AI is not just coming to newsrooms in Ireland, it’s already here

Mediahuis – which owns the Irish Independent, among other titles in Ireland – is experimenting to find out “at what level we can automate first line news”, Ana Jakimovska, its head of AI strategy, reportedly said on stage at a conference in London. 

Reach Plc – which owns DublinLive, the Irish Mirror and other titles in Ireland – has built its own AI tool, Guten, for its journalists to use, and roughly half of them in Ireland do, says Michael Marsh, Reach’s editor in chief for Ireland.

Breakingnews.ie, a website owned by the Irish Times, says it is experimenting “to see how Generative AI (GenAI) tools can support our production processes”.

A journalist at The Journal has been using Claude to help translate his articles from Irish into English, says managing editor Susan Daly.

AI is not just coming to journalism in Ireland, it’s here, in newsrooms, in use by journalists for things like recording and transcribing interviews, searching through large troves of documents, translation, polishing the writing in articles, suggesting headlines, speeding up the publishing process, and more. 

AI tools can make journalists’ jobs easier, and give them new capabilities, but their use damages the environment, can lead to mistakes that hurt news organisations’ credibility, and can help companies employ fewer journalists, journalists say.

“Different news organisations are embracing it at different speeds,” says National Union of Journalists (NUJ) Irish organiser Ian McGuinness. “Some aren’t even touching it yet. There's some, what you would call early adopters.”

“I'd be amazed if it’s not used in some form everywhere now,” said Marsh, the editor in chief at Reach Ireland.

Early adopter: Reach

One of the media companies leaning into the use of AI is Reach Plc. 

The UK-based company with “120 brands” has titles in Ireland including DublinLive, the Irish Daily Star, and the Irish Daily Mirror.

“As part of their mission to free up journalists to focus on high-value and impactful journalism, Reach’s Data Science team developed Guten,” according to a July 2025 post “co-authored by Lewis James, Senior Data Scientist at Reach plc and Dan Taffler, Group Director of Data and Analytics at Reach plc”.  

“Guten optimises every part of Reach’s editorial workflow including ingestion of news wires, wire article suggestions, article idea recommendations, and content generation,” they wrote. 

“For Reach’s Content Hub, a dedicated team of journalists produce traffic-driving content for multiple national and regional brands,” they wrote.

“Guten provides the ability to take a piece of Reach content and redraft it in the style and tone of their other publications without needing to rewrite every version of the same article from scratch,” they wrote.

Indeed, going big on AI is a core part of the company’s strategy. 

In the company’s most recent annual report, in a note dated 3 March 2026, Reach Plc chairman Nick Prettejohn writes that the company’s “three key priorities are Connecting with Audiences, Accelerating the use of Tech and AI and Diversifying Revenues”.

The annual report sits in the “investor” section of the company’s website, suggesting that that is its main audience. So what is the day-to-day reality of the use of AI in Reach newsrooms in Ireland?

In Reach’s Ireland operation, Guten is “one of our kind of approved AI tools”, said Marsh, editor in chief for Ireland, on a video call on 29 April.  

“I don't care if people use it or not,” Marsh said. “We don't say, right, you have to do eight stories a week using Guten, you know, there's nothing like that.”

“Some people might use it a couple times a day. Some people might use it once a month. It just depends, really,” he said. “One of the benefits of it is it can help get things into our CMS [content management system] really quickly.”

So if someone at Reach in Ireland wants to take an article from another Reach publication, or the Press Association (PA) newswire, and publish it on a Reach website, it takes some of the “heavy lifting” out of it, Marsh said.

If they’re looking at their personal dashboard, which would include recommendations based on what they’ve been interested in the past, “there might be something around, alright, okay, this is an interesting story from PA”, he said. “You can then hit a button that’ll then upload into the CMS, words, pictures, as it appears on PA.”

“You then can ask it to rewrite or tweak certain elements of the copy or the headline on that, but generally speaking, most people would just put it in without the rewrite option, and then you can then edit within our CMS,” he said.

Whatever role AI takes in creating an article, though, before it’s published, “somebody would have to read, go, right, I'm happy with this, and then press the button”, Marsh said. 

