AI Pioneers and Stragglers in Greece: Challenges, Gaps, and Opportunities for Journalists and Media

Abstract

Media organizations are experiencing ongoing transformation, increasingly driven by the advancement of AI technologies. This development has begun to link journalists with generative systems and synthetic technologies. Although newsrooms worldwide are exploring AI adoption to improve information sourcing, news production, and distribution, a gap exists between resource-rich organizations and those with limited means. Since ChatGPT 3.5 was released on 30 November 2022, Greek media and journalists have gained the ability to use and explore AI technology. In this study, we examine the use of AI in Greek newsrooms, as well as journalists’ reflections and concerns. Through qualitative analysis, our findings indicate that the adoption and integration of these tools in Greek newsrooms is marked by the lack of formal institutional policies, leading to a predominantly self-directed and individualized use of these technologies by journalists. Greek journalists engage with AI tools both professionally and personally, often without organizational guidance or formal training. This issue may compromise the quality of journalism due to the absence of established guidelines. Consequently, individuals may produce content that is inconsistent with the media outlet’s identity or that disseminates misinformation. Age, gender, and newsroom roles do not constitute limiting factors for this “experimentation”, as survey participants showed familiarity with this technology. In addition, in some cases, the disadvantages of specific tools regarding qualitative results in Greek are inhibiting factors for further exploration and use. All these points to the need for immediate training, literacy, and ethical frameworks.

1. Introduction

Over the last three decades, journalism has considerably transformed due to the introduction of information and communication technologies. The advent of the internet redefined the processes of news collection, creation, distribution, and consumption [1], and this digital revolution was the springboard for the introduction of artificial intelligence technologies and tools in journalism [2]. This transformation is now increasingly driven by developments in artificial intelligence (AI) technologies [3]. AI tools are rapidly transforming the media ecosystem, and media organizations worldwide are adopting and using them for many purposes. Media newsrooms started using AI tools in 2014, but since ChatGPT 3.5 was released on 30 November 2022, the use of AI in global media took a very big leap; however, Greek newsrooms still seem to be lagging [4].

The way news is produced is changing rapidly [5]. Newsgathering, writing, production, and dissemination have been significantly influenced by the advent of AI [6]. AI assists journalists in various ways, such as by automating repetitive and/or boring tasks, using data to find stories, and telling stories in new and interesting ways. Using intelligent computer programs, AI can even write news articles, produce videos or podcasts, and create images. This means that news can be delivered faster and in a more engaging way for the audience [7].

Greek media organizations have begun to experiment with AI tools, but their use remains at an early, exploratory stage, in contrast to international media. AI has not yet been fully integrated into Greek newsrooms, primarily due to the sector’s delayed adoption of technological developments and the long-lasting impact of the financial crisis that affected Greece from 2010 to 2018 [8,9]. As a result, many media outlets cannot afford the high costs associated with advanced AI systems. Consequently, journalists often engage with AI on their own initiatives, using it experimentally for personal purposes without institutional support, formal training, or established guidelines. Self-directed learning is the predominant approach, and there is currently no cohesive strategy for the adoption and implementation of AI across the Greek media landscape [4].

However, these changes raise some important questions. For example, some worry that AI may replace journalists altogether [10]. Others are concerned that news written by AI may not be accurate or truthful [11]. There are also questions about whether it is acceptable for machines to write the news. As AI is beginning to undertake more and more of the work [12] that journalists used to do, questions are being raised about the value and importance of human creativity in journalism compared with the efficiency of machines.

This study aims to examine the Greek media ecosystem, with a particular emphasis on how journalists engage with and incorporate artificial intelligence (AI) tools into their professional practices. It offers a comprehensive analysis of the insights, challenges, gaps, and opportunities associated with this technological integration. Additionally, the research investigates Greek journalists’ level of knowledge concerning AI technologies and their attitudes toward the implementation of AI in journalistic workflows. Finally, it explores the extent to which formal policies exist within Greek media organizations regarding the adoption, regulation, and governance of AI tools.

1.1. Literature Review

Developments in AI boost its functionalities, with automation and algorithmic processes reaching an advanced stage that allows them to perform fundamental journalistic tasks, thus transforming journalistic practices [2]. Newsrooms worldwide are exploring the adoption of AI to improve the sourcing, organization, and distribution of information, but a gap exists between organizations with abundant resources and those with limited means [13]. As noted, Greek media faces a lack of resources, and newsrooms have delayed adopting this technology [1,14]. AI applications can automate content creation, personalize news feeds, and process extensive datasets in real time [15]. Although AI has the potential to revolutionize various fields, its outcomes are often constrained by factors such as resistance to change, issues of literacy, regulatory barriers, ethical dilemmas, and technological limitations.

Major international media organizations had integrated AI applications into content production as early as 2014, generating text with structured data [16], presenting financial business reports, and producing sports news [17,18] with the Wordsmith tool, developed by Automated Insights, Durham, North Carolina, U.S.A (Automated Insights, n.d.). Essentially, algorithms converted numerical data into text, employing rule-based AI technology that had existed for decades [19]. Structured data algorithms, such as Wordsmith, rely on predefined templates that constrain linguistic variability and adaptability, rendering them most effective for repetitive, data-driven reporting tasks. In contrast, the rapid advancement of generative AI has introduced sophisticated language models capable of producing context-sensitive, nuanced, and creative content. This technological shift not only underscores the limitations of rule-based systems like Wordsmith but also reveals the transformative potential of AI-driven journalism. Generative artificial intelligence (GenAI) gained widespread attention with the release of ChatGPT 3.5 as a free preview on 30 November 2022 [20].

