Framing Artificial Intelligence in Journalism: Dominant Frames and Systematic Omissions in Spanish Digital News Outlets from Pre-Boom to Implementation (2022–2024)

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This is a timely and revenant study drawing on a sample size and region that highlights key trends in AI coverage and media exposure.

Aspects to the literature elements are at times a little dense and hard to follow with topic and theme jumping quickly from one to the next.

Methodology and discussion as well as results are clearly presented and easy to follow with a little more succinctness needed in the conclusion, referring more distinctly to the research aims.

Overall strong submission and engaging subject material.

Author Response

Comment 1

This is a timely and relevant study drawing on a sample size and region that highlights key trends in AI coverage and media exposure. Aspects to the literature elements are at times a little dense and hard to follow with topic and theme jumping quickly from one to the next. Methodology and discussion as well as results are clearly presented and easy to follow with a little more succinctness needed in the conclusion, referring more distinctly to the research aims. Overall strong submission and engaging subject material.

 

Response 1

We are grateful for this assessment and have acted on both observations.  

 

The theoretical framework has been reorganised so that it now follows a single line of argument, with each concept introduced before it is used, and a passage on the stages of news production, which interrupted that line, has been moved to the methodology section, where it functions as part of the coding instrument (page 7, lines 319–327). 

The conclusions have been shortened and now follow the order of the four research objectives, each addressed in turn, with a summary table linking every objective to its main finding and to the section in which it is reported (pages 17–18, lines 611–677). 

Reviewer 2 Report

Comments and Suggestions for Authors

1. This paper furthers our understanding on how AI is framed by Spanish media outlets. The research design is well built, but some results are unclear and do not overlap with current scholarship on topic. Hence, some changes are requested prior to publication. 

2. First, in the introduction the manuscript addresses the rise of AI in journalism. However, in my opinion it fails to provide enough information on framing and media representation of AI. For example, I miss a key reference on the ambivalent news coverage of AI: 

• Cools, H., Van Gorp, B., & Opgenhaffen, M. (2024). Where exactly between utopia and dystopia? A framing analysis of AI and automation in US newspapers. Journalism, 25(1), 3-21. https://doi.org/10.1177/14648849221122647 

Moreover, it is true that existing studies focus mainly on Anglophone media contexts, but some articles have examined the news framing of AI in Europe, including Spain. These results might be interpreted in a comparative manner. See, for instance: 

• Rivas-de-Roca, R., Pérez-Curiel, C., & Casero-Ripollés, A. (2025). Reporting AI in Political Disinformation: Journalistic Discourses in European Quality Newspapers. Journalism Practice, 1-18. https://doi.org/10.1080/17512786.2025.2505961 

3. Regarding methods, data selection and analysis seem appropriate. The coding procedure is also clearly explained. Nevertheless, I think RQ1 is too ambitious considering the limited number of items. This question should be clarified or removed.

4. Some results are noteworthy, particularly on how predominant frames focus on the processing phase of news production, credibility, or ethics. Even though the analysis of the association between journalistic variables and frames (RQ4) could be interesting, additional details are required to assess the findings. The media origin, ideology, authorship (at least a number of authors), journalistic genres, and main sources of each news outlet should be identified. 

On this matter, a description of each frame needs also to be provided in order to enhance transparency. For instance, which is “promoting social empathy”?

5. The discussion offered is insightful but should be nuanced as only a scant of longitudinal analysis was conducted. Besides that, I would recommend answering each research question directly in the conclusions section. This promotes the traceability of the study. 

6. In short, this article could be a relevant contribution that addresses a timely topic in the field of journalism in Spain, but the current version has some flaws. I suggest reworking the theoretical background introduction and clarifying the issues raised.  

Author Response

1. This paper furthers our understanding on how AI is framed by Spanish media outlets. The research design is well built, but some results are unclear and do not overlap with current scholarship on topic. Hence, some changes are requested prior to publication.

We are grateful for this assessment and for the precise guidance that follows. Every point raised has been addressed, and the corresponding changes are detailed below.

2. First, in the introduction the manuscript addresses the rise of AI in journalism. However, in my opinion it fails to provide enough information on framing and media representation of AI. For example, I miss a key reference on the ambivalent news coverage of AI: 

• Cools, H., Van Gorp, B., & Opgenhaffen, M. (2024). Where exactly between utopia and dystopia? A framing analysis of AI and automation in US newspapers. Journalism, 25(1), 3-21. https://doi.org/10.1177/14648849221122647 

Moreover, it is true that existing studies focus mainly on Anglophone media contexts, but some articles have examined the news framing of AI in Europe, including Spain. These results might be interpreted in a comparative manner. See, for instance: 

• Rivas-de-Roca, R., Pérez-Curiel, C., & Casero-Ripollés, A. (2025). Reporting AI in Political Disinformation: Journalistic Discourses in European Quality Newspapers. Journalism Practice, 1-18. https://doi.org/10.1080/17512786.2025.2505961 

We agree, and both suggestions have been incorporated. The theoretical framework now includes a passage on the ambivalence of AI news coverage between utopian and dystopian registers, drawing on Cools, Van Gorp and Opgenhaffen (2024), which also allows us to anticipate the simultaneous enthusiasm and alarm that our own findings document (page 5, lines 224–227). Rivas-de-Roca, Pérez-Curiel and Casero-Ripollés (2025) has been added at the point where the predominance of Anglophone contexts is identified, since it qualifies that claim for the European case, and it now supports a comparative reading of our results (page 6, lines 234–237). Both works have been added to the reference list.