“It would be up to the journalists, the individuals, to kind of do the checks and balances, kind of in the old-school way,” he said. “It's up to the person pressing the button is responsible for, you know, making sure that it's, it's legally sound, it's accurate, it complies with the Press Council.”

Journalists at Reach also use other AI tools, Marsh said. “Don't ask me to name them, because it's a very long list, but we have an approved AI list,” he said. But the big one besides Guten is Google’s Gemini, he said. 

“Google’s Gemini ‘AI assistant’ product is now available for all colleagues, with over 40% of our people actively using it,” according to a 3 March note in the company’s most recent annual report from Reach Plc chief executive Piers North.

People at Reach in Ireland might use Gemini to summarise long council reports to pull out the six key points, Marsh said. Or to help them come up with article ideas, he said. 

“So, if I went in now and said, give me a story suggestion about Temple Bar that is likely to resonate with people aged 18 to 24 and that is a fit with a DublinLive audience, it would give you, Oh here’s three ideas that might work,” he said. 

A lot of the benefit of all this is speed. Reach publishes tens of thousands of articles each month in Ireland alone, Marsh said. 

So shaving some time off the monotonous tasks associated with each of those – downloading photos from one place and uploading them to another, that kind of thing – can make a big difference, and free up journalists to do more stuff they’d like better. 

“People want to be interviewing. They want to be on the road. They want to be maybe doing a good long read on whatever it may be,” he said.

Is that really what happens though, does the use of AI really free up journalists to go out and interview people, and write long reads?

“It depends on the role,” Marsh says. “Some journalists, we allow – that depends on their role and their expertise. Others are in roles where they're more desk based.”

In a small online survey I ran in March, asking journalists working in Ireland about their use of AI, a respondent said they had worked in a Reach newsroom. Their account chimes, somewhat, with Marsh’s.

“A typical shift would consist of churning out 10/12 low effort local stories, with a significant amount being 'brought over' from other Reach sites in UK and Ireland,” they said. 

“You copy a link from another Reach site's story (one that has been noticed by the editor as doing well for views) into Guten and it will re-produce it with suggested tweaks to optimise SEO variables and article tone,” they said.

Could AI have done their job? “That job, yes, probably because much of the work was regurgitating Facebook posts and adding hundreds of words from other articles so that a short two sentence Facebook post about a café opening could be 400 words – long enough to reach ad optimisation targets.”  

Early adopter? Mediahuis

Mediahuis has also positioned itself as an early adopter, leaning into the use of AI.

It’s based in Belgium and has titles in Ireland including the Independent, Sunday Independent, Sunday World, and more.    

In her presentation at the News in the Digital Age 2026 conference in London in February, Jakimovska, Mediahuis’s head of AI strategy, talked about how the company could use AI. 

It could create a huge database – newswires, social media, other sources – and then employ a series of agents to create content, she said, according to notes on the presentation made by a member of the team at FT Strategies, which co-hosted the conference with the Google Digital News Initiative. 

This seems to support reporting in the Press Gazette about Jakimovska’s speech at the conference, which says that “Under an experimental project AI agents are involved in the writing, fact checking, legal checking and editing of news stories under the plan but a human would always check the finished product before it is published.” 

Then, according to the notes on the event, a monitoring agent would highlight readers’ reactions to the breaking news to build “signature content”. 

Or, as the Press Gazette quoted Jakimovska as saying, “When the polarisation becomes high, or when there’s different opinions, it triggers an opinion skeleton to send to an editor to say ‘this might be a really interesting thing to cover for a [piece of] signature journalism’.” 

Basically, it appears to be a vision of AI taking over the work of writing the kinds of articles that some journalists, sitting at desks all day in Dublin, are expected to write five to eight of, from online sources. 

This, the pitch goes, would free up journalists to instead leave their desks and go out into the world to write more in-depth, “signature journalism”, knocking on doors, reporting from the scene, and doing other things that AI can’t do (yet). 

In response to a 16 March request for an interview about how Mediahuis uses journalism now, and its future plans, Jakimovska on 20 March suggested a couple of possible days: “I have some availability next Wed or Thursday if that works for you?”

