograph is not
38 35
28

available
21
16 15

people
20

article
9

−1 −2
Translation into different

accompany the text of an

image/illustration to
Editing the spelling and
grammar of an article

languages

infographics

Data analysis

Writing a headline

Turning a written article

Creating a generic
Making charts and

into audio or video

article
-20 −13
−24

AI_news_tasks.
AI_news_tasks. InIn general,how
general, how comfortable
comfortable ororuncomfortable
uncomfortableare are
you with each each
you with of the of
following being being produced mostly by
the following
Base: Total sample across Argentina,
artificial intelligence with some human oversight? Base: Total sample across Argentina, Denmark, France, Japan, UK, USA =
Denmark, France, Japan, UK, USA = 12,217. Note: Figures are percentage point difference between very/somewhat comfortable and very/somewhat uncomfortable.
12,217. Note: Figures are percentage point difference between very/somewhat comfortable and very/somewhat uncomfortable.

When asked if it should be disclosed or labelled as such if news has been produced mostly by
AI with some human oversight, only 5% of our respondents say none of the use cases included
above need to be disclosed, and the vast majority of respondents say they want some form of
disclosure or labelling in at least some cases. Research on the effect of labelling AI-generated
news is ongoing, but early results suggest that although labelling may be desired by audiences,
it may have a negative effect on trust (Toff and Simon 2023).

32
 WHAT DOES THE PUBLIC IN SIX COUNTRIES THINK OF GENERATIVE AI IN NEWS?
Figure 22. Proportion that think each should be labelled
as such
Figure 22. if it has been
Proportion produced
that think using
each should AI as such if it has been produced using AI
be labelled
Averaging across six countries, up to half think that some tasks should be disclosed or labelled as such if they
have been
Averaging produced
across mostlyupby
six countries, to artificial
half think intelligence
that some taskswith some
should human oversight.
be disclosed or labelled as such if
they have been produced mostly by artificial intelligence with some human oversight.

Creating an image if a real photograph is

49%
not available

Writing the text of an article 47%

Data analysis 47%

Creating an artificial presenter or author 45%

Creating a generic image/illustration to

44%
accompany the text of an article

Translation into different languages 41%

Rewriting the same article for different

41%
people

Turning a written article into audio or

40%
video (or vice versa)

Making charts and infographics 38%

Writing a headline 35%

Editing the spelling and grammar of an

32%
article

AI_news_labelling. Which, if any, of the following should be disclosed or labelled as such if it has been produced
AI_news_labelling.
mostly Which,with
by artificial intelligence if any,
someof the following
human should
oversight? Base:be disclosed
Total or labelled
sample across as such
Argentina, if it has been produced mostly
Denmark,
by artificial
France, Japan,intelligence with some human oversight? Base: Total sample across Argentina, Denmark, France, Japan, UK,
UK, USA = 12,217.
USA = 12,217.

There is, however, less consensus on what exactly should be disclosed or labelled, except for
somewhat lower expectations around the back-end tasks people are frequently comfortable
with AI completing (Figure 22). Averaging across six countries, around half say that ‘creating
an image if a real photograph is not available’ (49%), ‘writing the text of an article’ (47%),
and ‘data analysis’ (47%) should be labelled as such if generative AI is used. However, this
figure drops to around one-third for ‘editing the spelling and grammar of an article’ (32%) and
‘writing a headline’ (35%). Again, variation exists between both countries and demographic
groups that are generally more positive about AI.

33
 THE REUTERS INSTITUTE FOR THE STUDY OF JOURNALISM

Conclusion

Based on online surveys of nationally representative samples in six countries, we have, with
a particular focus on journalism and news, documented how aware people are of generative
AI, how they use it, and their expectations on the magnitude of impact it will have in different
sectors – including whether it will be used responsibly.

We find that most of the public are aware of various generative AI products, and that many
have used them, especially ChatGPT. But between 19% and 30% of the online population in
the six countries surveyed have not heard of any of the most popular generative AI tools, and
while many have tried using various of them, only a very small minority are, at this stage,
frequent users. Going forward, some use will be driven by people seeking out and using stand-
alone generative AI tools such as ChatGPT, but it seems likely that much of it will be driven
by a combination of professional adaptation, through products used in the workplace, and the
introduction of more generative AI-powered elements into platforms already widely used in
people’s private lives, including social media and search engines, as illustrated with the recent
announcements of much greater integration of generative AI into Google Search.