3. Regarding methods, data selection and analysis seem appropriate. The coding procedure is also clearly explained. Nevertheless, I think RQ1 is too ambitious considering the limited number of items. This question should be clarified or removed.

We have clarified rather than removed it, for the following reason. The 242 items are not a sample but the population of analysable units retrieved under the eligibility criteria specified in the methodology, and the corpus was coded in its entirety, so the first objective does not involve inference from a sample to a wider universe. That said, the reviewer is right that its earlier formulation promised more than the corpus can deliver. The objective has been reworded so that it traces the emergence and the distribution of media attention across the three phases of the period, and the accompanying claims are now stated at that level (page 6, lines 275–276).

4. Some results are noteworthy, particularly on how predominant frames focus on the processing phase of news production, credibility, or ethics. Even though the analysis of the association between journalistic variables and frames (RQ4) could be interesting, additional details are required to assess the findings. The media origin, ideology, authorship (at least a number of authors), journalistic genres, and main sources of each news outlet should be identified. 

Although these factors (media origin, ideology, authorship, and journalistic genre) have traditionally influenced news production, in this case, as shown in Figures 3 and 4, only a consistent and strong association is observed between the main source and the different AI frames. The results also identify some specific associations between authorship—particularly agency-produced content, in-house journalists, non-journalist contributors, and generic bylines—and certain prominent or systematically omitted frames. Accordingly, the Conclusions section has been revised to highlight the limited influence of these other professional variables in relation to this objective.

On this matter, a description of each frame needs also to be provided in order to enhance transparency. For instance, which is “promoting social empathy”?

The variables representing our issue-specific frames are not based on our own definitions, but rather on concepts developed and defined by previous scholars, as we explain throughout the theoretical framework. For example, “promoting social empathy” corresponds to one of the functions attributed to and defined by Schudson (2008). Although we acknowledge that each variable could be further defined, we considered it more appropriate to rely on the academic references accompanying our conceptualizations throughout the article, thereby allowing readers who wish to do so to explore each concept in greater depth while also enhancing the readability of the article and avoiding unnecessary length. The database, which is openly available on Zenodo and accessible through the article’s Data Availability Statement, also includes the corresponding codebook.

5. The discussion offered is insightful but should be nuanced as only a scant of longitudinal analysis was conducted. Besides that, I would recommend answering each research question directly in the conclusions section. This promotes the traceability of the study. 

Both points have been taken up. On the first, the design is explicitly presented as three waves of analysis corresponding to distinct phases of an issue-attention cycle rather than as a continuous time series, and the claims made about the evolution of attention are stated at that level. The conclusions have been reorganised to follow the order of the four objectives, each addressed in turn, and a summary table now links every objective to its main finding and to the section reporting it. The limitations have also been expanded, and now note that two variables yielded tentative reliability coefficients and that the associations reported are not causal.

6. In short, this article could be a relevant contribution that addresses a timely topic in the field of journalism in Spain, but the current version has some flaws. I suggest reworking the theoretical background introduction and clarifying the issues raised.  

The theoretical background has been restructured around a single line of argument and condensed, following this comment and a converging observation from another reviewer. We hope the revised version answers the concerns raised.

 

Reviewer 3 Report

Comments and Suggestions for Authors

Thank you for the opportunity to review this manuscript. The study is well conceived, clearly structured and methodologically sound. It addresses a highly relevant topic at the intersection of generative artificial intelligence, journalism and media representation, while providing original empirical evidence from the relatively unexplored Spanish media landscape.

I appreciate the rational scope of the research design. The authors have chosen a clearly defined problem, appropriate time frame and a corpus that is large enough to provide valid and meaningful interpretive findings without unnecessarily expanding the analytical scope. The longitudinal quantitative content analysis of ten leading Spanish digital news media provides a solid basis for identifying both dominant frameworks and systematic omissions in the representation of AI in journalism.

The study makes an interesting contribution by examining the period between 2022 and 2024 not as three consecutive years, but as distinct phases in early adoption of generative AI in journalism. In this regard, my only substantive suggestion concerns the title of the manuscript. The current title ends with the specification “(2022–2024)”, which may give the impression that these years represent an arbitrarily chosen period. However, as the authors themselves convincingly explain, this period captures specific stages in the emergence and early implementation of AI in journalism practice. I would therefore recommend considering whether the title could reflect this logic more explicitly. This would, in my opinion, better communicate the conceptual rationale for the longitudinal design and further strengthen the connection between the title and the actual contribution of the study.

