For a long time, news publisher Raphaella Kathryn Crosby identified as the “most anti-AI person”. The impact of a fuel crunch sparked by the US-Israel war with Iran dramatically changed her perspective.
Using AI is now a part of daily life for Crosby and her staff, all of whom would have been out of a job months ago without the adoption of the technology, she says.
The change reflects are wider trend of AI uptake by Australian journalists; a Medianet report found more than half of journalists surveyed used generative AI or large language models in their work in 2025.
A soon-to-be released Local & Independent News Association (LINA) survey found member publishers are increasingly reporting AI use, particularly for research, administration and limited editorial support such as transcription.
LINA executive director Claire Stuchbery tells PIJI the organisation is “engaging with the government” to find solutions to address AI copyright infringement and lack of compensation. But it supports journalists using AI to improve efficiency and free up time to pursue “irreplaceable” work such as spending time in communities and analysing information with important cultural and locally relevant context.
Crosby independently publishes New England Times and North Coast Times, two free-to-read digital news outlets collectively covering 25 local government areas in regional New South Wales. She describes their output as an “unusual mix” of regional hard news and hyperlocal soft news.
The US-Israel war with Iran broke out weeks after the launch of the North Coast Times.
Living in a regional community that relies heavily on diesel, Crosby says the impact of the conflict saw panicked local businesses step back from running ads with her publications.
The effect was financially devastating.
“I had never seen an entire economy, in a regional sense, just shut down in milliseconds. Literally, we lost all of our major clients in three days,” she says.
“We lost pretty much our entire revenue stream, bar about $2000 a month, in three days.
“I didn’t have any choice but to lay off staff … cut down on the amount of news and cut down on the quality of news.”
That was the moment Crosby decided to lean on AI.
She says the decision saved some jobs and salvaged web traffic that would have otherwise been lost due to low quality and quantity of stories from a team running on a decimated budget.
“If it had not been for AI, we would not have gotten through,” she tells PIJI.
Productivity gains ‘impossible to ignore’
Prior to March, Crosby says some staff had been using AI for “glorified spell checking” and occasionally for promotional content.
After the sudden contraction of revenue, the technology quickly became part of daily operations, sometimes cutting down production time on long articles or big projects by several days.
“I’m using the tool to make the grunt work around the actual business of journalism easier,” Crosby says.
“Some of our guys that do more editorial stuff have got it down to a fine art of doing an interview, getting the transcript out [of AI transcription platform] Otter.ai, dumping it into ChatGPT and generating an advertorial in under 15 minutes.
“That kind of productivity gain is impossible to ignore.”
Crosby finds it best to play to the respective strengths of individual AI platforms.
She says ChatGPT is best suited to good news stories and advertorials, while Claude does better with detailed tasks and hard news.
Claude is used to “triage” the news desk inbox and automatically assign stories on specific rounds to certain writers.
It has also been added to a range of programs, such as the team’s New South Wales election commitments and analysis tracker, which Crosby says will help create a platform allowing people to search through every statement and policy related to the upcoming state election.
Crosby has begun to allow Claude to have a “first pass” at some stories. It is programmed to adhere to the respective style of her publications, adds backlinks and context, and automatically generates tags and social media excerpts.
“I can take five reports, six media releases, three interviews, dump it all into Claude and say, ‘Claude, give me a rough outline of what this story should be.’ I then write the guts of it,” Crosby explains. She will then go back and forth with the program to fine-tune the piece.
“There’s a level of snobbery, that AI-generated or AI-assisted news generation is poor quality,” she says.
“I don’t think that’s the case. It’s a tool, and just like the internet is a tool, you can use it badly, or you can use it well.”
Lines remain
If her application for the News Innovation Fund is approved, Crosby intends to use AI to expand operations.
This will include the production of a printable newspaper for the New England Times so readers can print out copies themselves; Crosby says demand for a print version of the publication is high, but with no local printers left, printing and delivering each edition would cost more than $30,000.
She also hopes to produce a short AI-generated daily podcast delivering a bulletin of the publications’ top headlines, a plan similar to recent moves from larger organisations such as the ABC and Seven to use AI to produce digital articles based off radio and TV reports, respectively.
“Does [using AI for a podcast] kill about six jobs in the normal media creation process? Yes,” Crosby says.
“Do I apologise for that? No, because there is no way that we would be able to viably do that.”
But there are lines she’s not willing to cross.
Crosby says human oversight in news is essential, so a human will always do the final check of every AI output and AI will never make final publishing decisions at her outlets.
She also says ChatGPT and Claude are only allowed to produce work based off sources she has personally found, verified and inputted.
“We treat AI like a stupid junior,” she says.
“They sometimes do some good work, but you need to check it every single time.”
Rather than completely replacing journalists, Crosby expects AI tools will allow journalists to focus on the core of their work and chase deeper stories, enhancing the quality of content they produce.
“The world is kind of overwhelmed by quantity of news. What’s missing, and has been missing for quite some time, is quality,” she says.
“What using these AI tools does is it frees up the journalists to go and do more interviews, to deep dive more, to get more sources, to dig deeper into the report.
“It still requires a human brain to do that because it’s a human rationalisation-type process to be able to sniff out the story, to know who to go and talk to, to get people to trust you. Those are soft skills that … cannot be replaced by AI.”
New England Times was launched in 2023 after Crosby noticed a lack of quality local news had led to a sense of “malaise” permeating the regional town of Armidale.
Her initial goal was to produce about one article a day, but interest “exploded” when her story about the town’s quick successive loss of multiple GPs was picked up by the ABC and Crosby realised she had “greatly underestimated” the community’s news appetite.
Years later, New England Times’ staff “still feel guilty” that they can’t publish more. The team of 18 – mostly consisting of freelancers – produces a minimum of seven stories per day, with at least one long read prepared for the weekend.
But Crosby says the team is “running very much on vapour” as they await the outcome of their News Innovation Fund application.
– Sezen Bakan