But then she became unavailable. Perhaps connected, at about the same time, unbeknownst to me, on 19 March, the newspaper NRC in the Netherlands had reported that Peter Vandermeersch had messed up while using AI. 

Vandermeersch, who was chief executive of Mediahuis Ireland from 2022 to late 2025, had then been appointed the company’s Media and Society Fellow. He was writing a Substack newsletter called Media and Society.

But, as Vandermeersch put it, while writing on 19 March about the mistakes in his work that NRC had revealed, “I summarised reports using AI tools and worked from those summaries, trusting they were accurate. In doing so, I wrongly put words into people’s mouths, when I should have presented them as paraphrases.”

“It is particularly painful that I made precisely the mistake I have repeatedly warned colleagues about: these language models are so good that they produce irresistible quotes you are tempted to use as an author,” he wrote. “Of course, I should have verified them. The necessary ‘human oversight,’ which I consistently advocate, fell short.” 

A Mediahuis Ireland spokesperson also didn’t respond to requests in mid-April for an interview with someone in leadership here, about the company’s use of AI in Ireland. 

In any case, there appears to be some distance between the aspirations and experiments Jakimovska talked about at the conference, and the reality in Mediahuis newsrooms in Ireland. 

In my small online survey in March, two responses were from people who said they write articles for Mediahuis titles.

One said their employer did not allow them to use AI, and they did not use it. The other said they are allowed to use AI in their work but they do not. 

Anything else to add, the survey asked? “Fuck AI,” one of the respondents added.

Slower movers: The Journal

The Journal, based in Ireland and owned primarily by brothers Brian and Eamon Fallon, has not moved quickly to formally adopt the use of AI in its newsroom in Dublin. 

Conversations about AI, and whether it was going to take journalists’ jobs, have been floating around the industry for a long time, says managing editor Susan Daly. 

But recently, there’s been a “tipping point”, with people integrating AI into their lives, and “other, I suppose parts of the industry adopting it very, very quickly”, Daly said.

“That makes it kind of a conversation about, have we gotten past the point of talking about whether we use AI at all, or is it now table stakes, and how do we use AI?” she said via video call. 

“I think there's probably the ethical discussion about the use of AI at all, which – the environmental impacts, I think about a lot, for example – I think that that ship has sailed for a lot of journalists,” she said.

The team at The Journal hadn’t had “any comprehensive discussion about it internally”, she said on 14 May. “But certainly, a month ago, we had kick started a process which was to have an objective, to have a very specific AI usage policy for content.”

She wants to make sure the approach they settle on is “a very truthful approach for us, that we’re not trying to get anyone ringfenced by something that they don’t believe in, either the use of or the non-use of”.

She wants to guard against “reputational risk”, and ensure that everything The Journal publishes is up to its standards, Daly said. 

“And also look at, okay, are we looking at opportunities here as well to improve the offering for readers in a way that makes sense to us and them,” she said. 

Not speaking for The Journal as an organisation, or for its newsroom, Daly said that perhaps there is an opportunity for journalism in AI.

“We – like the media industry, and journalists – have been just so restrained by resource scarcity that we are finding it difficult to invest resources into the journalism that's really deep and that's really helpful to people,” she said. 

The Journal has its investigations, and its fact checks, Daly said. “We’ve had to be really thoughtful about how we resource those and funds we pursue for those things.”

So maybe the use of AI could help small teams of journalists dig more effectively into huge troves of documents, like the Epstein files released by the US Department of Justice, Daly said. “There’s loads of examples of, kind of, data journalism that can be improved by the use of an AI tool.”

“So I think we’re really open to all of that,” Daly said. But she is just at the very start of a process of consulting with the journalists at The Journal, and nothing’s been decided yet, she said. 

Why has The Journal been more cautious than other media companies about adopting AI? 

“If journalism is the core, leading thing for you, then the issue is that AI, we can see the huge risk to journalism in it, and we can see the huge risk to information integrity for the public,” Daly said. 

“Therefore we're going to have a more cautious and thoughtful approach to it than somewhere else where it’s kind of stack ’em high, sell ’em cheap kind of a thing – and that’s no shade on anyone,” she said.