Besides, I have identified two minor formal issues. The citation of Dodds et al. seems inconsistent: in the main text it is listed as Dodds et al. (2025), while the bibliography lists the year of publication as 2026. This should be checked and harmonized. There also appears to be a typographical error in the citation of Chadwick (2013), whose name is written as “Chadwic” in the Discussion section.

However, these are very minor comments that do not affect my overall assessment of the manuscript. I congratulate the author(s) on a well-written study. I consider the manuscript suitable for publication and look forward to its publication.

Author Response

Dear Reviewer,

We would like to thank you for your careful evaluation of our manuscript. Below, we provide a point-by-point response to your comments, with our responses highlighted in red.

Thank you for the opportunity to review this manuscript. The study is well conceived, clearly structured and methodologically sound. It addresses a highly relevant topic at the intersection of generative artificial intelligence, journalism and media representation, while providing original empirical evidence from the relatively unexplored Spanish media landscape.

I appreciate the rational scope of the research design. The authors have chosen a clearly defined problem, appropriate time frame and a corpus that is large enough to provide valid and meaningful interpretive findings without unnecessarily expanding the analytical scope. The longitudinal quantitative content analysis of ten leading Spanish digital news media provides a solid basis for identifying both dominant frameworks and systematic omissions in the representation of AI in journalism.

The study makes an interesting contribution by examining the period between 2022 and 2024 not as three consecutive years, but as distinct phases in early adoption of generative AI in journalism. In this regard, my only substantive suggestion concerns the title of the manuscript. The current title ends with the specification “(2022–2024)”, which may give the impression that these years represent an arbitrarily chosen period. However, as the authors themselves convincingly explain, this period captures specific stages in the emergence and early implementation of AI in journalism practice. I would therefore recommend considering whether the title could reflect this logic more explicitly. This would, in my opinion, better communicate the conceptual rationale for the longitudinal design and further strengthen the connection between the title and the actual contribution of the study.

We are grateful for this reading of the design, and we have followed the recommendation. The title now makes the logic of the phases explicit while retaining the period, since the years also serve to situate the corpus: “Framing Artificial Intelligence in Journalism: Dominant Frames and Systematic Omissions in Spanish Digital News Outlets from Pre-Boom to Implementation (2022–2024)”.

Besides, I have identified two minor formal issues. The citation of Dodds et al. seems inconsistent: in the main text it is listed as Dodds et al. (2025), while the bibliography lists the year of publication as 2026. This should be checked and harmonized. There also appears to be a typographical error in the citation of Chadwick (2013), whose name is written as “Chadwic” in the Discussion section.

Both have been corrected. Dodds et al. is now cited as 2026 throughout the manuscript and in the reference list, and the spelling of Chadwick has been fixed. We have also reviewed the punctuation of in-text citations, since a small number of them carried a comma between author and year that the journal’s style does not use.

However, these are very minor comments that do not affect my overall assessment of the manuscript. I congratulate the author(s) on a well-written study. I consider the manuscript suitable for publication and look forward to its publication.

We thank the reviewer for this assessment and for the attention given to the manuscript.

Reviewer 4 Report

Comments and Suggestions for Authors

SOCIAL SCIENCES REVIEW

PROPOSAL TITLE:

 

 Framing Artificial Intelligence in Journalism: Dominant Frames and Systematic Omissions in Spanish Digital News Outlets (2022–2024)

This is a topic proposal of interest for advancing research on the impact of AI. Specifically, it focuses on the professional field of journalism, examining the frames and perspectives adopted by the selected Spanish digital media outlets. The research comprises three phases: an initial phase in 2022, covering the period prior to the public launch of ChatGPT, followed by two subsequent phases. The first is an expansion phase (2023), focused on the public dissemination and adoption of generative AI tools, and the second is an implementation phase (2024), marked by institutional and regulatory changes concerning the use of AI.

Overall, the article presents a structure consistent with the requirements of scientific research, providing results and conclusions on how AI may affect journalism and opening avenues for future studies focusing on the position of institutions (European regulations and legislation), the media themselves, and audiences regarding the benefits and risks associated with the use of AI.

However, the main criticism of the study is that it encompasses too many research variables across too many sources and media outlets. This results in a certain degree of scientific superficiality.

Indeed, considering that the final sample consists of 242 texts published in 10 Spanish digital media outlets, and taking into account the authors' justification for the sampling period:

“Coverage of AI began to increase in early 2023, with 39 articles published between January and June. The summer of 2023 marked a decline in the relevance of the topic (13 articles), before October 2023 became the starting point for a substantial increase that continued through June 2024 (132 entries). Another summer period, from July to September 2024, brought about a renewed decline (24 articles), which was not followed by another increase in media relevance, but rather by a period of stagnation during the final quarter of the analysis, from October to December 2024 (25 articles).”

The suggestion would therefore be to reduce the number of variables and even shorten the sampling period, with the aim of achieving greater analytical depth. In this way, the authors could have focused on part of 2023 and 2024, when the expansion and dissemination of generative AI tools in professional and public contexts took place, accompanied by an increase in media coverage that continued until June 2024.

