Artificial Intelligence in Slow Journalism: Journalists’ Uses, Perceptions, and Attitudes

Abstract

Through long-form, creative, high-quality stories, slow journalism seeks to counteract the effects of speed and immediacy in news production and consumption primarily driven by technological advancements. The advantages of artificial intelligence (AI) in journalism include generating and enhancing content, reducing workloads, and consequently giving journalists more time for non-routine and creative tasks. This raises the question of where AI fits into slow journalism. Twenty-one semi-structured interviews were conducted with practitioners of slow journalism in Spain to explore their use, attitudes, and perceptions of AI in their work. The findings indicate that the interviewees make rudimentary use of AI tools, and their attitudes range from a slight lack of interest to a willingness to learn more about them, alongside concerns regarding ethical boundaries and the potential for job losses. They assert that they have a moral and human responsibility when producing stories that AI cannot enhance in terms of quality, creativity, and emotional depth. It can be concluded that AI offers little to ‘slow’ journalists due to the significant limitations in enhancing long-form reporting. At most, it may enable them to streamline repetitive and non-creative work, thereby allowing the depth required in slow journalism, at least in its current state of development.

1. Introduction

The fast pace at which information is produced and consumed in a constantly changing digital world causes information overload, which in turn leads to continuous updates, fewer sources being consulted, poorly verified facts, and increasingly standardized journalistic narratives (Soliyeva 2024). In response to this, the concept of ‘slowness’ has gained traction across various media and communication practices in recent years, including slow communication, slow reading, slow blogging, slow television, slow news, and slow media (Craig 2020). Slow journalism, in particular, aims to counteract the effects of speed and immediacy in news production and consumption by using technology and media in a more mindful, critical, and reflective way (Barranquero-Carretero 2023). Ultimately, it takes deeper approaches that result in long-form, creative, good-quality stories (Greenberg 2012).

On the other hand, technological development has seen the rise in artificial intelligence (AI) in the media. From content creation to broadcasting and audience engagement, AI entails a major shift in journalism. It poses challenges, such as a range of ethical issues (Shi and Sun 2024) related to automated news writing, as well as broader societal concerns (Thurman et al. 2017), and potential threats to the profession (Peña-Fernández et al. 2023). Benefits include helping to generate and improve content, reducing workload and, consequently—and presumably—giving journalists more time for non-routine and creative tasks (DalBen and Jurno 2021; Simon 2024). The latter raises the question of where AI fits into slow journalism, a new perspective within the academic output on journalists perceptions on AI (e.g., Cools and Diakopoulos 2024; de Haan et al. 2022; De-Lara-González et al. 2022; Mayoral-Sánchez et al. 2023; Noain-Sánchez 2022; among others), which is addressed in this study.

1.1. Slow Journalism: Concept and Academic Approach

The US manifesto Not So Fast, published in The Wall Street Journal in 2009, and the German Slow Media, published online in 2010, lay the foundations for slow journalism. These documents were calls against the culture of superficiality and rush in the media and they advocated quality over quantity, diversity of voices, sustainability, transparency and active user participation (Greenberg 2007). Slow journalism is more a promoter of reflection that enhances journalism’s core mission in society (Min 2021) than just a term with a set of established characteristics (Le Masurier 2015). It is often known as in-depth and narrative journalism, and it shares similarities with other reform movements that came before it, such as New Journalism, which was an attempt to reconsider what good journalism is using long-form, personal, and narrative styles (Min 2021).

Most of the emerging slow journalism outlets have assumed an ‘alternative’ journalism role, taking time to investigate more focused issues, and using narrative storytelling (Le Masurier 2015). News magazines and book-length journalism, for instance, “do not only function at different speeds from more daily forms of journalism but they can also claim authority and prestige within the journalistic field because of their slower, more analytical approaches” (Craig 2020, p. 468). Other outlets diversify revenues without relying only on advertising. They adapt formats to each platform and develop communities of readers who identify with their product (Albalad Aiguabella 2018). An example of this is The Washington Post launching Deep Reads, a home for immersive and narrative reporting where stories are labelled at the top of the article for readers to easily access them and pieces are accompanied by human-read narration (WashPostPR 2023). In Spain, the newspaper El Periódico de España launched a special newsletter, offering more in-depth narrative stories and revealing how they were created with a view to, they claim, achieving higher quality journalism and strengthening links with the audience (Laboratorio de Periodismo 2023).

In academic terms, it was in 2007 that Susan Greenberg, a professor at the University of Roehampton, first adopted the term ‘slow journalism’. Founded on ‘the end of the middle’ concept of marketing, she argued that people obtain basic news cheaply on air and online, whereas in the middle is traditional print journalism. At the other end is slow journalism, which is composed of “essays, features and other non-fiction writing and takes the time to find things out, looks at stories that others miss, and communicates it all with the highest quality” (Greenberg 2007) as well as telling readers where the information comes from and how it was gathered. From then on, the idea of slow journalism began to have a space in scholarly research (Craig 2020; Greenberg 2012; Le Masurier 2015, 2016; Neveu 2016; Rauch 2011; Rosique-Cedillo and Barranquero-Carretero 2015).