On 5 May, not long before this interview, The Journal had published an article that included at the bottom the text “Sonnet 4.6 Claud is AI and can make mistakes. Please double-check responses”, which someone noticed and posted about on Reddit.

That bit of text was quickly removed, and a clarification appended to the article, saying that the original article had been written by the journalist, Concubhar Ó Liatháin, in Irish, and then Claude had been used to translate it into English. 

“In March, I suppose, he started using Claude, and he explained to us it was giving good context versus other translations,” Daly said on 14 May. They put a process in place where the reporter would read over the translation, and it would also be reviewed by a human editor, “so two, two sets of human eyes afterwards”. 

But there was an update to the story, “later in the evening”, and the new version was translated again, “and an editor didn’t see it, and that little line that indicates that Claude had been used was left in”. 

They weren’t labelling the English versions of Ó Liatháin’s articles as having been translated with the help of AI, because they hadn’t finished their process of making an AI policy and which will include, “how do we talk about it to our audience”, Daly said.

But now they are labelling his articles. The English translation of his article from 19 May includes a line at the bottom saying, “This article was originally written in the reporter’s native Irish and has been translated into English here. AI was used as part of the translation process before final edits.”  

Slower mover: The Irish Times

The Irish Times is, according to the Media Ownership Monitor: Ireland, from DCU, “operated by The Irish Times Group DAC which is ultimately owned by The Irish Times Trust CLG”.

The editor of the Irish Times, Ruadhan Mac Cormaic, in an interview with Alan English for The Currency, published on 15 May, talked about his approach to AI at his paper.

“At the moment, I’m happy not to be first mover. I’m happy to observe what’s happening,” Mac Cormaic is quoted as saying. 

“That’s not to say we’re not experimenting, that we’re not looking at it all the time. We really are,” he said. “If AI can help us to free up journalists to work on the things journalists can do, but that AI can’t, I’m all for that.”

He talked about trying to create “an intelligence search engine that you can interact with in a Q&A format” to explore the Irish Times archive, “that delivers you answers and then provides you with the links to the original pages”.

“The other thing I’m quite excited about is its potential to unlock new forms of investigative journalism for us,” he said. “AI can certainly give us access to data sets that we couldn’t otherwise access in any practical way. And we’re already playing around with that in our investigations.” 

This is not to say AI is not being used by some journalists at the Irish Times. One of them said recently that they’d been sent a video on how they could use AI.

Breakingnews.ie, which “is owned and operated by The Irish Times Group”, according to the Media Ownership Monitor, has been quite public about its use of AI.

There’s a page on its website, although it’s not clear when it was published, that says “In recent months, we’ve been carrying out a pilot project to see how Generative AI (GenAI) tools can support our production processes.”

This involves, it says, “taking a selection of articles and requesting that an AI tool offer several headline suggestions and SEO tag suggestions. These are then reviewed by journalists who ultimately create the final headline and tags.”

“We’ll be gathering data on how well the tools perform, where they fall short, and the production benefits they deliver,” it says.

Slow movers: Dublin Inquirer

It would hardly be fair to dig into what all these other news organisations are doing with AI and not tell you what we are doing with it ourselves. 

We do not have any in-house tools or official AI policy. Some of us – including me – do use AI tools in our work. 

For example, I recorded the interviews I did for this article on Otter.ai, which automatically transcribed the audio. 

There are lots of mistakes in the transcripts, but even still, it’s a lot easier to search through the text – by reading it, or using the search bar – to find the bits I want to use, than it would be to search through an audio-only file. 

Then I can easily listen to the audio of just the few seconds or minutes I want to use, and properly transcribe those to use in my article. 

Also, I find Google less and less useful, so I also use Claude, and Kagi, to help me research things on the web, find the sources I am looking for, and give me links to them. 

I have also used Claude to write complex (to me) formulas to use in Excel or Google Sheets to help me clean and/or analyse big data sets. 

I once used Microsoft Copilot to generate an image for an article by Michael Lanigan about Dublin City Council moving towards deploying AI tools – to demonstrate its capabilities.

And I used Claude Code to “vibecode” some updates to our councilvotetracker.ie website, to get it back online after a yearslong hiatus, and expand it so it can now be used to track votes on any council – not just its original Dublin City Council.

