From fact-checkers to editors, reporters across the region are discovering AI’s benefits, and drawing firm ethical lines to protect journalistic quality.
As part of preparing this article, we recorded a 30-minute interview in Slovak, transcribed it with Google’s Pinpoint, and used Gemini to translate the transcript into English. This made the work with the text much easier and saved us considerable time.
Journalists are discovering the substantial benefits of AI for efficiency and data management, confirms a Thomson Foundation report that mapped and compared the use of AI in newsrooms in the Czech Republic, Hungary, Poland, and Slovakia.
“But AI is not writing the articles for us,“ says Ondrej Podstupka, Deputy Editor-in-Chief of SME Daily, a prominent Slovak newspaper that has always welcomed technological advances that can help increase the quality of journalism. But, like many other outlets, SME is still quite conservative in its approach.
“AI is used for mechanical content processing, which means transcription, text editing, and error detection,“ Podstupka explains.
The reasons for this conservative use are several.
The first is quality. “As a newsroom, at this moment we are not willing to trust these tools enough to be sure that they will consistently produce good, quality, truthful content,“ says Podstupka. He doesn’t expect this to change soon.
The second obstacle is the fear that journalists will stop thinking creatively.
“I have never seen a robot write a good headline. They write an average one, but an excellent headline, in my opinion, has only ever been written by a human,“ says Podstupka.
Defining how far AI can be integrated into journalistic work without hitting these limits is an ongoing debate. Some newsrooms, such as SME, are updating their codes of conduct with clear principles. “But at the individual level, I don’t think it’s possible to write rules that cover all the edge cases,“ Podstupka notes.
Eliminating mechanical work
Over the course of the last few parliamentary elections, the SME website was filled with automatically generated articles showing results for each Slovak municipality. “This wasn’t artificial intelligence in the true sense of the word, it was just a script that inserted election results into pre-written sentences,“ explains Podstupka.
The adoption of AI in the region is progressing slowly, especially due to ethical concerns.
Today, when an ordinary reader opens SME, they might see a few AI-generated images or one or two AI-assisted sentences—always after being verified by an editor. Most AI use remains invisible, hidden in individual journalists’ workflow.
As Deputy Editor-in-Chief, Podstupka is trying to establish clear boundaries for AI use in the newsroom, preferring it to eliminate mechanical work.
For instance, today AI can transcribe an interview in minutes, a task that once took hours. But the text still needs substantial editing afterwards, simply due to obvious changes between the spoken and written language that a human transcriber would deal with as they work. The next step, he says, might be asking AI to clean the text of linguistic clichés.
“But we must have very good control over how much we let the robot into the text,“ Podstupka warns. “The moment the robot decides which question goes into the interview and which doesn’t, we are beyond the boundary.“
Code of conduct
AI use across the region varies from practical assistance to cautious experimentation.
“AI is most useful in the initial phases – when I’m mapping a subject, identifying patterns, or condensing interview transcripts,” says Neda Uzunkoleva from Boulevard Bulgaria, an online outlet offering in-depth explanations and analyses of current events.
For Boglárka Balogh, a fact-checker at Lakmusz.hu, AI helps with comprehension: “For example, I ask it to simplify legislation so I can understand it better and explain it more clearly.”
In smaller newsrooms, use often depends on individual initiative. “In our newsroom, the average age is over 60, so it’s mostly younger colleagues who use AI,” says Kata Tihanyi from Veszprém Kukac, a rural Hungarian outlet. “As younger people, we’re aware of the risks, like unreliable information, so we use it within clear ethical boundaries.” Similarly, Zsuzsanna Bilás from Napunk.sk notes that her newsroom, part of the Denník N group, uses AI “very minimally.”
The new code of conduct in preparation at SME will set out general principles for AI use in the newsroom. But according to Podstupka, it won’t be able to answer every situation.
Bulgarian economic weekly paper Kapital is also working on an ethical code.
“We have a team of journalists and a technical person that have a mission to experiment with AI and suggest how to use it so it is helpful for journalists,” says Anina Santova.
Smaller outlets usually don’t have any formal AI policies.
The journalists interviewed did not mention any misuse within their organisations. But AI-generated articles can easily be found online. “A lot of mainstream and gossip websites are fully using AI for their articles and calling it journalism, which is frustrating,” says Delia Dascalu from Info Sud-Est, a local newspaper in Constanța, Romania.
Small and big outlets
According to the mentioned study mapping the use of AI by newsrooms in the Czech Republic, Hungary, Poland, and Slovakia, published by Thomson Foundation and Media and Journalism Research Center in 2024, the average percentage of AI use in newsrooms does not exceed 15%. AI is mainly used for technical tasks, while journalists and editors continue to handle the more complex operations.
“This is particularly evident in small newsrooms that lack the necessary human and technical resources to make more extensive use of AI tools, even though they would benefit most from them,“ the report states.
Bigger publishers can generally invest more. A good example is The Washington Post, which offers “AI Overviews“ for its articles. With one click, readers can see a short summary and a few bullet points.
Kapital in Bulgaria uses a similar feature, providing three-bullet summaries of articles. “But we always have to edit them,“ says journalist Anina Santova.
The Washington Post also offers its own chatbot, Ask The Post AI, a generative-AI tool that answers questions based only on content from The Washington Post. The risk of hallucination is therefore smaller, though the chatbot may miss very recent events or ones not covered by the paper.
SME Daily has no plans to create something similar. “We considered it, but there are two problems,“ says Podstupka.
First, it’s expensive to train the tool to be consistently accurate and reliable without human supervision. Second, the added value for readers would be lower than that of general-purpose chatbots. “The benefit for the reader is smaller than if we use the same resources to make the work of our journalists easier,“ he says.
Next steps
For smaller languages like Slovak, AI development lags slightly behind English. Tasks the one generation of chatbots managed well in English were only being handled smoothly in Slovak in the next generation.
SME is currently implementing a new CMS with a first-generation AI correction tool. Podstupka estimates it can catch 80–90 percent of mistakes in Slovak—always spotting typos, but still struggling with commas. “Even ordinary people often don’t place them correctly,“ he notes.
“I don’t think it will replace a good proofreader, especially for more complex or stylistically rich texts,“ he adds.
After the new CMS is in place, SME editors will be able to experiment with tools for generating keywords or social media posts.
Podstupka sees the biggest potential in working with video, but doesn’t expect it to catch on widely. Instead, he predicts more sophisticated text applications, like converting articles into different formats or supplementing them with infographics.
The work of journalists is changing, and younger reporters are very aware of it.
Kata Tihanyi from Veszprém Kukac recalls a famous Hungarian copywriter saying: “AI won’t take our jobs, it will be the AI users who take jobs from those who don’t use it.“
“Of course, I also feel some anxiety about this, as many others probably do. But I don’t think journalism itself is at particular risk, especially not opinion writing,“ she says.
According to Henrietta Bocz from Átlátszó Erdély, a Romanian investigative portal writing in Hungarian for the minority living in Transsylvania, AI poses a risk to simpler news reporting, since it can quickly generate articles or rephrase stories from other outlets. “It may also threaten readership, as AI can search for and summarise news content independently,“ Bocz warns.
“I believe AI will replace routine content and low-quality production, but not intellectual or narrative journalism,“ says Uzunkoleva from Bulgarian. “The more personal, analytical, or investigative a piece is, the more it needs a human voice, judgment, and lived experience.“
Author: Katarína Jakubjaková
This article was produced as part of the Thomson Foundation’s Internship programme for young female journalists implemented under the Deepening Digital: Reinforcing Resilience project.