Many studies have addressed the effects of speed and use the concept of ‘slow’ to refer to a way some emerging media outlets are producing journalism and experimenting with new business models (Le Masurier 2016). Journalistic handicaps like haste, lack of sources, simplistic and stereotyped language, and lack of context and time prevent journalists from reflecting on the facts, length to explain the events, depth, and time to elaborate quality narratives (Gómez-Mompart 2021). Hence, some newsroom innovators are pausing to consider what producing good journalism means (Min 2021, p. 89). They conclude that slowness helps journalists respond to the complexities of modern public life (Craig 2020) and fulfil their social function. They stress the need for less ‘fast food journalism’ and more slow journalism. That is, a more relaxed, elaborated, contrasted, better oral–written–audiovisual and infographically narrated (Gómez-Mompart 2021), and teamwork journalism (Gómez-Mompart 2021; Le Masurier 2016).

There is a growing consensus that only two types of media that move at different speeds will survive in the future. On the one hand, the mass producers of short news. On the other hand, the minority ‘alternative’ media that focus on delivering hard-earned information (Boynton 2015; Craig 2020) to an audience that has time and money and values quality, as opposed to fast-moving consumer journalism aimed at massive audiences. Although less common, slow and fast journalism could even co-exist within one publication. The Guardian’s introduction of in-depth reporting, essays, and profiles under The Long Read section is “an example of a commitment by a major, mainstream news company to cater to different temporalities of news production and consumption” (Craig 2020).

1.2. Intersection of AI and Journalism

Studies have highlighted the efficiency gains that AI can bring to newsrooms as well as the ethical and quality challenges it poses in an evolving media environment (Clerwall 2014; Diakopoulos 2019; Túñez-López et al. 2019; Parratt-Fernández et al. 2021). AI can be used for generating basic news stories, particularly in sectors with structured data, such as finance, sports, and weather; and to assist with journalistic tasks like research, selection, verification, composition, finding connections in large datasets, and the presentation and distribution of news (Graefe 2016; Simon 2024; Thurman 2019). News executives and managers underscore the use of AI for back-end automation tasks (56%) such as transcription and copyediting as a top priority, followed by recommender systems (37%), content creation (28%) with human oversight, and commercial uses (27%) (Newman 2024, p. 27). However, the positive side of these applications is not so clear among editors, who remain wary of the potential job losses to which the implementation of AI in media outlets could lead (Mayoral-Sánchez et al. 2023).

In fact, the fear of job cuts is pointed out as the main reason for rejection of AI among practitioners across the world (Beckett 2019), who also express a desire to retain control at all stages of news production (Wu et al. 2019). AI can reduce the time journalists spend on repetitive tasks—such as fact-checking and data analysis—and they will have more time to spend on work that cannot be automated (DalBen and Jurno 2021; Wu et al. 2019; Young and Hermida 2014). But it also raises the potential threat of journalists being replaced by AI as it becomes increasingly proficient at tasks once handled by humans (Marconi and Siegman 2017). In the efficiency discussion, there are those who believe that AI will allow journalists to focus on more creative and strategic functions while technology takes care of grunt work. Others are more skeptical, arguing that AI’s impact on productivity is likely to be more limited and varies depending on what is being automated (Simon 2024). In the debate on whether to see AI as a competitor or a partner, journalists generally choose the latter (Peña-Fernández et al. 2023), although recent studies have claimed that there are no guarantees that this will remain the case (Simon 2024).

Furthermore, journalists’ role has undergone a profound metamorphosis, and they are now expected to be proficient in the use of AI tools in addition to traditional investigative journalism skills. This convergence of abilities results in more hybrid jobs (Deuze and Beckett 2022) and close collaboration between journalists, data scientists, information designers, and computer programmers (De-Lara-González et al. 2022). However, while data streams can offer historical context, static facts, and sentiment analysis, they are unlikely to grasp the full complexity of human expression, which shapes how events are reported or identifies the most crucial angle of a story. As a result, automation might actually heighten the demand for the essential human qualities that define good journalists—news judgment, curiosity, and skepticism (Thurman et al. 2017).

1.3. AI and In-Depth Journalism

As mentioned above, automated content production has mainly been used for algorithm-generated news articles (Dörr 2016), for which interpretation and explanation are not required (Diakopoulos 2019). But many authors argue that AI can now offer benefits for other types of journalism as well. Among them, enabling deeper investigation (Schapals and Porlezza 2020), contributing to add more complexity and meaning to texts (Brennen 2018), and enhancing creativity, listening, and sourcing (Guzman and Lewis 2019). They say it can also generate news stories according to the style of news organizations and even semantically reframe them for different audiences (Shi and Sun 2024). This could not only increase journalists’ productivity, but also improve the quality of news reporting (Shi and Sun 2024), while addressing information issues like overload or loss of credibility (Ali and Hassoun 2019).