With the aim of helping the article improve some of the sections comprising the proposal, the following suggestions and recommendations are made to the authors:

The abstract should be revised. The hypothesis and the general objective of the study are not clearly identified. The results obtained are, however, clearly outlined.

Regarding the introduction, greater reference should be made to the context of the Fourth Industrial Revolution, particularly with regard to the digital media environment in Spain.

In another respect, although the objectives are subsequently broken down in the methodology, the introduction would benefit from a clearer presentation of the hypothesis and the general objective, which are not yet sufficiently well defined.

The theoretical framework draws on important and relevant scientific literature. However, given that this is a study conducted in Spain and within the European context, some gaps can be identified in terms of recent references published in high-impact journals. Among others:

• Rubén Rivas-de-Roca, Concha Pérez-Curiel & Andreu Casero-Ripollés (14 May 2025): Reporting AI in Political Disinformation: Journalistic Discourses in European Quality Newspapers, Journalism Practice, DOI: 10.1080/17512786.2025.2505961.
• Gutiérrez-Caneda, Beatriz; Vázquez-Herrero, Jorge; López-García, Xosé (2023). “AI application in journalism: ChatGPT and the uses and risks of an emergent technology”. Profesional de la información, Vol. 32, No. 5, e320514. https://doi.org/10.3145/epi.2023.sep.14

In any case, given the ambitious scope of the study, the theoretical framework also reflects the superficiality that inevitably results from addressing too many topics. It could have been more narrowly focused on the thematic agenda and framing of AI in journalism and in the selected digital media outlets. Approaches such as the ethical and social implications of AI, as well as its regulatory framework, which are particularly addressed in the 2024 period of the study, could instead form part of the development of future research.

The general and specific objectives are presented at the end of the theoretical framework and are well formulated. However, as they address the evolution of AI in journalism, its expansion, and the regulatory and legislative effects, they are still considered ambitious and difficult to achieve.

With regard to the methodology, several points deserve consideration:

First, the way in which the ten Spanish digital media outlets were selected should be explained. The authors refer to leadership, but references to rankings or reports are lacking.

It is noteworthy that no research questions are formulated in the methodology, particularly given that the hypothesis has already been identified as insufficiently defined. It is therefore suggested that research questions be incorporated, as these are important for guiding the analysis of the results, discussion, and conclusions.

The methodological section requires the study variables to be clearly stated and their function explained. This is the appropriate section in which these indicators should be delimited and defined, rather than in the results section.

The sample, consisting of 242 newspaper articles published in 10 Spanish digital media outlets over a three-year period, raises some questions. Among them is whether 242 texts can be considered representative, taking into account the number of newspapers analyzed and the sampling frequency established.

Furthermore, there appears to be an intention to increase the number of study variables in an attempt to compensate for the limited number of publications.

It is always recommended that scientific rigor should not be compromised by including time periods that do not contribute to understanding the impact of AI on the journalism profession, or by incorporating variables that, given this relatively small sample of texts, do not allow representative and conclusive results to be obtained.

Regarding the Results section, after reviewing all the figures, charts, and tables, it is noteworthy that no data are provided broken down by each of the ten selected newspapers. All the charts present overall percentages, making it difficult to assess how each individual newspaper behaves. This issue is also not explained in a disaggregated manner in the text. It is therefore suggested that the authors include some stacked bar charts that would make it possible to examine each individual case and compare it across the different study variables.

Furthermore, given that sources—particularly the main source—play a central role in most of the texts and correspond primarily to actors within the sector, no comparative chart is provided showing the presence of the different types of sources, such as institutional sources and experts, among others.

The results do not adequately address the multiplicity of variables examined. Therefore, it was important to select a more focused set of indicators and variables and analyze them in greater depth.

With regard to the Discussion, greater emphasis should be placed on interpreting the data rather than summarizing ideas that have already been explained earlier.

As indicated above, none of the sections of the article provides specific, disaggregated references to each of the media outlets analyzed. This generalization and lack of specificity in the analysis may affect the methodological rigor of the quantitative approach employed. Similarly, the Discussion does not address how each of these media outlets operates, nor does it examine their similarities, differences, or particular approaches to the production and treatment of information about AI.

Regarding the Conclusions, it would be essential for them to respond directly to the objectives, hypotheses, or research questions and to be more clearly supported by previous literature.

With regard to the study's limitations, the authors state in the Conclusions that the results should be interpreted with the understanding that the research examines a single media system, namely the Spanish one, and ten digital media outlets, which calls for caution when extrapolating the observed trends.

In this regard, some of the conclusions, particularly those concerning the regulatory role of institutions such as the European Union, are not supported by verified results. Everything related to this indicator should therefore be reviewed, and some aspects that have not been demonstrated should not be interpreted subjectively. In fact, I would recommend reviewing all conclusions concerning the position of institutions, particularly those relating to texts published in 2024.

The authors do not consider the study sample (242 texts analyzed) to be a limitation. However, given that the study refers to ten digital media outlets over a three-year period (from 2022 to 2024), the sample does not appear to be representative.