When it comes to public expectations around the impact of generative AI and whether these
technologies are likely to be used responsibly, we document a differentiated and nuanced
picture. First, there are sectors where people expect generative AI will have a greater impact,
and relatively fewer people expect it will be used irresponsibly (including healthcare and
science). Second, there are sectors where people expect the impact may not be as great, and
relatively fewer fear irresponsible use (including from ordinary people and retailers). Third,
there are sectors where relatively fewer people expect large impact, and relatively more people
are worried about irresponsible use (including government and political parties). Fourth, there
are sectors where more people expect large impact, and more people fear irresponsible use by
the actors involved (this includes social media and the news media).

Much of the public is still undecided on what the impact of generative AI will be. They are
unsure whether, on balance, generative AI will make their own lives and society better or
worse. This is understandable, given many are not aware of any of these products, and few
have personal experience of using them frequently. Younger people and those with higher
levels of formal education – who are also more likely to have used generative AI – are generally
more positive.

Expectations around what generative AI might mean for society are more varied across the six
countries we cover. In two, there are more who expect it will make society worse than better, in
another two, there are as many pessimists as optimists, and in the final two, more respondents
expect generative AI products will make society better than expect them to make society
worse. These differences may also partly reflect the current situation societies find themselves
in, and whether people think AI can fundamentally change the direction of those societies. To
some extent we also see this pattern reflected in how people think about AI in news. Across a
range of measures, in some countries people are generally more optimistic, but in others
more pessimistic.

34
 WHAT DOES THE PUBLIC IN SIX COUNTRIES THINK OF GENERATIVE AI IN NEWS?

Looking at journalism and news media more closely, we have found that many believe
generative AI is already relatively widely used for many different tasks, but that they are, in
most cases, not convinced these uses of AI make news better – they mostly expect it to make it
cheaper to produce.

While there is certainly curiosity, openness to new approaches, and some optimism in parts of
the public (especially when it comes to the use of these technologies in the health sector and
by scientists), generally, the role of generative AI in journalism and news media is seen quite
negatively compared to many other sectors – in some ways similar to how much of the public
sees social media companies. Basically, we find that the public primarily think that the use of
generative AI in news production will help publishers cut costs, but identify few, if any, ways in
which they expect it to help them as audiences, and several key areas where many expect news
made with AI to be worse.

These views are not solely informed by how people think generative AI will impact journalism
in the future. A substantial minority of the public believe that journalists already always
or often use generative AI to complete a wide range of different tasks. Some of these are
tasks that most are comfortable with, and are within the current capabilities of generative
AI, like checking spelling and grammar. But many others are not. More than half of our
respondents believe that news media at least sometimes use generative AI to create images
if no real photographs are available, and as many believe that news media at least sometimes
create artificial authors or presenters. These are forms of use that much of the public are
uncomfortable with.

Every individual journalist and every news organisation will need to make their own decisions
about which, if any, uses of generative AI they believe are right for them, given their editorial
principles and their practical imperatives. Public opinion cannot – and arguably should not
– dictate these decisions. But public opinion provides a guide on which uses are likely to
influence how people judge the quality of news and their comfort with relying on it, and thus
helps, among other things, to identify areas where it is particularly important for journalists
and news media to communicate and explain their use of AI to their target audience.

It is still early days, and it remains to be seen how public use and perception of generative AI in
general, and its role in journalism and news specifically, will evolve. On many of the questions
asking respondents to evaluate AI in different sectors and for different uses, between roughly
a quarter and half of respondents pick relatively neutral middle options or answer ‘don’t
know’. There is still much uncertainty around what role generative AI should and will have, in
different sectors, and for different purposes. And, especially in light of how many have limited
personal experience of using these products, it makes sense that much of the public has not
made up their minds.

Public debate, opinion commentary, and news coverage will be among the factors influencing
how this evolves. So will people’s own experience of using generative AI products, whether
for private or professional purposes. Here, it is important to note two things. First, younger
respondents generally are much more open to, and in many cases optimistic about, generative
AI than respondents overall. Second, despite the many documented limitations and problems

35
 THE REUTERS INSTITUTE FOR THE STUDY OF JOURNALISM

with state-of-the-art generative AI products, those respondents who use these tools
themselves tend to offer a reasonably positive assessment of how well they work, and how
much they trust them. This does not necessarily mean that future adopters will feel the same.
But if they do, and use becomes widespread and routine, overall public opinion will change – in
some cases perhaps towards a more pessimistic view, but, at least if our data are anything to go
by, in a more grounded and cautiously optimistic direction.

36
 WHAT DOES THE PUBLIC IN SIX COUNTRIES THINK OF GENERATIVE AI IN NEWS?

References

Ada Lovelace Institute and The Alan Turing Institute. 2023. How Do People Feel About AI? A
Nationally Representative Survey of Public Attitudes to Artificial Intelligence in Britain.
https://adalovelaceinstitute.org/ report/public-attitudes-ai.