Despite the potential benefits, AI still has significant limitations in improving some long-form reporting (Nishal and Diakopoulos 2024). Its ability to contribute to better quality is disputed (Simon 2024) because, according to Gómez-Mompart (2021), the long-term sustainability of quality journalism will depend on those who have creativity and intelligence, among other abilities. AI cannot offer any of this to an audience (Graefe 2016) that particularly values the narrative quality and context of human-written articles. The rise in AI in journalism also entails ethical and professional concerns. One of the most pressing is the potential erosion of journalistic integrity and human touch in news reporting, since AI systems lack the nuanced understanding of context and ethical judgement that journalists bring to their work (Diakopoulos 2019).

It is also questioned whether robots will be able to fully replace journalists (Lindén 2017), considering that journalists’ professional craft is not just ordering or adding specific data (Gómez-Mompart 2021) but is a most human endeavor, as is reporting (Anderson 2013). Moreover, some wonder if the efficiencies gained from AI will lead to an expectation that journalists will produce more content. In essence, the point is whether AI will bring an increase in the quality of journalism or simply in its quantity (Simon 2024). Gómez-Mompart (2021) argues for the recovery of the best in-depth journalism, narrative and/or investigative, “relying more on competent humans than AI machines in terms of quality and emotion”. Others, such as Graefe (2016), believe that a hybrid approach may be the most effective.

In this context, in this study, the aim is to explore the place of AI in slow journalism in Spain. The following questions are raised:

RQ1. How are journalists using AI in slow journalism?

RQ2. Do journalists see AI as an efficiency gain in the production of slow content, or as a threat to creativity and quality?

RQ3. How do journalists see AI affecting the future of slow journalism?

As for the theoretical basis for this study, according to Ayyad et al. (2023), the technology acceptance model (TAM) has two primary factors that influence users’ perceptions and attitudes towards using a technology. Firstly, “perceived ease of use” refers to the degree to which someone believes that using a particular technology would be effortless. Secondly, “perceived usefulness” refers to the degree to which a person believes that using a particular system would improve his or her job performance. In addition, TAM has been theoretically extended to include “job relevance”, which refers to an individual’s perception of the degree to which the system is applicable to his or her job, and “output quality”, which refers to the quality of the result produced by the system.

Additionally, Wu et al. (2019) draw the following three theories of how AI will change the news business: (1) AI replaces journalism, (2) AI swallows journalism, and (3) AI improves journalism. The latter believes that AI tools, such as GPT, will simply make certain journalistic operations more productive, partially automating processes and addressing concerns about accuracy, bias, and plagiarism by emphasizing the final role of the human writer.

2. Materials and Methods

This research is based on the qualitative method. The technique used was the semi-structured interview, specifically interviews with slow journalism practitioners. In line with Guion et al. (2011), the interviews seek to understand phenomena from the point of view of the protagonist, and they are used for collecting data based on opinions, feelings, emotions, and experiences that “need to be explored in depth and in detail rather than simply reported in a word or two” (Denscombe 2010, p. 174). Before identifying the interviewees, a purposive sampling (Krippendorff 1990) of representative slow journalism publications in Spain was carried out, taking into account the following criteria established by Rosique-Cedillo and Barranquero-Carretero (2015): (a) non-daily reference publications that set their own news agenda independently from mainstream media, prioritizing slow forms of investigative and/or narrative journalism; (b) the average length of their pieces is over 2000 words; (c) they either describe themselves as slow journalism or call for quality journalism amid the dominance of fast culture.

The supplement El País Semanal, Jot Down quarterly magazine, and the publisher Libros del K. O. were chosen because of their different formats—and speeds—and their well-established reputations. Since 1976, El País Semanal, the Sunday supplement of the El País newspaper, deals with current affairs, culture, and society in an in-depth way. Its debate generation has made it one of the most prestigious publications in Spain. Jot Down, a magazine born in 2011 in digital form before moving to paper, stands out for its timeless, in-depth, analytical and slow, reflective journalistic approach, and for its interviews with well-known personalities (Rosique-Cedillo and Barranquero-Carretero 2015). Its distinctive style, editorial excellence, quality writing and visuals, have turned it into a reference in the Spanish cultural scene with international recognition. As for Libros del K. O., it publishes different literary genres, as well as slow journalism both by renowned and emerging authors and journalists and stands out for its literary quality and careful book design. It supports independent voices and commits to unconventional topics at national and international levels (Libros del K. O. n.d.).

An intentional selection of professionals was then made according to three criteria. Firstly, the journalists with leading positions in the three media mentioned above. Secondly, the periodicity of the publications was considered. Thirdly, the aim was to achieve as much parity as possible. In some cases, after initial contact by email, it was not possible to arrange an interview, and it was necessary to contact another professional. Thus, nine journalists from the supplement El País Semanal, five from Jot Down magazine and four from the publisher Libros del K. O. were interviewed. Ultimately, a total of 21 respondents—three leaders, three staff editors and fifteen occasional contributors, of whom nine were women and twelve were men—were interviewed (

Table 1. Slow journalism professionals interviewed.