The Conclusions also refer to future research expanding the study to the international level. This prospective approach is valuable, although it may be more appropriate to pursue such an expansion after further exploring other aspects related to journalism and AI within the national context.

Regarding the bibliographic references, they are up to date, without diminishing the value of previous publications that laid the scientific foundations for research on AI.

However, some relevant references in studies on AI, journalism, and the media are missing. For example:

• Rubén Rivas-de-Roca, Concha Pérez-Curiel & Andreu Casero-Ripollés (14 May 2025): Reporting AI in Political Disinformation: Journalistic Discourses in European Quality Newspapers, Journalism Practice, DOI: 10.1080/17512786.2025.2505961.
• Gutiérrez-Caneda, Beatriz; Vázquez-Herrero, Jorge; López-García, Xosé (2023). “AI application in journalism: ChatGPT and the uses and risks of an emergent technology”. Profesional de la información, Vol. 32, No. 5, e320514. https://doi.org/10.3145/epi.2023.sep.14

At a semantic level, I would suggest that the authors review the concepts related to AI to ensure that they are properly defined and contextualized, so that they can be clearly understood, particularly given the innovative field generated by AI.

Finally, from a formal perspective, I would recommend using simpler prose with fewer subordinate clauses, as this would also facilitate a clearer understanding of the ideas, particularly in the Introduction and Theoretical Framework sections.

 

 

Author Response

Dear Reviewer,

We would like to thank you for your careful evaluation of our manuscript. Below, we provide a point-by-point response to your comments, with our responses highlighted in red.

This is a topic proposal of interest for advancing research on the impact of AI. Specifically, it focuses on the professional field of journalism, examining the frames and perspectives adopted by the selected Spanish digital media outlets. The research comprises three phases: an initial phase in 2022, covering the period prior to the public launch of ChatGPT, followed by two subsequent phases. The first is an expansion phase (2023), focused on the public dissemination and adoption of generative AI tools, and the second is an implementation phase (2024), marked by institutional and regulatory changes concerning the use of AI.

Overall, the article presents a structure consistent with the requirements of scientific research, providing results and conclusions on how AI may affect journalism and opening avenues for future studies focusing on the position of institutions (European regulations and legislation), the media themselves, and audiences regarding the benefits and risks associated with the use of AI.

However, the main criticism of the study is that it encompasses too many research variables across too many sources and media outlets. This results in a certain degree of scientific superficiality.

Indeed, considering that the final sample consists of 242 texts published in 10 Spanish digital media outlets, and taking into account the authors' justification for the sampling period:

“Coverage of AI began to increase in early 2023, with 39 articles published between January and June. The summer of 2023 marked a decline in the relevance of the topic (13 articles), before October 2023 became the starting point for a substantial increase that continued through June 2024 (132 entries). Another summer period, from July to September 2024, brought about a renewed decline (24 articles), which was not followed by another increase in media relevance, but rather by a period of stagnation during the final quarter of the analysis, from October to December 2024 (25 articles).”

The suggestion would therefore be to reduce the number of variables and even shorten the sampling period, with the aim of achieving greater analytical depth. In this way, the authors could have focused on part of 2023 and 2024, when the expansion and dissemination of generative AI tools in professional and public contexts took place, accompanied by an increase in media coverage that continued until June 2024.

We are grateful for the care this review reflects, and a substantial number of its recommendations have been adopted, as set out point by point below. On the central criticism, we would ask the reviewer to consider three clarifications. First, the study is exploratory by design and deliberately panoramic in scope: its purpose is to establish which aspects of AI enter the media agenda and which do not, and that question requires a wide inventory of issue-specific frames rather than a narrow set examined in depth. Breadth here is a design decision with a cost we accept, and it is the condition for the subsequent studies of this project, in which individual dimensions such as disinformation or the tone of coverage will be analysed in depth. Second, the 242 items are not a sample but the population of analysable units retrieved under the eligibility criteria specified in the methodology, and the corpus was coded in its entirety; shortening the period would therefore not deepen the analysis but reduce the number of units available. Third, the three years are not treated as consecutive annual slices but as distinct phases of an issue-attention cycle, and removing 2022 would remove the pre-boom baseline against which the expansion and implementation phases acquire their meaning. The methodology now makes both points explicit (page 6-7, lines 275–287; page 7, lines 301–304; and page 8, lines 330-338).

With the aim of helping the article improve some of the sections comprising the proposal, the following suggestions and recommendations are made to the authors:

The abstract should be revised. The hypothesis and the general objective of the study are not clearly identified. The results obtained are, however, clearly outlined.

The abstract has been rewritten to state the general objective at the outset. On the hypothesis we should be explicit: the study follows a descriptive and exploratory design and is guided by research objectives rather than by hypothesis testing. No hypothesis is formulated because none is being contrasted, and the abstract now says so rather than leaving the matter implicit.

Regarding the introduction, greater reference should be made to the context of the Fourth Industrial Revolution, particularly with regard to the digital media environment in Spain.