Alm, C. O., Alvarez, A., Font, J., Liapis, A., Pederson, T., Salo, J. 2020. ‘Invisible AI-driven HCI
Systems – When, Why and How’, Proceedings of the 11th Nordic Conference on Human-Computer
Interaction: Shaping Experiences, Shaping Society, 1–3.
https://doi.org/10.1145/3419249.3420099.

Angwin, J., Nelson, A., Palta, R. 2024. ‘Seeking Reliable Election Information? Don’t Trust AI’,
Proof News. https://www.proofnews.org/seeking-electioninformation-dont-trust-ai/.

Becker, K. B., Simon, F. M., Crum, C. 2023. ‘Policies in Parallel? A Comparative Study of
Journalistic AI Policies in 52 Global News Organisations’, https://doi.org/10.31235/osf.io/c4af9.

Beckett, C., Yaseen, M. 2023. ‘Generating Change: A Global Survey of What News Organisations
Are Doing with Artificial Intelligence’. London: JournalismAI, London School of Economics.
https://www.journalismai.info/research/2023-generating-change.

Bender, E. M., Gebru, T., McMillan-Major, A., Shmitchell, S. 2021. ‘On the Dangers of Stochastic
Parrots: Can Language Models Be Too Big?’, in Proceedings of the 2021 ACM Conference on
Fairness, Accountability, and Transparency, 610–23. New York: Association for Computing
Machinery. https://doi.org/10.1145/3442188.3445922.

Broussard, M. 2018. Artificial Unintelligence: How Computers Misunderstand the World. Reprint
edition. Cambridge, MA: The MIT Press.

Broussard, M. 2023. More than a Glitch: Confronting Race, Gender, and Ability Bias in Tech.
Cambridge, MA: The MIT Press.

Caswell, D. 2024. ‘AI in Journalism Challenge 2023.’ London: Open Society Foundations.
https://www.opensocietyfoundations.org/publications/open-society-s-applied-ai-in-
journalism-challenge.

Diakopoulos, N. 2019. Automating the News: How Algorithms Are Rewriting the Media. Cambridge,
MA: Harvard University Press.

Diakopoulos, N., Cools, H., Li, C., Helberger, N., Kung, E., Rinehart, A., Gibbs, L. 2024. ‘Generative
AI in Journalism: The Evolution of Newswork and Ethics in a Generative Information
Ecosystem’. New York: Associated Press. https://doi.org/10.13140/RG.2.2.31540.05765.

Fletcher, R. 2024. How Many News Websites Block AI Crawlers? Reuters Institute for the Study of
Journalism. https://doi.org/10.60625/risj-xm9g-ws87.

37
 THE REUTERS INSTITUTE FOR THE STUDY OF JOURNALISM

Fletcher, R., Adami, M., Nielsen, R. K. 2024. ‘I’m Unable To’: How Generative AI Chatbots Respond
When Asked for the Latest News. Reuters Institute for the Study of Journalism.
https://doi.org/10.60625/RISJ-HBNY-N953.

Humprecht, E., Herrero, L. C., Blassnig, S., Brüggemann, M., Engesser, S. 2022. ‘Media Systems
in the Digital Age: An Empirical Comparison of 30 Countries’, Journal of Communication 72(2):
145–64. https://doi.org/10.1093/joc/jqab054.

Mellado, C., Cruz, A., Dodds, T. 2024. Inteligencia Artificial y Audiencias en Chile.
https://www.noticiasyperiodismo.cl/audiencias-e-inteligencia-artificial.

Newman, N. 2024. Journalism, Media, and Technology Trends and Predictions 2024. Reuters
Institute for the Study of Journalism. https://doi.org/10.60625/risj-0s9w-z770.

Newman, N., Fletcher, R., Eddy, K., Robertson, C. T., Nielsen, R. K. 2023. Reuters Institute Digital
News Report 2023. Reuters Institute for the Study of Journalism.
https://doi.org/10.60625/risj-p6es-hb13.

Newman, N., Fletcher, R., Robertson C. T., Ross Arguedas, A. A., Nielsen, R. K. 2024. Reuters
Institute Digital News Report 2024 (forthcoming). Reuters Institute for the Study of Journalism.

Nielsen, R. K. 2024. ‘How the News Ecosystem Might Look like in the Age of Generative AI.’
https://reutersinstitute.politics.ox.ac.uk/news/how-news-ecosystem-might-look-age-
generative-ai.

Nielsen, R. K., Fletcher, R. 2023. ‘Comparing the Platformization of News Media Systems: A
Cross-Country Analysis’, European Journal of Communication 38(5): 484–99.
https://doi.org/10.1177/02673231231189043.