No.	Gender	Position	Publication	Code
1	Female	Journalist in leadership position	El País Semanal	L1
2	Male	Journalist in leadership position	Jot Down	L2
3	Male	Journalist in leadership position	Libros del K. O.	L3
4	Female	Staff editor	El País Semanal	S1
5	Female	Staff editor	El País Semanal	S2
6	Male	Staff editor	El País Semanal	S3
7	Male	Occasional contributor	El País Semanal	C1
8	Male	Occasional contributor	El País Semanal	C2
9	Male	Occasional contributor	El País Semanal	C3
10	Female	Occasional contributor	El País Semanal	C4
11	Male	Occasional contributor	El País Semanal	C5
12	Male	Occasional contributor	El País Semanal	C6
13	Male	Occasional contributor	Jot Down	C7
14	Male	Occasional contributor	Jot Down	C8
15	Female	Occasional contributor	Jot Down	C9
16	Female	Occasional contributor	Jot Down	C10
17	Female	Occasional contributor	Jot Down	C11
18	Male	Occasional contributor	Libros del K. O.	C12
19	Female	Occasional contributor	Libros del K. O.	C13
20	Female	Occasional contributor	Libros del K. O.	C14
21	Male	Occasional contributor	Libros del K. O.	C15

).

Conducting semi-structured interviews requires prior elaboration and flexibility, where the sequence and formulation can be adapted to each interviewee (Rubin and Rubin 2011). For this reason, seven dynamic questions (Appendix A) in a friendly language and closer to the interviewees (Kvale 1996) were developed. Some were shared with all participants, while others were adapted to the specificities of each one. The interviewees were asked, among other things, about their use of AI tools, their perceptions about AI in slow journalism, and how AI affects their work. The interviews were conducted between March and July 2024 via the Zoom platform, lasting between 30 min and 4.5 h, in most cases over two days. All interviews were recorded with the consent of the participants and then transcribed using Converterapp software (https://converter.app/es/, accessed on 28 November 2024). As for the qualitative data analysis method, the transcripts were manually edited and coded to organize the responses according to thematic patterns.

3. Results

The results are shown below according to the thematic criteria used to analyze the interviews.

3.1. Integration of AI into ‘Slow’ Journalistic Work

Among the 21 slow journalism practitioners interviewed—three journalists with leading positions (L), three staff editors (S), and fifteen contributors (C)—eleven reported that they used AI tools in their work. Leaders—two out of three—and contributors—eight out of fifteen—were the most inclined to integrate AI into their daily tasks. In contrast, only one staff editor did so (

Table 2. Slow journalism practitioners who use AI in their work.

Journalists	Use of AI
L1	Yes
L2	No
L3	Yes
S1	No
S2	Yes
S3	No
C1	No
C2	No
C3	Yes
C4	Yes
C5	Yes
C6	Yes
C7	No
C8	No
C9	No
C10	Yes
C11	Yes
C12	Yes
C13	No
C14	Yes
C15	No

).

Although slightly more than half of respondents said that they frequently used AI in their work, most of them were cautious about the changes brought about by the advent of these tools. In fact, many of the responses contained phrases such as “for the time being” and “up until now”. This caution is also found in the ways they described their use of AI, “People tinker a lot, they try it out, but it isn’t very reliable at the moment” (L1); “I did some tests” (S1); “I just experimented with it” (C6); “I tried several” (C7); “I tried ChatGPT, but it was disappointing” (C9); “I started playing with Midjourney” (C13).

Practitioners’ awareness of the rapid development and adoption of AI technologies supports the idea that the technologies have great potential and, like it or not, they will eventually become part of the daily practice of journalism. For C2, “AI tools will be standardised, and this will benefit journalism, as has happened with many technological developments”. C1, on the other hand, believes that “they’ll end up complementing and benefiting not only basic news writing, but also deep reporting”. C7 has a similar opinion, “Maybe in a few years they’ll improve to the point where they can write good reports, slow journalism and novels”. However, the evidence suggests that the use of AI tools is currently limited to a few very specific tasks and the outcomes do not consistently meet quality standards. For example, eight of the eleven professionals who used AI do so to transcribe interviews, and four to translate texts. Suggesting keywords and headlines, proofreading, and creating illustrations and layout are performed by one out of eleven. The most sophisticated application, the writing of press releases, is a task also performed by just one (Figure 1).

Transcribing interviews is therefore the task that most professionals entrust to AI, mainly for two reasons, that it frees them from tedious labor, and it saves time. “I used to hate interviews just because I had to transcribe them, so these tools help me a lot”, says S2. C6 agrees: “It’s invaluable for transcribing interviews because it saves time and speeds up work”. The most remarkable aspect for C11 is that “you only have to check for errors in the transcription, you don’t have to do it all from scratch”. It is precisely these errors that C7 highlights as the reason why, despite having tried AI tools, they do not use them, “It produces a text full of typos and misunderstandings. Having to listen to the recording again and revise the text doesn’t really save much time. I haven’t found it useful at the moment”.