The Fourth Industrial Revolution is invoked in the opening paragraph of the introduction, following Schwab (2016) and Quian and Sixto-García (2024) (page 2, lines 39–41). Following this comment, the connection between that framework and the Spanish digital media environment has been made explicit at that point, rather than left for the reader to infer. The environment itself is characterised further on in the same section, where the contraction of digital news traffic in Spain and the disintermediation of journalistic content are documented with industry audience data (pages 3, lines 92–95).

In another respect, although the objectives are subsequently broken down in the methodology, the introduction would benefit from a clearer presentation of the hypothesis and the general objective, which are not yet sufficiently well defined.

The general objective is now stated at the close of the introduction (page 3, lines 112–115) and again at the end of the theoretical framework (page 6, lines 248–252), immediately before the methodology section, where the four specific objectives are set out (page 6-7, lines 275–287). As indicated above, the study formulates no hypothesis, and this is now made explicit at the same points.

The theoretical framework draws on important and relevant scientific literature. However, given that this is a study conducted in Spain and within the European context, some gaps can be identified in terms of recent references published in high-impact journals. Among others:

• Rubén Rivas-de-Roca, Concha Pérez-Curiel & Andreu Casero-Ripollés (14 May 2025): Reporting AI in Political Disinformation: Journalistic Discourses in European Quality Newspapers, Journalism Practice, DOI: 10.1080/17512786.2025.2505961.
• Gutiérrez-Caneda, Beatriz; Vázquez-Herrero, Jorge; López-García, Xosé (2023). “AI application in journalism: ChatGPT and the uses and risks of an emergent technology”. Profesional de la información, Vol. 32, No. 5, e320514. https://doi.org/10.3145/epi.2023.sep.14

In any case, given the ambitious scope of the study, the theoretical framework also reflects the superficiality that inevitably results from addressing too many topics. It could have been more narrowly focused on the thematic agenda and framing of AI in journalism and in the selected digital media outlets. Approaches such as the ethical and social implications of AI, as well as its regulatory framework, which are particularly addressed in the 2024 period of the study, could instead form part of the development of future research.

Both references have been incorporated and the framework has been restructured. Rivas-de-Roca, Pérez-Curiel and Casero-Ripollés (2025) now qualifies the claim about the predominance of Anglophone contexts and supports a European comparison, and Gutiérrez-Caneda, Vázquez-Herrero and López-García (2023) has been added where the uses and risks of generative tools in journalistic practice are discussed (page 4, lines 139–141, and page 5, lines 234–237). On the breadth of the framework, we would clarify its function: it does not survey a series of separate topics but presents the inventory of issue-specific frames identified in previous research, which is the inventory operationalised in the coding instrument. The ethical and regulatory dimensions are among the variables coded in this study, and the framework therefore has to introduce them; what has changed is that they are now presented as entries in that inventory rather than developed as thematic discussions in their own right (pages 3–5, lines 116–247). The framework as a whole now follows a single line of argument, from the definition of AI to agenda-setting, framing and issue-specific frames, and finally to selectivity and omission, with the digressions removed and the three-phase model of news work relocated to the methodology, where it belongs (pages 3–5, lines 116–247, and page 8, lines 330–338).

The general and specific objectives are presented at the end of the theoretical framework and are well formulated. However, as they address the evolution of AI in journalism, its expansion, and the regulatory and legislative effects, they are still considered ambitious and difficult to achieve.

The objectives concern the treatment of AI during the first phases of media attention, not the long-term evolution of the technology or of its regulation. Following this comment and a related observation from Reviewer 2, the first objective has been reworded so that it claims no more than the corpus supports, and the scope of the remaining objectives has been stated in the same terms (page 6-7, lines 275–287). The interpretation of the phases rests on Downs’s (1972) issue-attention cycle, which is precisely a model of how attention to an issue rises and falls over a limited period.

With regard to the methodology, several points deserve consideration:

First, the way in which the ten Spanish digital media outlets were selected should be explained. The authors refer to leadership, but references to rankings or reports are lacking.

The reviewer is right that the criterion was asserted but not evidenced. The methodology now specifies that the ten outlets were selected on the basis of GfK DAM audience data for 2022, 2023 and 2024, taking each outlet’s share of unique users as the criterion and preserving the balance between legacy titles and digital-native platforms (page 7, lines 301–304).

It is noteworthy that no research questions are formulated in the methodology, particularly given that the hypothesis has already been identified as insufficiently defined. It is therefore suggested that research questions be incorporated, as these are important for guiding the analysis of the results, discussion, and conclusions.

The four research objectives were placed at the end of the theoretical framework, which explains why they were not visible where the reviewer looked for them. They are now restated at the opening of the methodology section, numbered RO1 to RO4, so that each subsequent section can be read against them, and the conclusions address them one by one, with a summary table mapping each objective to its finding (page 6-7, lines 275–287, and pages 17–18, lines 631–701). We have retained the form of objectives rather than questions for the reason given in Response 2: the design is descriptive and exploratory and contrasts no hypothesis.