AI is already being used by journalists and news organizations in various ways. The Open Society Foundation investigated 880 application scenarios in journalism and found that structural changes are expected in the coming years [21]. According to that study, GenAI is being applied by media organizations in three key areas: newsgathering, news production, and news distribution [2,22]. By exploring the relevant literature and the dynamic database of the LSE ‘As illustrated in Case Studies: Exploring the Intersection of AI and Journalism (JournalismAI, n.d.), the integration of artificial intelligence into journalistic practices reveals a diverse range of applications and challenges, developed within the framework of the Journalism AI program, and it becomes evident that GenAI tools for journalism are rapidly evolving and impacting all media and all forms of journalistic work, including newsgathering, fact-checking, content production, personalization, news distribution, audience engagement, and comment moderation.

1.2. Background of the Use of AI in Media

Some of the leading tech minds in the world and in media have heralded generative AI as a next-generation technology. For publishers, it seems offers great potential for workflow efficiency, text creation, correction, search/research, and translation—factors that most people say will free up journalists and editors, allowing them to focus on producing core, quality content and improving along the way. Moreover, this offers opportunities for personalization. However, generative AI, particularly chatbots, raises several questions, challenges, and serious concerns. Nevertheless, GenAI is developing rapidly, and mistakes have already popped up. The Human Error Project1 is one of the organizations mapping such mistakes. Errors, such as the 2025 incident in which ChatGPT falsely identified a Norwegian man as the murderer of his two children or the 2023 case involving an Australian mayor who was incorrectly named as the perpetrator in a corruption case from the early 2000s, highlight the serious risks associated with AI-generated misinformation and must be proactively prevented [23,24]. Essentially, these tools need to “learn” more and more and be tested repeatedly, both by their creators and those on the frontlines, like the publishers experimenting with it. According to a survey, preventing the spread of misinformation, data privacy, and regulation are issues associated with social media, but they are even more pertinent with GenAI [25].

The same survey suggests that most newsrooms perceive the supportive role of generative AI tools, considering them an important means to increase productivity and efficiency in various processes. Several newsrooms are already working with generative AI to create summaries and bullet texts. One could argue that this function constitutes both content creation (quality improvement) and workflow help (supportive). In the survey, 39% of the respondents see the use of AI in this way, 50% see these tools as purely supportive for newsrooms, and 8% see them as quality improvements.

Furthermore, with the varied types of content, journalists and editors need to produce for different platforms. Text summaries/bullets seem a logical and practical use of tools like ChatGPT, as it can learn from a text written by a journalist, as opposed to the riskier scenario of asking it to write a summary about “Joe Biden’s speech to Congress”. Notably, 54% of the participants in the survey are doing just that. More than 40% also use it for simplified searches/research, correcting texts, and improving workflows [23]. While the respondents rated text creation as the most useful area for GenAI tools (63%), newsrooms are clearly promising for further development in the areas of workflow/efficiency, translation, and personalization, at least compared with the actual use so far. In total, 61% rated workflow/efficiency as an area where GenAI can help the most. In fact, 43% of the newsrooms already use these tools for this purpose. While “translation” is considered valuable by 51%, only 32% have used tools in this area so far. Furthermore, 42% see potential in extra personalization, but only 19% use it. For the highly rated text creation, 54% are already active in this area [25].

Artificial intelligence is not a novel technology; in fact, it has been employed by numerous media organizations worldwide since as early as 2014. However, a significant breakthrough occurred with the release of ChatGPT (Generative Pre-trained Transformer), a chatbot developed by OpenAI and launched in November 2022, which has profoundly reshaped the media landscape. Trained on an extensive corpus of textual data, ChatGPT is capable of performing a wide range of natural language processing (NLP) tasks, including question answering, language translation, text summarization, and text-to-speech generation—applications that are increasingly being utilized by Greek media outlets.

The Associated Press started using AI in 2014 [26,27]. In collaboration with Automated Insights, it uses the Wordsmith news platform to produce automated news using its own algorithm, programmed to write in the style of the Associated Press. At the end of each report written by the algorithm, a note indicates this and makes it clear. The AP now makes extensive use of this system, which can produce about 4400 reports in four months. Thus, it validly and quickly covers a wide range of facts. It also covers sporting events, including all US baseball leagues, i.e., 142 teams in 13 divisions and 10,000 games per year. Given its coverage of baseball with automated journalistic content, the AP also announced the hiring of its first automation editor in 2016 [26,28,29,30].

Among the pioneers in the use of AI is The Washington Post, which makes extensive use of algorithms in its online edition, especially in the coverage of important events, such as the Rio Olympics in August 2016 or the presidential elections of the USA in November 2016. These two events were covered by The Washington Post with the help of the Heliograf algorithm, which was manufactured in-house [30,31].