Not all professionals who choose not to use AI for transcription cite the poor quality mentioned before. Some point out reasons related to the very nature of storytelling and the purpose of journalism, especially when it comes to slow journalism. For example, C2 is categorical, “I wouldn’t even think of using a transcription and translation app for interviews. I have hundreds of hours of transcriptions that take up a lot of time and it’s horrendous. Still, in no way would I use an AI for that, because I lose information that can be used to build scenes. I lose hesitation, a cough, the sound of birds or a coffee machine. All that is fundamental to the story”. The same goes for S3, “I’ve used it a few times, but I didn’t like it. Only if I was in an extreme situation or felt very lazy, I’d use it. It’s more work transcribing myself, but it’s a kind of ceremony. Having an AI do this is like a filmmaker having a film edited by a machine. Transcribing and listening are part of the story, they put me into it, it’s like warming up before a game”. C4, although more flexible, also highlights the advantages of listening to the recording and transcribing it themselves, “I like to transcribe manually sometimes because that way I remember things”.

Following interview transcription, text translation is the task most frequently performed by interviewees with the help of AI. AI’s usefulness is associated with functions considered supplementary to journalistic work, such as reading documents and information, and checking specific expressions. For instance, C5 mentions using AI to verify “certain details” when writing texts in English, though they note: “I haven’t yet made it part of my regular work process”. S2 found AI translation particularly valuable while working as a foreign correspondent and needing to read information in unfamiliar languages, “When I first arrived in the country, I didn’t know the language, so translation apps were essential to start reading local newspapers. I still use them to translate words, in addition to the dictionaries I turn to all the time. The translations might not be perfect, but they give you a general idea”.

Regarding other tools, L1 confirms that their newsroom does not generate text with AI, but it is used to generate illustrations “only when the subject requires it”. L3 makes it clear that they turn to AI “rarely, only to carry out tasks that are repetitive and tedious”, such as “suggesting highlighted phrases”. Generating titles for audiovisual pieces is something that C11 has occasionally asked the AI to do, “I do it because it is the hardest thing for me”. C10 uses it to layout texts and C6 to correct them, “For grammatical corrections I use Grammarly, which immediately detects errors or suggests a comma. I don’t use it in a sophisticated way, just to facilitate time-consuming tasks”. Finally, C14 says they have used ChatGPT for press releases, “I think it’s a fantasy that this can be done”.

3.2. Changes Brought About by Integrating AI into Slow Journalism

Except for two journalists (C6 and C11) who noted time savings in transcription, all interviewees agreed that AI has not changed their work so far. However, they have no doubt that substantial changes will come in the near future, when the tools improve. Two of the three professionals with leading positions believe that it is important to have a document to regulate the use of AI in newsrooms. L1 says, “Although AI is just beginning, it would be ridiculous not to consider it. We have a team working on how we should use it, and we want to include a section on it in our style guide”. L3’s newsroom has already published a guide “stating that we’ll use AI as a tool that’ll always be supervised by humans and won’t replace anyone’s work”. But this is not a priority for L2, “We don’t intend to create a regulation now; I don’t think people are very interested in that. If we were going to use it, we’d say so […]. We have 180 different writers, and AI has hardly changed how they work. I’m not saying they don’t use it, but they’re still going to places, talking, documenting, contextualizing, choosing the beginning of a sentence or a metaphor… I have the feeling that although it’s an extremely important tool, it still doesn’t have the weight that it’s given in public conversation”.

According to the testimonies collected, neither the way of investigating nor the way of presenting information has been affected by the use of AI yet, and the changes in journalists’ routines are minimal. These new technologies are still recent, and it will take time to get to know them and integrate them into journalism. This is what L1 thinks, “There’s always a lot of confusion and fear at first when there are big technological leaps”. C1: “The point is to learn to live with them and to handle the advances”. And S2 says, “We have to see where it goes and what we want to make of it. There is a part of collective responsibility and an individual part. We have to adapt it to our needs so that it serves the best journalism we do”.

As reported by most interviewees, their contact with AI does not go beyond the individual use of free tools with which they fiddle. Only two (C7 and S2) claim to have paid versions, “We’ve built a number of very useful tools into the editing platform that help. For example, they provide alternatives to more clickbait-friendly headlines, or 600-word agency press releases” (C7). The majority of those who use the more limited free versions are aware of the errors they can generate, “My experience is that they’re not entirely reliable. You have to use the paid ones to be sure the data they provide is correct and not misleading” (S1); “I haven’t paid for any yet, I’ve just been experimenting, but if I ever needed to, I’d definitely pay” (C6); “I’ve been changing tools, but I always try to use the free ones’ (C4).