The methodological section requires the study variables to be clearly stated and their function explained. This is the appropriate section in which these indicators should be delimited and defined, rather than in the results section.

Each variable derived from an issue-specific frame is defined throughout the theoretical framework, either explicitly or through supporting citations that provide access to the original definition proposed by the respective author. The codebook can be accessed through the article’s Data Availability Statement. In the Results section, the variables themselves are not defined; rather, we explain how the dimensions were constructed based on the greater presence or absence of each frame.

The sample, consisting of 242 newspaper articles published in 10 Spanish digital media outlets over a three-year period, raises some questions. Among them is whether 242 texts can be considered representative, taking into account the number of newspapers analyzed and the sampling frequency established.

The 242 items are not a sample but the population of units of analysis retrieved under the eligibility criteria specified in the methodology, and the corpus was coded in its entirety, so no inference from sample to population is involved and statistical representativeness is not at stake. Retrieval through MyNews returned over 2,000 items, which were filtered by a relevance threshold and then screened manually to retain only those in which AI in journalism constitutes the central focus rather than a passing mention; the 242 items that met these criteria are the whole of what the search yielded, not a selection from it. Probability sampling had been envisaged at the design stage and ceased to be applicable once screening was complete. The methodology now states the selection criteria, and the terminology has been corrected throughout, since the earlier wording referred to a sample and thereby invited precisely this objection.

Furthermore, there appears to be an intention to increase the number of study variables in an attempt to compensate for the limited number of publications.

It is always recommended that scientific rigor should not be compromised by including time periods that do not contribute to understanding the impact of AI on the journalism profession, or by incorporating variables that, given this relatively small sample of texts, do not allow representative and conclusive results to be obtained.

We would respectfully clarify the sequence. The inventory of issue-specific frames derives from frames identified in previous research on the coverage of AI, and was fixed in the coding instrument through two pilot tests before the corpus was coded; it was not expanded in response to the number of items retrieved. The frames are moreover not treated as independent variables competing for a small number of cases: they are grouped into dimensions and subdimensions, and it is at that level that the analysis operates. We refer also to Response 1 on the exploratory purpose of the design.

Regarding the Results section, after reviewing all the figures, charts, and tables, it is noteworthy that no data are provided broken down by each of the ten selected newspapers. All the charts present overall percentages, making it difficult to assess how each individual newspaper behaves. This issue is also not explained in a disaggregated manner in the text. It is therefore suggested that the authors include some stacked bar charts that would make it possible to examine each individual case and compare it across the different study variables.

The aim of the study is not to provide a detailed account of the individual responses of each media outlet, but rather to offer a broader view of the Spanish media system. Therefore, we did not consider it appropriate to disaggregate the results by individual outlet. Nevertheless, a table listing the selected media outlets, their origin, ideological orientation, and number of news articles published has been added to the Materials and Methods section. In addition, the database is openly available, allowing readers to consult the results for each individual media outlet.

Furthermore, given that sources—particularly the main source—play a central role in most of the texts and correspond primarily to actors within the sector, no comparative chart is provided showing the presence of the different types of sources, such as institutional sources and experts, among others.

While it is true that no table has been created to show the frequency of each category of main sources, the frequency of each category is reported in the text to document their use (page 13, lines 478-480). We do not delve further into other characteristics of the sources, in line with the aim of providing a broad initial overview.

The results do not adequately address the multiplicity of variables examined. Therefore, it was important to select a more focused set of indicators and variables and analyze them in greater depth.

Certain aspects are not examined in greater depth, as priority is given to examining all items in accordance with the aim of providing a broad initial overview.

With regard to the Discussion, greater emphasis should be placed on interpreting the data rather than summarizing ideas that have already been explained earlier.

We understand that, following an extensive and complex Results section, particularly given the configuration of the dimensions within which the variables are organized, the Discussion section needs to strengthen its connection with the theoretical concepts in order to substantiate the extent to which the study objectives are addressed in the Conclusions. We acknowledge that some of these points may overlap with those presented in the theoretical framework; however, we consider it appropriate to demonstrate the extent to which these theoretical assumptions are supported by our findings.

As indicated above, none of the sections of the article provides specific, disaggregated references to each of the media outlets analyzed. This generalization and lack of specificity in the analysis may affect the methodological rigor of the quantitative approach employed.

To present the selected media outlets, we have added Table 1, “Media Population and Characteristics,” which provides information on the outlets included in the sample, their media origin, ideological orientation, and the number of news articles published on the topic during the period under analysis.

Similarly, the Discussion does not address how each of these media outlets operates, nor does it examine their similarities, differences, or particular approaches to the production and treatment of information about AI.

Differences in publication volume across media outlets are now addressed in table 1, where an explanation of these differences has been incorporated. The study’s aim of providing a broad overview, rather than conducting individual case studies, together with the analytical approach adopted, means that we do not consider outlet-level results to be particularly relevant to the objectives of the study. Nevertheless, results for each individual media outlet can be extracted from the database, which is accessible through the article’s Data Availability Statement.