With the help of artificial intelligence, in addition to content production, there are now virtual news anchors on TV. This is a significant development that has created a new paradigm. China’s Xinhua News Agency has built such robot news anchors. With the help of a computer, in April 2018, Xinhua built and displayed a simulation of one of its news anchors, an avatar with the same movements and speech as that journalist. Its similarity with the real journalist is such that one would hardly be able to tell the difference. In November 2018, Xinhua produced a second news anchor, this time a woman, and today, it has increased its use of virtual news anchors. Xinhua has been experimenting with artificial intelligence for years. In 2017, it introduced a robot called Inspire that worked experimentally at the agency. This practice is now followed by several television stations and websites around the world [30,32].

The Washington Post still uses Heliograf, which creates content and covers major events. However, in early November 2024, they launched a new tool named “Ask The Post AI”, an experimental generative AI tool that delivers answers to users’ questions in keeping with the world-class journalism of The Washington Post [33]. The New York Times also makes wide use of AI. The ΝΥΤ began using algorithms for content recommendations in 2011 but only recently started applying them to home page modules; it now uses them on its website and for other applications [34]. The NYT won the 2024 Pulitzer Prize for using an AI tool. In the international reporting category, a December 2023 report by The New York Times’ visual investigations desk was one of several stories recognized about the war in Gaza. The Pulitzer-winning team trained a tool that could identify craters left behind by 2000-pound bombs, some of the largest in Israel’s weapons arsenal. The Times used the tool to review satellite imagery and confirm that hundreds of these bombs were dropped by the Israeli military in southern Gaza, particularly in areas that had been marked as safe for civilians. Two 2024 Pulitzer winners disclosed using AI in their reporting, a first for these prizes. Apart from the NYT, local media based at the Chicago City Bureau and Invisible Institute won the prize for a series entitled “Missing in Chicago,” using thousands of police misconduct files [35].

On the other hand, the aforementioned examples are drawn from North American media organizations operating within a high-resource media landscape. Greece, while geographically and politically situated within the Global North, exhibits several characteristics commonly associated with the Global South—a point we elaborate on below. Following a prolonged financial crisis that severely impacted its media sector, the capacity of Greek media organizations to invest in and adopt advanced technological systems remains limited.

A pronounced digital divide between the Global North and South continues to shape the Greek media environment. Disparities in digital access, infrastructure, and investment represent significant obstacles to the widespread adoption of emerging technologies, including those transforming the media industry.

Apart from the above, many other media use AI, including Yahoo Sports. Their Daily Fantasy platform generates millions of stories every week, providing details not found on any other sports websites and creating a huge amount of material. In Europe, the BBC, Sky News, The Telegraph, the Daily Mail, and The Economist use AI and AI tools. In Norway, the National News Agency uses AI, and in Sweden, Mitt Media also uses AI tools.

1.3. Challenges and Risks

The use of AI raises many questions regarding its effects. Many issues need to be discussed, such as work and the moral and political dimensions, which include issues that touch on democracy itself. As mentioned above, large news organizations abroad make extensive use of AI from big technology companies, such as OpenAI, Google, Amazon, and Microsoft, while developing their own tools. Most publishers, especially smaller ones, choose third-party solutions from platform companies due to the high cost of developing their own tools [36]. However, the complexity of AI increases the control of platform companies over news organizations, limiting their autonomy. In addition, the lack of transparency in AI systems raises concerns, and intellectual property issues arise regarding the content the algorithms were trained on, the content they produce, and issues of legal liability regarding the content produced. The New York Times sued OpenAI for copyright infringement, alleging that the creator of ChatGPT used the newspaper’s material without permission to train the massively popular chatbot [37]. There is also a risk that the extensive use of AI will degrade journalists, raise concerns regarding potential job displacement their jobs, and decrease their credibility due to the production of fake news.

Among the challenges and risks outlined in the Open Society Foundation’s research scenarios are addressing the filter bubble and polarization phenomenon; the production and distribution of fake and deepfake news; malicious threats to manipulate the public; the inability of legacy media to adapt; and the simultaneous development of human and machine producers and opinion makers [21]. Extensive research has shown that AI can make incorrect decisions in cases of content control (moderation) [38], with errors disproportionately affecting marginalized communities; in a political context [39], minority voices are silenced and censored more often, stereotypes are reproduced [40], and topics of particular interest to them are at higher risk of deletion [41] or downgrading. For example, a 2023 analysis of more than 5000 images created with the generative AI tool Stable Diffusion found that it simultaneously amplifies both gender and racial stereotypes [42]. Another study found that racist biases against particular groups are baked into ChatGPT itself [43]. Rona Wang, an Asian-American graduate of MIT, sought to generate a professional LinkedIn photograph using an AI tool; however, the system modified her appearance to resemble that of a white woman, prompting public debate over racial bias in AI design [44]. Similarly, a separate study [45] revealed that language models exhibit covert forms of racism through dialect-based prejudice. Specifically, they reinforce negative stereotypes associated with African American English, resulting in biased outputs in contexts such as employment, legal judgments, and sentencing—despite simultaneously expressing overtly positive sentiments about African Americans.