The impact of AI on daily work not yet being noticeable also has to do with the lack of quality and the limitations of AI tools. Again, leaders (L1: “Maybe in five years we’ll forget about the original recording when automatically transcribing, but we’re not there yet), editors (S1: “At the moment, AI is not developed enough”), and contributors (C14: “The challenge is that you need a critical perspective to use these tools”) agree. Only C7 seems more convinced that journalists will eventually be replaced by AI, even in their work as copywriters and in in-depth news stories and long interviews, “According to technology experts, these machines may well be able to replace humans in ten years’ time. Maybe not completely, but I’ve no doubt that with a good AI, an editor-in-chief could go from having ten editors to none”.

Is there any aspect that they feel will not change? Except for one of the interviewees who claims to use AI to write press releases (C14), most only use it for tasks that do not involve text writing. Some are even clearly against doing so in the future, “Using tools that make your job easier is one thing and having an AI writing news articles is very different. I don’t think that’ll ever happen in my newsroom. I can ask the AI to draw three ideas from a human-written text for a Twitter version, but by no means can the mother piece be written by AI. A quality newspaper wouldn’t do that because it’d be anti-journalistic” (L1). Leader L3 agrees, “We decided that our use of AI would be ethical—it says so in our guide—, which means it’ll help us do repetitive work, but it’ll never replace anyone’s work”. For C10, using it for copywriting is “absolute nonsense”. If it comes to the point where copywriting is no longer journalists’ regular activity, C14 says that “we’ll still need professionals to think, to find stories, to find clever approaches. Only a human can do that”.

3.3. AI in Slow Journalism: Increased Efficiency or a Threat to Creativity and Quality?

As seen, most interviewees are still reluctant to use AI for news writing, but they do believe that it will become the norm in the not-too-distant future. It is a different matter when they are asked about AI in slow journalism. To write ‘slow’ texts, which are more in-depth, time-consuming, in a freer and more creative narrative style, and not always linked to breaking news, most of them do not currently use AI tools neither do they believe that they will do so in the short term. The reasons given have to do with the very nature of slow journalism, “One of the premises of our magazine is to report from where news events take place. That’s why you can’t have AI in long-form stories”, C13 asserts. Similarly, L1 says: “Journalism consists of going to a place, being with people, interviewing them and reporting. That cannot change because it’s the essence of respectable media”.

Another stumbling block is the belief that AI cannot replace observation and the human point of view, “Narrative journalism is by definition humanistic. In principle, we journalists are the ones who have the privilege of being able to pursue it” (C12). Likewise, S1 says “AI can help with certain slow journalism issues, but not all. What I appreciate about renowned professionals is their ability to connect emotionally. I don’t think AI can do that”. The human element, people’s sensitivity and their emotional ability, is again an argument when it comes to distinguishing between the tasks that an AI can do and those that journalists will still have to undertake in slow journalism. In this sense, C15 underlines professionals’ aims when approaching certain issues, “You can use AI tools for mundane things like searching for documents or transcribing interviews. But at the end of the day, it is the journalist who has to write the text and decide what the intention is. You can write a guideline for a washing machine with AI, but a long piece like Leila Guerriero’s, I very much doubt it”.

Interviewees C7, S2, and S1 refer to the ability to connect with other people and build a relationship with them taking into account their circumstances, “When it comes to slow journalism, to in-depth reporting, AI isn’t able to write or find the right sources. I wish it could interview people, but that’s a long way off” (C7); “I find it hard to imagine asking an AI to produce a long story. I’m one of those who still look for people, I interview them, I ask them to suggest other people… I would use AI to search large databases, but once I had that, I’d do journalism like in the 19th century, like in the Crimean War” (S2). S1 does not rule out the possibility that AI tools will eventually generate narrative writing but does not think they will ever behave like humans, “It can be a threat, but it hasn’t developed the ability to understand 100% of individual’s complexity yet. I’d like to believe it’ll make life easier for us and that instead of working sixteen hours a day we’ll work eight, which is what we should be doing”.

According to L1, creativity being inherent to slow journalism makes it impossible for an AI to replace it, “Readers aren’t stupid and want to read a story and be surprised, they want to see a new dimension that they may not have seen before. AI cannot go somewhere and describe what it smelled like or what the temperature was or how a character behaved. Summarising is a different thing altogether”. This is why L1’s newspaper backs the idea that slow journalism must be accomplished by humans, “You have to discover something new—a person, a place, a fact, or an approach. Our reporting isn’t a rehash of existing information. It’s very creative work that AI can’t do”. Even C11, who believes that AI tools are capable of producing original work, thinks they will not be able to replicate human qualities, “A machine has a different conception of creativity from an author’s voice”. C15 goes a step further and questions how readers would react to a story whose authorship and intentions are unknown, “I’d find AI writing an interview or a feature story strange for two reasons: I’d find it odd that it captures the atmosphere and the details; and when I read a text I often wonder who’s written it, why and who the writer serves. I don’t know who AI serves, and since there’s no signature, I’ll think something is being hidden”.

