People in the UK are generally pessimistic about the potential impact of AI on journalism – but are more open to it when given examples of specific tasks it could be used for.
A Yougov survey of 2,046 adults carried out in January and shared with Press Gazette found that 48% of people fear AI will have a negative effect on journalism.
Just 6% thought it would do more good than bad – versus 36% who fear it will do more bad. Almost a quarter of people (23%) remain unsure.
Justin Marshall, head of digital, media and technology at Yougov, told Press Gazette: “There is a general perception that people think this could be a bad thing – that is something that news organisations are definitely going to have to take note of.”
Related
Yougov carried out the data to get a better understanding of how the public actually feels about the introduction of AI.
Marshall said: “AI is a hot topic in general last year and this year. There’s a lot of talk about it but we wanted to start digging into what it means more specifically to different industries or different groups of people – drilling down into what it’s going to potentially be used for and how people might react to that.”
Content from our partners
The data indicated that younger people are much more open to the use of AI. Among 18 to 24-year-olds, 17% felt it would do more good than bad to journalism versus 8% of 25 to 49-year-olds, 2% of 50 to 64-year-olds and 3% of over-65s.
"There is more nuance by age, so 18-24s are slightly more positive towards the use of AI in journalism whereas the over-65s are much less so," Marshall said.
"The younger generation has grown up with more technology and are a bit more open to it, but the majority of them are still against it," he said, referring to the 32% of 18 to 24s who think AI will be a bad thing.
More 18 to 24s (34%) took a neutral view that AI will do both bad and good compared to 21% or 23% for other age groups.
Despite the apprehension shown by many, just 7% said they have read a news article written by AI to their best of their knowledge.
Just over half (53%) said they weren't sure if they had, while 40% said they definitely had not.
Marshall explained that many people were more amenable to the use of AI for several specific journalism tasks - with the primary exception of in-depth writing.
He said: "A lot of the articles that media are writing on AI often have a negative angle, which may be influencing the public's perception that ‘AI is all bad anyway... it's going to be bad for journalism because it's potentially going to be bad for everything.’
"But then when you actually ask what journalism tasks is it acceptable for AI to do, when you give them a specific thing, some of them think, ‘yeah, actually that might be okay. I'd be happy for newsbrands to use AI for this task'."
For example, 73% of people were happy with AI performing spelling and grammar edits on an article, 67% of people were open to AI being used to translate articles into other languages, 51% said it was acceptable for AI to perform analysis of datasets for journalists to write about, 51% were open to it suggesting interview questions for journalists to ask someone and also to it suggesting article topics, and 50% were happy with it generating summaries of articles.
Many of these are frequently automated tasks already, Marshall noted, saying people are "more familiar with spell checks and things like Google Translate so language checks are often part of their day-to-day anyway".
Several publishers are already using AI to translate content directly for audiences to consume. In 2022 Bloomberg Media started using an AI firm to translate its news videos into Spanish to reach a new Spanish-speaking audience base in the US and Latin America.
Some publishers have begun using AI-generated summaries at the top of articles. For example Nottinghamshire Live editor Natalie Fahy told readers last summer the Reach website was trialling the use of bullet points at the top of articles to "help you get a sense of what you're about to read and improve the experience for you". Although these are generated by AI, they are always checked by a human editor before publication, she added.
Swedish publisher Aftonbladet discovered to its surprise last year that readers spent more time with articles with AI-generated summaries. Deputy editor Martin Schori told Press Gazette that because readers get a more general understanding of an article upfront, they are more likely to go on and read the whole thing.
Other areas where people were open to the use of AI were choosing an image to put in an article from stock photo websites (49%) and fact-checking claims made in a draft by a human journalist (46%).
The tasks deemed most unacceptable for AI were conducting interviews on behalf of journalists (70% said unacceptable), conducting research on a topic for a journalist to use (43% said this was unacceptable versus 39% who would be okay with it) and writing article headlines (42% unacceptable versus 41% acceptable).
"The other theme within this is that the more in-depth the article is, the less acceptable people think it is for AI to do," Marshall explained.
For example, people are much happier for AI to write and publish "short routine data-driven stories such as financial reports, sports scores or weather updates" (48% acceptable, versus 35% unacceptable) than to write and publish "short articles reporting the content of press releases put out by a company/organisation/government and drawing in simple additional context relevant to the story" (52% against versus 26% on board).
This trend continues when people were asked about how they would feel if AI wrote and published "long articles that investigate an issue in depth that would typically include expert or eyewitness interviews" with 67% against this and 16% saying it was acceptable.
Younger people are more willing to consider this though, with 8% of 18 to 24 year-olds saying it is completely acceptable and 22% saying it is somewhat acceptable, versus 2% and 11% of 65 and overs.
The crucial factor across all age groups was that most agreed media organisations should be required to display any ways in which a news article has been assisted by AI. In total 79% backed this, although there was again a split across age groups with 17% of 18 to 24-year-olds thinking this was unnecessary versus 5% of respondents aged 65+.
Marshall said: "Where AI has been used to produce an article, people generally think that this should be flagged or noted on the article just so people are aware. This might help trust in newsbrands that are using this technology as long as it's signposted."
How use of AI affects trust in news
Yougov also investigated the extent to which the use of AI affects trust in news.
It conducted a randomised test asking people to what extent they would trust an article on a randomly-selected news organisation's website that had been written and overseen by a random combination of human journalists, AI journalists, human editors and AI editors.
The overall theme of the test, Marshall said, was that "having an AI journalist or AI editor reduces trust, and if both roles are AI, then the trust is lower again".
For example, when people saw both the journalist and editor in question were human, 48% said they would have a great deal or fair amount of trust in the article. The change in news outlet shown to respondents varied trust only to the extent that there was more trust in broadcast/"upmarket" newspaper websites than mid-market and tabloid sites.
Trust was lower when one of the key roles involved was done by AI: trust was 25% when AI was the journalist and 23% when it was the editor. Where AI fulfilled both roles, trust was 13%.
Email pged@pressgazette.co.uk to point out mistakes, provide story tips or send in a letter for publication on our "Letters Page" blog