International organizations have issued guidelines for understanding the use and impact of artificial intelligence systems to mitigate and address disinformation and hate speech [46]. The European Union has set a regulatory framework for the use of AI [47] following a risk-based approach. The higher the risk, the stricter the requirements. The definition of risk is “the combination of the probability of harm occurring and the severity of that harm”. Transparency is essential for news organizations according to the AI Act (AIA). In other words, a newsroom using AI tools and systems for news production should be transparent to the audience, labeling the content as generated with the assistance of AI. Synthetic media content, called deepfakes in the Act, is also covered. At the same time, several news organizations have developed guidelines for the responsible, ethical, and impartial use of artificial intelligence by their journalists, but according to a study [48] in 17 countries analyzing 37 guidelines, institutional isomorphism can homogenize rules. The geographical concentration of AI governance guidelines—primarily originating from Western nations, particularly North America and Europe—raises persistent concerns about global power asymmetries in the regulation of emerging technologies. This dominance frequently leads to institutional isomorphism, wherein countries and organizations outside these core regions adopt similar AI frameworks. Such adoption is often driven not by contextual suitability but by external pressures to align with internationally recognized standards. Institutional isomorphism may manifest through coercive, mimetic, or normative mechanisms, potentially marginalizing culturally or economically appropriate alternatives and reinforcing Western-centric models of technological development, ethics, and governance. European media emphasize protecting the public in contrast to North American media, which focus on journalistic values and the responsible integration of AI in newsrooms. However, only in non-Western, educated, industrialized, rich, and democratic countries has human involvement in the development and integration of tools been adopted. International and European journalist associations and journalism-related NGOs under Reporters Without Borders (RSF) published the Paris Charter on AI and Journalism, while the French Council for Journalistic Ethics and Mediation adopted the EU risk-based model [49] to uphold the Code of Journalistic Ethics and foster transparency toward the public when using AI models.

All of the above are expected to significantly change the work of journalists. Tools like ChatGPT can accelerate the implementation time of 14% of specific tasks across hundreds of professions by up to 50% [50]. In journalistic specialties, research estimates that only 16.7% of reporters’ and journalists’ tasks will be beyond the capabilities of large language models (LLMs) with software adaptations. For editors, this percentage rises to 23.8%. In other words, algorithms will be able to perform 83% and 76% of the tasks of these respective journalistic specialties. Consequently, for small newsrooms and local media outlets, news coverage could be managed with just two to three journalists, a tempting prospect for owners but a frightening one for professional journalists.

2. Materials and Methods

A survey in Greece in November 2024 shows that one in three Greeks (36%) uses ChatGPT and one in four (26%) uses other AI tools. The same survey shows that two in three Greeks (63%) believe that AI will help them improve their professional prospects, especially those (79%) who already use AI tools. On the other hand, 42% believe that AI is a threat, and four in five (79%) believe that the biggest risk of AI is misinformation. This is the reason that three in five (62%) say that there is a need for a strict framework for the use of AI [51].

The media landscape in Greece is characterized by considerable media concentration, digital fragmentation, and a notable lack of public trust in news outlets [52,53]. Political polarization and concerns regarding the influence of politicians and powerful business figures further complicate the media environment. In this context, digital media in Greece largely adhere to advertising-supported business models that were initially developed in the 1990s and early 2000s. Compared with other European countries, the adoption of online subscription models has been relatively slow, primarily due to concerns over audience engagement and potential loss of influence. Kathimerini, one of the country’s leading legacy newspapers, became one of the few to introduce a soft paywall in 2023.

Although Greece is classified as part of the Global North, its digital technology sector exhibits characteristics typically associated with the Global South. A notable digital divide persists between the Global North and South, which continues to influence Greece’s media landscape. The disparity in digital access and infrastructure remains a significant barrier to the widespread adoption of modern technologies, including those within the media sector.

Recent findings from a survey conducted by [4] indicate that AI is not yet fully integrated into Greek newsrooms. At present, journalists are using AI in an experimental stage, primarily for personal projects, without formal training or established guidelines for its application. This trend points to a reliance on self-directed learning rather than structured, organizational strategies for AI adoption. There is also no cohesive approach to integrating AI into Greek media practices, a fact that hinders its effective utilization.

Despite these challenges, AI technologies are beginning to make their way into the operations of Greek media outlets. Notably, protothema.gr launched an AI-supported English version of its platform in July 2024 [54] and introduced AI-powered comment moderation tools [55]. Around the same time, To Vima, a digital edition of a prominent Sunday newspaper, incorporated AI-cloned voices for its opinion pieces, including those of prominent columnist Yannis Pretenteris [56]. In addition, robonews.gr, a new AI-only news website, was launched in May 2024. Previous research has also shown that Gazetta.gr, a leading sports and infotainment site, has begun using AI tools for video production and social media content creation.

Given these developments, the purpose of this study is to provide an in-depth analysis of how AI is reshaping the Greek media landscape, focusing on the challenges and opportunities it brings to the field of journalism and media organizations.

Specifically, through qualitative analysis, this study will explore how leading media outlets in Greece are preparing for and utilizing AI within the framework of their journalistic practices, examining the potential implications for the future of news production and distribution in the country.