However, some respondents are confident that when AI tools become commonplace, they will have more to offer. This is the view of C14 and C9, “AI will encourage more and more creativity. It’ll be able to write texts, but we’ll still need to think, to find stories and ingenious approaches. It won’t do that” (C14); “ChatGPT lacks creativity, but it can enhance yours. If you have a fantastic idea, it’ll give you a summary. It’s amazing for that. Then you adapt it to your style” (C9). C1 is the only interviewee optimistic enough to imagine an AI capable of describing places, contextualizing situations, and even describing smells, “I’m convinced that AI will allow for better reports. It may not be able to describe the smell in Gaza now, but we have to give it time. In fifty years, it’ll be able to perfectly describe the environment and the context. In a century’s time all the data, the atmosphere and the details of a story, will be supplied by a small device”.

3.4. Perceptions About AI Among Practitioners Who Have Not Used It

As seen above, not all journalists are adapting to AI in the same way. Ten of the twenty-one respondents say they have never used it in their work for various reasons. For C5—who does use it—“journalists are afraid because all new technology has taken its toll. There’ve been many overlapping crises, and this one looks bad too”. In the same vein, C1 says that “Journalists tend to be reticent to any kind of progress. They always see it from a negative perspective. But in modern journalism, especially in big newsrooms, many employees aren’t journalists. It’s important to have engineers, statisticians, illustrators, philosophers”. Indeed, the implementation of AI in newsrooms has led to building multidisciplinary teams that include professionals from outside the field of communication. For C2, these are the ones who use it the most, whereas journalists are still reluctant, “Some people use AI for everything. But journalists don’t. I cannot see my colleagues using AI against their profession in 2024”.

One interviewee (C10) who has integrated AI into their work, notices “quite a lot of doom and gloom” among colleagues, “If someone thinks your job can be done by a machine, we’re off to a bad start. Some people don’t admit they use it, because it’s frowned upon right now. Thinking it’s going to eliminate jobs is something that’s happened whenever there’s been technological change”. The fear of significant job destruction lies behind the reluctance of some professionals to embrace AI tools. ‘Fear’ and ‘vertigo’ are the most common words used. “It makes me dizzy. I think journalists, like translators and illustrators, could disappear. They’ll have to defend their work a lot to convince their bosses that they can do it better than a machine” (C8).

In addition to the concern about the potential effects in the workplace, some professionals—as seen in point 4.3.—believe that AI is not (yet) applicable to the storytelling of slow journalism. C13 explains that “Many are afraid of the unknown, of what might happen. It moves very fast, and the first reaction is rejection. But I don’t use it because one of the premises of our magazine is to report from where facts take place, which makes the use of AI unfeasible”. Another interviewee (C15) does not believe that AI can help with ‘slow’ writing yet either, “I don’t use it because I think I can do it much better. If one day it’s the other way round, I won’t say a word. With the current state of these tools, if you write with AI and it’s accepted, something has gone wrong along the way”. This refusal also shows that there is still mistrust, “I don’t use them to avoid feeding the dragon, because it grows with the work of others. I have no rational justification; I just don’t like them”. Still, this attitude does not mean that the interviewees are unaware of the future of journalism being linked to the development of AI, “If you bring up what I said in fifteen years’ time, I’ll look like a fool”.

As C1 states, the point does not seem to be whether AI tools will be used by journalists in the coming years, but how they will be used, “We need to see how we can integrate them to improve our stories. It’s not about us journalists writing them, it’s about making better content so that readers get better quality and better material. And if the stories are better with the help of AI, then go ahead, the reader will appreciate it. I’ll be sorry it’s not done by humans, but that’s not the point”.

3.5. Do ‘Slow’ Journalists Think Readers Recognize AI-Generated Content?

According to most of the professionals interviewed, readers can distinguish AI-generated texts from those written by journalists because AI tools are not sophisticated enough to produce quality content. However, when using expressions such as “for the time being” or “still”, they also show their belief that this will change. “Readers are aware of it today, but the outcome may be perfected and indistinguishable in two or four years”, states C3. L3 agrees, saying that “what’s been done with AI so far is quite primitive, it still lacks narrative and that’s why it’s noticeable. We’ll see how it develops in the coming years”. C1 assures that “It can be distinguished because it’s untrained. But considering the tool learns and feeds back every day, if it can write a simple article now, imagine what it’ll be able to write in five, ten, or fifty years’ time”. C5 goes one step further, pointing to what will happen when readers can’t see the difference between automated and non-automated content, “When that time comes, we’ll have an issue of ethics, not of quality”.

The difference between automatically generated texts and human-written ones is particularly clear when it comes to slow journalism. This is what S1 believes, “So far it seems to me that readers of this kind of journalism, who are assumed to have a certain taste for literature, can notice the difference”. C14 (“In narrative reporting, of course they do”) is of the same view. C12 points out where the differences are, “At the moment it’s noticeable in linguistic proficiency, the choice of words, the rhythm, the lack of purpose, the impersonal language. There’ll come a time when it won’t make a difference, but it does now”. As for who readers look at when encountering a poorly written text, opinions differ. C7 thinks they blame an underdeveloped AI, “When they notice lack of quality, they think AI did it. We get a lot of feedback from our readers; after 600 words, it’s very noticeable”. But C15 says that “readers can see it’s badly written—there are very bad texts published every day—, but I don’t believe they think it’s written by an AI”. It remains to be seen whether, as C7 predicts, “in ten years’ time, if a piece is well done, people won’t care if it’s done by a human or a machine”.