I’ve also had a freelancer whose first language was not English tell me they used Grammarly to help polish up their article before submitting it to me.

Lois Kapila here also uses Claude for research, particularly to help parse company documents, and has used it to translate documents from other languages into English so she can understand them for her reporting.

Other members of the team here do not use AI at all, they tell me.   

Jobs, jobs, jobs

AI has the potential to be “labour saving”, said Ian McGuiness, the NUJ organiser, in a 30 April video call.  

But isn’t “labour saving” just a way to say it’s a way of not hiring someone, or making someone redundant? 

“We don't mind labour saving if it's not replacing journalists,” McGuinness said. “In other words, if it's not a), making journalists redundant, b), essentially taking jobs of freelancers.”

“If it's used to try and replace journalists, and it has been used to replace – there have been redundancies in places like Reach, which is using AI, and we're absolutely convinced that that's no coincidence – if it's used to replace journalists, that's a danger,” McGuinness said. 

The average number of people employed by Reach overall in “Production and editorial” fell from 2,587 in 2024 to 2,494 in 2025, according to its most recent annual report. RTÉ and other media reported in September on potential job losses in Ireland.

Not in response to these particular numbers, but in general, does Reach Plc’s adoption of AI allow it to employ fewer people and have lower costs? 

“I wouldn’t say it was allowing us to – I wouldn’t say that,” said Marsh, the editor in chief for Reach in Ireland. 

“The industry is challenged bigtime at the moment,” he said. “I wouldn’t say our newsrooms are smaller because of AI as such, I think there’s multiple factors that play into that, if I’m being honest.”

In any case, for journalists in newsrooms in Ireland, the introduction of AI can make life harder and more unpleasant, said McGuinness. 

“You start off with a newsroom which was already under pressure. You then have a certain number of redundancies,” he says. Then the people who are left behind have to do their own job, plus the work of the people who’ve left, he says. 

“And then you add in AI. I mean, those people are struggling to keep their head above water as it is doing their non-AI jobs, and now they have to oversee AI, which sometimes hallucinates or writes absolute garbage, so they have to rewrite it,” he says. 

Credibility

Beyond the potential for companies to employ fewer journalists – not hiring new ones, or cutting existing ones – and make the jobs of remaining journalists worse, the use of AI also has the potential to undermine readers’ trust in media, McGuinness says.

“The problem with AI is, you know, people have said this to me, you know, friends of mine said this to me, how do we know that this has actually been researched and written, double checked by a journalist, and authenticated?” he says. 

News organisations should tell readers how they are using AI, McGuinness says. “We want everything that is AI-created or partially created, labelled as such,” he says.

“I think where AI is being used to create something, certainly, whether it’s text whether it’s images, whatever,” that should be labelled. And even if it’s just used as a tool, that readers should be told about that too. 

“Even if it’s something as simple as AI was used to analyse the data mentioned in this article – that simple,” McGuinness said.

Said McGuinness: “We’re not saying, you know, all AI is bad and all use of AI is bad, it’s just that we have to try and be careful about how it’s used.”

However, there are still some types of journalism that AI cannot yet do, says McGuinness. 

It can’t cultivate sources, who’ll pass on tips and information with a quiet word or quick message, McGuinness says. 

“It can’t go and cover something out in the field, so, you know, it can’t go out where there’s a car crash and talk to Gardaí and ask what’s going on, it can’t go to protests which turn violent and tell you what’s going on there,” he says.

“It's not intelligent with a capital ‘I’, in the way of human beings, and it’s unable to physically be there and to do certain things that human beings can do,” he says. But “who knows where it will be in 10 or 20 years time”.

There may well be a backlash against the introduction of AI into journalism, though, says McGuinness.

“I think it could be, sooner or later, that the scepticism of readers, listeners, viewers is going to be such that they're going to want, possibly want, less AI and more journalists’ boots on the ground, because that's the only thing that they know they trust,” he said.

Because, for example, how can you make a complaint against something that’s written by an AI? It doesn’t care, he says.

You won’t see the writer around and have a chance to give out to them about what they’ve got wrong. 

“And how can AI speak truth to power when it’s literally owned by the most powerful people in the world?” he says.