2.1. Research Methodology

This study employed purposive sampling to recruit journalists from a range of media sectors, including print, television, radio, and digital platforms, as well as across different age groups, in order to capture a broad spectrum of experiences and perspectives regarding the adoption of AI in journalism. Purposive sampling is particularly well-suited for identifying information-rich cases that provide in-depth insights into the phenomenon under investigation [57].

As AI advances rapidly, we have reassessed the Greek media ecosystem and updated our previous research [4] to uncover new findings. In our earlier study, we conducted 25 semi-structured interviews with 25 journalists between March and May 2024. Of these, 6 journalists work for national television stations, 4 for radio stations in the Attica region, 4 for nationwide newspapers, 6 for news websites (3 of which are part of media groups) boasting unique monthly users exceeding 2 million, 1 for a public service news website, 1 for a regional news website, 1 for a sports website, and 2 for lifestyle websites.

From September to November 2024, we then conducted 14 semi-structured interviews with various journalists from the same media outlets (2 from national television, 2 from radio stations in the Attica region, 5 from news websites, 2 from sports websites, 1 from a regional site, and 2 from a newspaper) (

Table 1. Sample of Interviews.

1st Sample		2nd Sampleample
Medium	n	n
TV	6	2
Online	6	5
Radio	4	2
Newspaper	4	2
Public Service (online)	1
Online Regional	1	1
Online Sports	1	2
Online Lifestyle	2
Total	25	14

). Despite the differing number of interviewed journalists in this second phase, the sample remained consistent, as we examined the evolution of AI across the same media, as in the first study. Additionally, in both instances, we interviewed journalists aged 24 to 55, representing both genders and all editorial roles (editors, reporters, chief editors, editors-in-chief, and news managers). The interviews were conducted via Zoom, lasting an average of 25 min each, recorded, and transcribed with the assistance of AI, under the researchers’ supervision and with verbatim corrections. NVivo14 software was utilized to systematically code, categorize, and analyze qualitative data obtained from interviews with journalists. The analysis involved iterative close reading and thematic coding, guided by the study’s research questions.

2.2. Research Questions

• RQ1. Do Greek journalists incorporate AI in their workflows?
• RQ2. Are Greek media using third-party AI tools instead of developing their own?
• RQ3. Are Greek journalists well equipped and educated for the responsible use of AI in their workflows?

3. Results

Artificial intelligence (AI) is increasingly recognized as a workflow component in journalism, facilitating a hybrid process in which reporters and editors engage with self-learning systems throughout various stages of media production.

3.1. Flux AI Use in Greek Newsrooms

The participants in this research come from different backgrounds, cultures, education, ages, professional experiences, and different day-to-day routines. For example, a TV journalist has different needs regarding the production of a story compared with a journalist working for a website. As an experienced TV reporter notes, “‘I don’t need a story recommendation tool, but I need a credible research tool. I’m in the field and I see the story evolving. What I need is data about similar stories or for the area”. On the other hand, a website editor’s needs are different, as they must rewrite and produce image(s), videos, graphics, social media videos in different formats, newsletters, push notifications, etc. This finding is fundamental regarding the way AI tools are used and implemented in a workflow as useful assistants for journalists and newsrooms. In this survey, journalists unanimously acknowledged that AI tools offer significant potential for advancing journalistic practices by enabling the creation of new forms of storytelling that were previously constrained by resource limitations or technical complexity. In our previous research, journalists were cautious about the use and the benefits of AI. “AI tools can provide us the skills we lack”, mentions a journalist who started producing short videos for social media. No matter their backgrounds or ages, journalists can be categorized as tech-savvy, curious, or indifferent regarding AI. Tech-savvy journalists represent a small portion of our sample (7 out of 25 in the first study; 3 out of 14 in the second). The curious are the majority of our sample (11 out of 25 in the first study; 6 out of 14), and the rest are indifferent (

Table 2. Classification of Journalists Based on AI Engagement.

1st Sample			2nd Sample
	n	%	n	%
Tech-Savvy	7	28	3	21.43
Curius	11	44	6	42.85
Indifferent	7	28	5	35.72
Total	25	100	14	100

). Tech-savvy journalists are self-learners and systematically follow technology evolution, including AI; attend seminars; experiment with new tools; and apply these tools according to their needs. Journalists who experiment with tools but are not using them in their daily routines can be defined as curious. Indifferent journalists do not engage with any AI tools. This category believes that “journalism has nothing to do with technology”.

“I’m a journalist and I have no need for tools in order to get in touch with my sources and obtain information to write a story” says a journalist working at a newspaper.

This last category is dominated by the ideology of journalistic authority.

However, in our previous research [7], a journalist from the same newspaper was using AI tools to gather data and information for her stories, underlining her colleagues’ attitude of indifference.

Most journalists believe that original journalistic work will not be undertaken by AI tools since communication with sources is one of their core responsibilities. However, they understand that when a story is published, follow-up stories will be easier to produce with the assistance of AI tools. The speed of follow-up production is an important issue since, as one of the respondents underlined, a new story with extra background and data can be produced in a few minutes and might be better, gaining more attention than the original one. The copyright issue was raised at this point, as the journalist claims that “my story should be protected and no tool should be allowed to use it, for at least 24 h”. There is a risk that original journalism will be subordinated to the technological superiority and familiarization of other journalists and media with AI.