4. Discussion and Conclusions

In this study, the increasing body of literature on the integration of AI in journalism is extended by eliciting and analyzing slow journalism professionals’ experiences, attitudes, and opinions about the use of AI tools in their work. To this end, semi-structured interviews of twenty-one slow journalism practitioners in Spain were carried out.

The results of the research suggest that just over half of the professionals surveyed use AI in their work as ‘slow’ journalists. The most common uses are transcribing interviews and, to a lesser extent, translating texts (RQ1), similar to what recent studies on the adoption of AI in journalism (Newman 2024, p. 27) show. The role of journalists has changed significantly, and they are increasingly expected to master the use of AI tools in addition to traditional journalistic skills. As a result, more hybrid jobs and collaboration between journalists, data scientists, information designers, and computer programmers emerge (Deuze and Beckett 2022; De-Lara-González et al. 2022). According to interviewees, the latter are less reluctant to adopt AI than journalists themselves.

When journalists do use AI, it is often limited to individual experiences with free tools that do not provide the desired quality, so a thorough review of the content they generate is needed. This suggests that while news organizations are training in the use of AI in newsrooms, they do not yet have high-quality (paid) tools available to their journalists. Hence AI’s ability to contribute to better quality—one of the factors that, according to the TAM, influence an individual’s intention to use new technology—is questioned by the interviewees (RQ2). This contradicts some who claim that AI not only increases journalists’ productivity but also improves the quality of news reporting (Shi and Sun 2024). Curiously, however, two of the three leading journalists interviewed ensure they have already drafted documents to guide the appropriate use of this new technology, presumably reflecting on what good journalism means (Min 2021, p. 89), or anticipating what is to come.

Despite their limited use of AI, all the journalists are aware that changes will arrive in the not-too-distant future, and they will be profound. Most believe that generative AI will be able to produce simple news articles in a few years’ time, although they doubt that it will go as far as writing narrative pieces and thus doing slow journalism (RQ3). In particular, the majority doubt that it can bring creativity to this type of text (RQ2) (in line with Graefe 2016; Gómez-Mompart 2021). Nor do they seem to agree with the idea (supported by Brennen 2018; Schapals and Porlezza 2020) that it offers benefits, such as allowing for deeper investigation and helping to add more complexity and meaning to texts—TAM’s factor of “perceived usefulness”. On the contrary, it is precisely the human qualities that characterize journalism—and to a greater extent slow journalism—that automation lacks, such as identifying the most important angle of a story, news judgment, curiosity, and skepticism (Thurman et al. 2017).

It is not surprising, therefore, that most respondents claim that their way of working has not changed since the introduction of AI in journalism. According to them, neither the way they research nor the way they present content has changed. They do not even seem to have benefited enough to say that they have more time to spend on work that cannot be automated—TAM’s factor of “ease of use”—as claimed by some authors (e.g., DalBen and Jurno 2021; Wu et al. 2019; Young and Hermida 2014). They tend to agree with the more skeptical, arguing that AI’s impact on productivity is likely to be more limited (Shi and Sun 2024), and that it varies depending on what is specifically automated (Simon 2024). In sum, practitioners believe that AI’s contribution to slow journalism will be limited to freeing journalists from routine and less creative tasks—TAM’s factor of “job relevance”—(RQ2).

In terms of how AI could affect the journalistic labor market, fear of job destruction is cited as the main reason for rejection of AI among professionals around the world (Beckett 2019). In the present case, respondents are also aware that AI could lead to significant job losses, although they do not currently seem to believe it could affect slow journalism practitioners that much (RQ3). According to their statements, it is not so much a question of seeing AI as a competitor or partner (a debate pointed out by Peña-Fernández et al. 2023), but rather as a useful support tool—in line with Wu et al.’s (2019) theory about how AI will change the news business—that as yet produces more quantity than the quality content (as suggested by Simon 2024) expected from slow journalists (Greenberg 2007, 2012).

In short, it can be concluded that according to the responses of the twenty-one slow journalism practitioners in Spain, AI has little to offer these minority-media (Boynton 2015; Craig 2020; Le Masurier 2015) professionals due to the notable limitations in improving long-form reporting (recently highlighted by authors such as Nishal and Diakopoulos 2024; Simon 2024). At most, it will allow them to streamline repetitive and non-creative tasks and thus better engage in in-depth journalism without it significantly threatening their jobs, at least in its current state of development. Still, it must be acknowledged that this study is not without limitations. While the sample of interviewees encompasses a diverse range of slow journalism formats in Spain, it does not represent all of slow journalism. Furthermore, future research could investigate whether there are variations between countries, given that the implementation of AI in journalism varies globally.