Pew. 2023. ‘Growing Public Concern about the Role of Artificial Intelligence in Daily Life’.
https://www.pewresearch.org/short-reads/2023/08/28/growing-public-concern-about-the-
role-of-artificial-intelligence-in-daily-life/.

Pew. 2024. ‘Americans’ Use of ChatGPT is Ticking Up, but Few Trust its Election Information’.
https://www.pewresearch.org/short-reads/2024/03/26/americans-use-of-chatgpt-is-ticking-
up-but-few-trust-its-election-information/.

Simon, F. M. 2024. ‘Artificial Intelligence in the News: How AI Retools, Rationalizes, and
Reshapes Journalism and the Public Arena’, Columbia Journalism Review.
https://www.cjr.org/tow_center_reports/artificial-intelligence-in-the-news.php/.

Toff, B., Simon, F. M. 2023. ‘Or They Could Just Not Use It?’: The Paradox of AI Disclosure for
Audience Trust in News. https://doi.org/10.31235/osf.io/mdvak.

Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., Polosukhin,
I. 2017. Attention Is All You Need. https://doi.org/10.48550/ARXIV.1706.03762.

38
 WHAT DOES THE PUBLIC IN SIX COUNTRIES THINK OF GENERATIVE AI IN NEWS?

Vogler, D., Eisenegger, M., Fürst, S., Udris, L., Ryffel, Q., Rivière, M., Schäfer, M. S. 2023.
Künstliche Intelligenz in der journalistischen Nachrichtenproduktion: Jahrbuch Qualität der
Medien Studie 1 / 2023. https://doi.org/10.5167/UZH-238634.

RISJ PUBLICATIONS

SELECTED BOOKS
Avoiding the News: Reluctant Audiences for Thomas Hanitzsch, Folker Hanusch, Jyotika
Journalism Ramaprasad, and Arnold S. de Beer (eds)
Benjamin Toff, Ruth Palmer, and Rasmus (published with Columbia University Press)
Kleis Nielsen (published with Columbia
University Press) NGOs as Newsmakers: The Changing Landscape of
International News
Hearts and Minds: Harnessing Leadership, Culture, Matthew Powers (published with Columbia
and Talent to Really Go Digital University Press)
Lucy Kueng
Global Teamwork: The Rise of Collaboration in
Worlds of Journalism: Journalistic Cultures Around Investigative Journalism
the Globe Richard Sambrook (ed)

SELECTED RISJ REPORTS AND FACTSHEETS

‘I’m Unable to’: How Generative AI Chatbots Paying for News: Price-Conscious Consumers Look
Respond when Asked for the Latest News for Value amid Cost-of-Living Crisis
Richard Fletcher, Marina Adami, and Rasmus Nic Newman and Craig Robertson
Kleis Nielsen (factsheet)
Strategies for Building Trust in News: What the
Race and Leadership in the News Media 2024: Public Say They Want Across Four Countries
Evidence from Five Markets Sayan Banerjee, Camila Mont’Alverne, Amy Ross
Amy Ross Arguedas, Mitali Mukherjee, and Arguedas, Benjamin Toff, Richard Fletcher, and
Rasmus Kleis Nielsen (factsheet) Rasmus Kleis Nielsen

Women and Leadership in the News Media 2024: Reuters Institute Digital News Report 2023
Evidence from 12 Markets Nic Newman, Richard Fletcher, Kirsten Eddy,
Amy Ross Arguedas, Mitali Mukherjee, and Craig T. Robertson, and Rasmus Kleis Nielsen
Rasmus Kleis Nielsen (factsheet)
News for the Powerful and Privileged: How
How Many News Websites Block AI Crawlers Misrepresentation and Underrepresentation of
Richard Fletcher (factsheet) Disadvantaged Communities Undermines Their
Trust in News
Journalism, Media and Technology Trends and Amy Ross Arguedas, Sayan Banerjee, Camila
Predictions 2024 Mont’Alverne, Benjamin Toff, Richard Fletcher,
Nic Newman and Rasmus Kleis Nielsen

Changing Newsrooms 2023: Media Leaders Struggle How Publishers are Learning to Create and
to Embrace Diversity in Full and Remain Cautious on Distribute News on TikTok
AI Disruption Nic Newman
Federica Cherubini and Ramaa Sharma
How We Follow Climate Change: Climate News Use
Climate Change News Audiences: Analysis of News and Attitudes in Eight Countries
Use and Attitudes in Eight Countries Waqas Ejaz, Mitali Mukherjee, Richard Fletcher,
Waqas Ejaz, Mitali Mukherjee, and Richard Rasmus Kleis Nielsen
Fletcher

39
 With support from

SAMPLE

https://reutersinstitute.politics.ox.ac.uk
https://reutersinstitute.politics.ox.ac.uk