3.2. Segmented AI Integration in the Newsroom

So far, media organizations have not embedded AI in the newsroom workflow. Some media are pioneers, some watch the evolution and prepare, and some are indifferent. The same pattern occurs among journalists and media, but this does not mean that they characterize each other. In other words, tech-savvy journalists exist in indifferent media, and vice-versa. According to our research, Greek legacy media are in the “watch and prepare” category. However, at this phase, leading websites are pioneering in “safe zones.” Different segments in news production use AI without interfering with their workflows. In our study [4], we presented case studies from Greek media using AI. For example, protothema.gr uses AI to translate Greek content into English, for semantic searches of these English versions, and to moderate comments. According to our research, the English version increased content production from four to five stories per day to almost 70 per day from January 2024 to November 2024. Furthermore, this increased moderated comments using AI by 1000% while reducing the number of journalists working on moderation from eight to two. Voice cloning for text-to-audio is used in two other websites: gazetta.gr and tovima.gr.

Sports24.gr is one of the oldest and a leader in sports news in Greece, and it is launching AI-assisted sections to present results and produce stories. It is also preparing an AI department, meaning a dedicated department is expected to be established within the newsroom, led by a chief editor responsible exclusively for AI-related initiatives, including the implementation and strategic use of emerging tools and techniques.

There is also a beta version of a website called robonews.gr, which produces images and text with the prompts of its founder. However, this content is labeled as AI-produced according to the provisions of the EU AI Act and its implementation in journalism [58].

3.3. Which AI Tools Are Used by Greek Journalists?

Automated applications are being developed to produce all content types across all media formats and platforms, enabling their distribution through all channels (e.g., social media) in various formats tailored to audience preferences.

In summary, numerous applications designed for daily use in newsrooms are emerging, including the following:

• Searching for and retrieving information, for instance, from documents and structured data;
• Fact-checking and accurate verification of information;
• Monitoring and moderating comments;
• Generating topic suggestions based on audience engagement and interest;
• Providing ideas for content themes;
• Transcription, translation, and summarization of texts;
• Writing articles, for instance, from structured data;
• Producing photos, graphics, and videos from textual inputs in multiple formats;
• Converting text to speech, audio editing, and sound reproduction;
• Generating voiceovers from text, voice cloning, and creating audio files from text inputs;
• Narration production, music composition, and subtitle generation (including real-time subtitles);
• Dubbing, voice cloning for dubbing synchronized with video, and content classification;
• Archiving images and videos, audio and video editing, and identifying filming locations;
• Personalizing content and recommendations, including website customization, article suggestions, and newsletters;
• Tailored notifications, enabling dynamic subscriber registration and analyzing audience metrics.

These applications illustrate the transformative potential of AI in streamlining newsroom operations and enhancing content production and distribution processes. However, according to our findings, Greek journalists who use AI tools have integrated them into a small number of tasks, which are limited to transcribing, translation, video subtitling, short video production for social media, summarization, text rewriting, research, ideas production, and image creation.

The respondents also report inefficient results for other tasks because the Greek language is still a barrier regarding text. In many cases, poor linguistic results disappoint Greek journalists and make them abstain from AI learning procedures, or they downgrade AI’s capabilities. “There is always a chance to have wrong or poor results using a mis-trained tool, but it is possible, in a few days or weeks, to have excellent results, since it is always trained and re-trained”, says a participant in the survey, underlining the fact that most journalists judge these tools as static and not dynamic. Similar limitations are observed with the Georgian language, as ChatGPT may occasionally produce imprecise outputs due to limitations in its training data [59].

Given these barriers and attitudes—most of them based on self-perceptions and not substantial knowledge—the Greek media ecosystem seems to abstain from the seismic changes coming from the use of AI.

RQ1 is not confirmed since AI tools are partially integrated into newsroom workflows. Greek media and journalists use AI tools, but when applied to the media workflow, this use does not affect all the segments of the news. It is applied in small sections, such as translation, comment moderation, semantic search, transcribing, text-to-speech, short video production, and image creation. Currently, AI tools are not incorporated into core journalistic workflows, such as writing, editing, the creation of ideas, fact-checking, and newsgathering. The language barrier and/or the journalist’s ideology constitute a hindrance to the expansion of AI in the newsroom workflow. Considering the lack of knowledge of AI, the attitudes of the media, and the journalists’ stances (indirectly mentioned in this RQ section), there is a need for AI literacy, not only for journalists but also for media organizations. The use of AI is limited for the majority of journalists, and it is mainly exploited for tasks and skills missing from the newsroom (video editing and production, image creation, etc.) or boring tasks (transcription).

3.4. Third-Party Tool Integration and Lack of Funding

At the moment, no Greek Media have developed their own AI tools. All of them are using “solutions from the shelf”. They test and choose the most adequate—in their opinion—tools to satisfy their needs. According to the journalists who participated in our survey, due to a lack of AI experts and poor financial resources, the media are not involved in producing their own AI tools. Another reason is that “we prefer to have short time subscriptions, because we believe that in the near future we are going to have more and better options for the tools we need, since it is an ongoing developing process”.