Regarding the Conclusions, it would be essential for them to respond directly to the objectives, hypotheses, or research questions and to be more clearly supported by previous literature.

The conclusions have been reordered to follow the sequence of the four objectives, each addressed in turn and connected to the previous literature on which the interpretation rests, and a summary table now links every objective to its main finding and to the section reporting it (pages 17–18, lines 631–701).

With regard to the study's limitations, the authors state in the Conclusions that the results should be interpreted with the understanding that the research examines a single media system, namely the Spanish one, and ten digital media outlets, which calls for caution when extrapolating the observed trends.

In this regard, some of the conclusions, particularly those concerning the regulatory role of institutions such as the European Union, are not supported by verified results. Everything related to this indicator should therefore be reviewed, and some aspects that have not been demonstrated should not be interpreted subjectively. In fact, I would recommend reviewing all conclusions concerning the position of institutions, particularly those relating to texts published in 2024.

We have revised this passage substantially. Every claim concerning the position of European institutions, the terms of the agreements between media organisations and technology companies, and the compensation of media workers has been removed, since none of it follows from the coded data. What remains is a single interpretive paragraph, explicitly marked as a reading that the present data can raise but not settle, anchored in two findings the analysis does establish: the prominence of the legal-responsibility frame and the dominance of media-sector sources. The question it raises is taken up in the lines for further research.

The authors do not consider the study sample (242 texts analyzed) to be a limitation. However, given that the study refers to ten digital media outlets over a three-year period (from 2022 to 2024), the sample does not appear to be representative.

As set out in a previous response, the corpus is the population of analysable units under the stated eligibility criteria rather than a sample, so its size is a property of the phenomenon during this period and not a sampling decision. The limitations section has nonetheless been expanded, and now records that the analysis covers a single media system, that two variables yielded tentative reliability coefficients, and that the associations reported between journalistic variables and frames are not causal.

The Conclusions also refer to future research expanding the study to the international level. This prospective approach is valuable, although it may be more appropriate to pursue such an expansion after further exploring other aspects related to journalism and AI within the national context.

We agree and have reordered the lines of further research accordingly: the deepening of individual dimensions within the Spanish context now precedes the international comparison, which is presented as a subsequent step (page 18, lines 717–719).

Regarding the bibliographic references, they are up to date, without diminishing the value of previous publications that laid the scientific foundations for research on AI.

However, some relevant references in studies on AI, journalism, and the media are missing. For example:

• Rubén Rivas-de-Roca, Concha Pérez-Curiel & Andreu Casero-Ripollés (14 May 2025): Reporting AI in Political Disinformation: Journalistic Discourses in European Quality Newspapers, Journalism Practice, DOI: 10.1080/17512786.2025.2505961.
• Gutiérrez-Caneda, Beatriz; Vázquez-Herrero, Jorge; López-García, Xosé (2023). “AI application in journalism: ChatGPT and the uses and risks of an emergent technology”. Profesional de la información, Vol. 32, No. 5, e320514. https://doi.org/10.3145/epi.2023.sep.14

Both have been incorporated, as detailed in a previous response, and added to the reference list.

At a semantic level, I would suggest that the authors review the concepts related to AI to ensure that they are properly defined and contextualized, so that they can be clearly understood, particularly given the innovative field generated by AI.

The concepts related to AI are defined in the theoretical framework, where each one is either explained explicitly or accompanied by the reference to the author who originally formulated it, so that its meaning can be traced to its source (page 2, lines 62–68; page 3, lines 101-108; and pages 3–6, lines 116–252).

Finally, from a formal perspective, I would recommend using simpler prose with fewer subordinate clauses, as this would also facilitate a clearer understanding of the ideas, particularly in the Introduction and Theoretical Framework sections.

We accept this. The introduction and the theoretical framework have been rewritten with shorter sentences and reduced subordination, and the manuscript as a whole has undergone a further language revision (pages 1–3, lines 32–115, and pages 3–6, lines 116–252).

 

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The authors have made a substantial effort to address and respond to all of the reviewer’s concerns. All the points raised have been carefully addressed, including a stronger connection to the previous literature and greater transparency regarding the media sample. The research questions are also clearly answered. Therefore, given the academic value and contribution of the article to the field, I believe it is now ready for publication. The only remaining suggestion is to check the alphabetical order of the references.

Reviewer 4 Report

Comments and Suggestions for Authors I wanted to thank the authors for their response to each of the requested amendments. I have checked that all possible changes have been made to the text, which is highlighted. Other details or points that were causing confusion have been reorganised and placed in the correct location. Table 1, detailing the measures selected for the study and the number of publications in each case, has also been included. Furthermore, they have taken care to include key bibliographical references and to revise the wording, providing a less subordinate and clearer narrative for the article’s readers. Taking all these factors into account, as far as I am concerned, the article meets the requirements for publication, should the editorial committee so decide. Whilst maintaining anonymity, please convey my thanks to the authors for all the work they have done, which represents a significant improvement on the initial version. Many thanks Best regards