Media organizations that have already integrated AI tools that cooperate with third-party companies and/or receive funding.

Protothema.gr collaborates with Cloudevo.ai and partners with Google in case studies. The AI-supported English edition was integrated with the cooperation of Google [40]. To automate comment moderation, Protothema collaborated with a US-based company [41]. In November 2024, the English edition implemented an AI Semantic Search system, developed in collaboration with Google, FT Strategies, and Cloudevo [60].

The Makedonia newspaper is developing its own tool, PNN (Personalized Notifications and Newsletters), with a grant from Polis LSE and Google Initiative (Journalismai.info, n.d.). For this project, Makedonia is collaborating with experts from Greek academia. The convergence of AI and data presents novel opportunities, including enhanced reader engagement, innovative monetization strategies, and personalized news feeds. However, these advancements also pose a challenge in balancing the creation of echo chambers with the journalistic commitment to serving the public interest.

The rest of the media use AI tools from third-party companies, such as Eleven Labs for audio; Veed.io for short videos; Midjourney for images; and ChatGPT, Gemini, and Perplexity for many other tasks and purposes. They also use a variety of tools for transcription and translation, either for free or with subscription. However, without funding, collaboration, or support from experts from abroad, Greek media organizations face significant challenges in implementing AI tools without external support. This finding is important for the Greek media ecosystem because it has caused a significant delay regarding the transition to AI-assisted journalism and has increased the gap between Greek media and media across the Global South. Thus, our findings confirm RQ2. Media organizations and journalists in Greece are using third-party AI tools, while other media, with the support and guidance of organizations such as Google Initiative, cover their needs in terms of human resources (translating content for English editions, moderating comments, etc.) and technology expertise (semantic search). However, certain efforts can help to create AI tools, such as personalized push notifications and newsletters with Greek experts. While personalized news strategies present new opportunities for enhancing reciprocity between media organizations and their audiences, reconciling the often conflicting principles of journalistic judgment, individual user preferences, and algorithmic selection remains a complex and ongoing challenge [61]. The lack of funding and skilled experts undermines the possibility of applying AI tools in Greek newsrooms. However, there are many areas in which AI can assist journalists and media given its use in international media. This interaction spans the entire production pipeline, from gathering information and research to story ideation, text, image, audio, and video creation, and even the dissemination, consumption, and monetization of journalistic content on a (hyper)personalized basis. AI technologies can assist in generating story ideas, with algorithms determining which topics are likely to resonate most with audiences based on website or app usage data. Furthermore, AI plays a role in gathering information for stories, as reporters often interact with AI-driven systems during this process. In the production and postproduction phases of news stories, AI enhances software functionality, aids in data analysis and visualization through statistical tools, and influences how stories and headlines are distributed across various online platforms. Sporadic, segmented use of AI is probably the beginning, but it is not the solution. The absence of a working AI-assisted newsroom workflow—one that can be incorporated into a particular stage of a journalist’s routine—might cause confusion as far as the needs and capabilities of AI tools are concerned.

3.5. The Hybrid Model of Self-Learners and the Lack of Strategy

Most journalists who participated in our survey are self-learners. The media organizations they work for do not offer education and training programs, and more specifically, they do not seem to have a strategy for AI implementation, neither in the journalistic workflow nor in other functions, as the journalists note in our research. Despite the volume and the intensity of AI tool use, those who make progress complain about the lack of equality regarding their colleagues in Europe; they must devote extra time to learn new tools and practices and pay for seminars, as there is “a lack of strategy and vision in the majority of the media organizations”. This finding agrees with a previous quantitative survey regarding 148 professionals representing 86 media organizations examining the state of digital transformation across four European countries: Greece, France, Portugal, and Cyprus [62]. The findings revealed a significant absence of digital transformation strategies within Greek media organizations, with 53.5% of Greek respondents affirming this gap.

Although AI could transform the field of journalism, it is expected to complement rather than replace the work of journalists. Unanimously, journalists ask for training, not only on how AI works but also on how they can practically apply tools to their daily needs and routines. According to our findings, training should be extensive, covering different tasks and newsroom workflow needs. As our research has shown, different types of media (TV, radio, podcasts, and websites) and their journalists have different needs and workflows.

Most importantly, the effective deployment of AI in journalism necessitates continued human supervision. A notable challenge arises from the knowledge and communication gap between technologists who design AI systems and journalists who employ them. This gap can result in journalistic malpractice if not adequately addressed.

“We need to know how the tools are trained and there should be a dedicated journalist in the newsroom who will have AI governance responsibilities. Somehow like the DPOs work” mentions one participant. The BBC has introduced new editorial policy guidelines concerning the use of generative AI and, in line with this initiative, has appointed two dedicated roles: an Editorial Executive for Generative AI and a Programme Director for Generative AI. These individuals are tasked with overseeing the implementation of the BBC’s AI-related policies [63]. Similarly, Reuters has appointed a Head of AI Strategy, responsible for defining and executing the organization’s overall AI strategy. This role includes leading AI initiatives aimed at enhancing content production, distribution, and personalization across the organization [64].

However, only a few Greek journalists are familiar with these technologies, laws, and ethics. Here, a need arises for separate expertise within the framework of news media organizations. An AI ethics and governance officer is needed to permit the use of AI tools. There is also a need for transparency regarding the audience.

Ethical considerations concerning data use remain paramount. Journalists must address issues related to the collection, storage, analysis, and sharing of user information.

Transparency in methodology is imperative when using AI tools in journalism. Readers have the right to understand how these tools were employed in processes, such as analysis, pattern recognition, or reporting findings within a story. According to the EU AI Act and its implementation in journalism [58], AI-produced content should be labeled or forbidden in specific cases.

The PFJU (Panhellenic Federation of Journalist Unions) created a code of ethics aimed at satisfying the need for AI guidelines in the Greek media landscape. Unions and media organizations all over the world have published guidelines for newsrooms, but isomorphism—adopting similar practices, even when organizations lack the necessary capacity or operate within significantly different contextual environments—is an issue [48]

The inherent unpredictability of AI systems complicates efforts, as major problems may emerge. This unpredictability underscores the need for vigilance among both technologists and journalists to ensure these systems are responsibly monitored and managed. Media organizations such as The New York Times have passed editorial judgment to AI in many steps of their algorithmic functions [34]

Greek journalists are not well trained or well-equipped regarding AI and its tools’ strengths and weaknesses. Thus, RQ3 is not confirmed. Training is urgently needed not only concerning the use of AI in newsrooms as an assistant for journalists but also to provide journalists with skills to stand against malpractice when using and/or dealing with AI-produced content in fact-checking. According to our findings, there are different needs for hands-on training to fill existing skill gaps and to satisfy different needs for different types of media. However, one major issue is a lack of strategy within Greek media organizations. If these organizations cannot quickly understand this major shift, their journalists and media will face many challenges, which might affect their viability.

4. Discussion

Our findings suggest that there is a lack of strategy and, therefore, AI integration in Greek media organizations. A few pioneering organizations are collaborating with international technology providers such as Google, widening the gap within the domestic media sector. However, AI use is limited in specific segments of news workflows, and linguistic barriers have become an important issue. In addition, AI is a supplement for missing skills, such as video editing and subtitling. Notably, the use of AI by journalists in these organizations for other tasks and purposes is sporadic and personal. It is not officially integrated, and most of these journalists use free tools. Indeed, third-party tools are also used by Greek media and newsrooms. We found that tech-savvy journalists are pioneering in the use of AI tools, while curious journalists are self-trained in automation systems. These tools are mostly used for repetitive and boring tasks, such as transcription and translation, and a few sophisticated tools are used for image and video production areas in which journalists lack skills. Greek journalists are not using tools for news writing or rewriting, which they believe is the “Holy Grail” of the profession. A few of them are using tools for story ideas and research, but the Greek language is still a barrier since the outcomes of these tools are, in many cases, poor.

The PFJU Code was implemented in May 2025, making it premature to fully assess its effectiveness and significance. However, both the Greek government and the Union of Website Owners have publicly committed to adopting the code.

Although Greece is classified as a Global North country, it ranks 20th out of 23 Western European nations in the 2024 AI Readiness Index2, with a score of 57.7—significantly below the regional average of 69.56. Eastern European countries have an average score of 57.8.

Investing in the training of Greek editors and reporters is essential as AI tools are integrated into newsroom practices. Our findings suggest that training is crucial for all members of the newsroom and should be extensive, as AI tools can be integrated into separate tasks. However, all of them are part of overall news production. Greek newsrooms are small. Forty percent of the top 40 news websites in terms of web traffic operate with journalistic teams consisting of only 10 members [14], and a lack of human resources is a major issue. This means that most journalists deal with many different tasks, such as newsgathering, verification, text creation, image and video creation, and translation. More research is needed to break down the workflows in all media types. This will result in a targeted training scheme and will probably be more beneficial for journalists and media organizations. The issue is not age or background but attitude and stance regarding technology.

There should be training on tested tools that provide quality results in the Greek language so users will not be disappointed, as disappointment increases the rejection of AI tools overall. A persistent challenge is the hidden biases of AI, which, although frequently overlooked, are inherent due to human programming. Newsrooms must commit to mitigating these biases through transparency in reporting.

Journalists must acquire the skills to use these new technologies both ethically and efficiently to enhance storytelling. Developing and disseminating shared guidelines for ethical data use and the transparent public disclosure of methodologies is equally imperative for fostering collaboration between journalists and technologists. Initiatives such as the Code of Ethics of the PFJU may prevent the possible indiscriminate use of AI in media organizations for journalists and citizens.

AI applications offer valuable opportunities to reflect on and integrate editorial standards into the development of journalism-specific technologies. Custom-built AI solutions often exceed the financial capacity of smaller organizations, but forming partnerships with academic institutions could provide a viable alternative, such as the one between Makedonia and Greek academia.

Funding initiatives are an opportunity for the Greek media and journalists but should be extensive and not limited to specific media. Intensive tailored training for journalists and the media will provide a larger spectrum for the faster and more inclusive development of AI in the Greek media ecosystem.