Artificial Intelligence in Journalism and Media Practice: A Systematic Review of Applications, Limitations, and Research Approaches
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsI consider the article—and, consequently, the research—to be relevant to the field of journalism studies. As the author(s) highlight, it would be valuable to replicate the methodology and research focus in leading scientific journals within countries where the subject matter is particularly significant; I would cite Brazil in this regard, as authors and researchers there often face challenges in accessing top-tier international scientific journals. Furthermore, I believe it is important to disseminate the proposed article among journalism research associations worldwide.
Author Response
Reviewer Comment:
I consider the article—and, consequently, the research—to be relevant to the field of journalism studies. As the author(s) highlight, it would be valuable to replicate the methodology and research focus in leading scientific journals within countries where the subject matter is particularly significant; I would cite Brazil in this regard, as authors and researchers there often face challenges in accessing top-tier international scientific journals. Furthermore, I believe it is important to disseminate the proposed article among journalism research associations worldwide.
Response:
We sincerely thank you for your encouraging assessment and for recognizing the relevance of our work to the broader field of journalism studies.
We fully concur with your suggestion to replicate this methodology in under-represented regions. Your specific mention of Brazil is particularly well-taken, as Latin American media markets offer distinct institutional, cultural, and regulatory contexts that would meaningfully extend the generalisability of our findings. In response, we have added the following sentence to the Future Research Directions section:
"Extending this methodology to under-represented media markets—particularly in Latin America, Africa, and South Asia—would test the generalisability of our findings across distinct regulatory, cultural, and technological contexts."
This explicitly names Latin America (including Brazil) as a priority region for replication. We will also actively seek appropriate conferences and academic venues to disseminate these findings following publication.
Changes made:
- Added replication recommendation to Future Research Directions section (p. 20), specifically naming Latin America, Africa, and South Asia.
Reviewer 2 Report
Comments and Suggestions for AuthorsCongratulations for selecting this object. You have provided an interesting review about the application of AI and machine learning in journalism and media research since if started. The article is original and provides useful information to the research community studying the applicability of intelligent systems to the media industry, as well as to journalism faculty and researchers.
Since no methodological or other flaws have been identified, we note below only a few improvements to the figures , some bibliographic references that need to be corrected and the References must be in alphabetical order:
- Some figures has illegible text due to their small size: Figures, 1, 8, 9, 10. Presumably, Figures 4 and 5 can be enlarged when the text is published and viewed on the journal's website, since any attempt to make out the words is impossible in a pdf or printed version.
2. The bibliographic references for Balloreta & Sandoval-Martín need to be corrected in both articles cited (2023 and 2024).
2.1. In the case of the 2023 reference published in the journal Cuadernos.info, the first author is Sandoval-Martín and Leonardo’s first last name is La-Rosa not Barrolleta (this one is the second name of this hispanic researcher).
Therefore, the complete reference in References must be change to:
Sandoval-Martín, Teresa, La-Rosa Barrolleta, Leonardo (2023). Research on the quality of automated news...
In the text the 2023 cited work must be change to (Sandoval-Martín & La-Rosa, 2023) instead of (Barrolleta & Sandoval-Martín, 2023).
2.2. In the case of the 2024 reference the correct one is not Barrolleta etc. it must be start with
La-Rosa Barrolleta, Leonardo, Sandoval-Martín, Teresa (2024). Artificial intelligence versus journalists...
In this case you also have to search for "Barolleta" and the 2024 references in the text must be change to (La-Rosa & Sandoval-Martín, 2024).
3. All the References are not in alphabetical format. Please, check them before you send an article.
4. In Results, 3.1., "The concentrarion of publications in 2025 and 2026 confirms that AI journalism research remains an actively devolping field". In the article it must be clear the period with the exactly months, specilly until which month of 2026 is done the systematic review. In this case, point 3.1., you can explain a little bit more, that since XXXX month of 2026 there are more articles than in the whole year of 2024, 2023 or 2021.
The exactly period of the systematic review, with the month when it starts in 2020 and when it finish must be cited in point 2. Materials and Methods.
Aside from these details, the methodology—which employs different approaches to the subject of study, thereby allowing for its validation—is considered a success and is viewed very positively in terms of recommending the publication of this work, which goes beyond a mere review by adding value through the use of different methods to validate the results.
I recommend reviewing the bibliographic references totally because, there appear to be more errors, such as the following:
- A comma is missing after the name María.Ángeles in the entry for Chaparro-Dominguez...
- A comma is missing after the name Mathias -Felipe in the entry de-Lima-Santos, Mathias-Felipe, Schapals, Aljosha, probable another come miss after Aljosha...etc.
- Kim, Daewon...coma
- Thurman, N, you use here ;
- Sonni... you use here ;
etc-. Please check.
A least, congratulations for writing the names when you know them, but please try to write all the possible ones due to applied the gender perspective in science as much as possible.
Author Response
Reviewer Opening Comment:
Congratulations for selecting this object. You have provided an interesting review about the application of AI and machine learning in journalism and media research since if started. The article is original and provides useful information to the research community studying the applicability of intelligent systems to the media industry, as well as to journalism faculty and researchers.
Response:
We are grateful for your positive assessment and for the detailed, constructive feedback that has substantially improved the manuscript. Each comment has been addressed as detailed below.
Comment 1 — Figure legibility:
"Some figures has illegible text due to their small size: Figures, 1, 8, 9, 10. Presumably, Figures 4 and 5 can be enlarged when the text is published and viewed on the journal's website, since any attempt to make out the words is impossible in a pdf or printed version."
Response:
We have taken the following measures to address figure legibility:
(a) Figure 1 (PRISMA flow): Redesigned using the PRISMA2020 R package for improved readability and layout accuracy (p. 4).
(b) Figures 8, 9, and 10: Replaced with LaTeX tables (Tables 4, 5, and 6 in the revised manuscript, pp. 12–13), providing clean, fully legible data display with appropriate font sizes. Figure 8 (top 10 highest-probability words) and Figure 9 (FREX words) are now presented as Table 4 and Table 5, respectively; Figure 10 (document-topic proportion distribution) is now Table 6.
(c) Figures 3 and 4 (keyword co-occurrence and author collaboration networks): Completely redesigned using the ggraph R package for improved legibility, with clearer node labels and colour-coded communities (pp. 8–9).
Comment 2.1 — Barrolleta & Sandoval-Martín (2023) reference correction:
Response:
We have corrected the 2023 bibitem to begin with "Sandoval-Martín, Teresa and La-Rosa Barrolleta, Leonardo." and updated the citation key from barrolleta2023 to sandoval2023. All in-text citations have been updated accordingly.
Comment 2.2 — Barrolleta & Sandoval-Martín (2024) reference correction:
Response:
We have corrected the 2024 bibitem to begin with "La-Rosa Barrolleta, Leonardo and Sandoval-Martín, Teresa." and updated the citation key from barrolleta2024 to larosa2024. All in-text citations have been updated accordingly (3 occurrences).
Comment 2.3 — Alphabetical ordering of references:
Response:
We have completely reordered all bibitem entries in alphabetical order by first author surname. The entire bibliography of 181 entries has been restructured accordingly.
Comment 2.4 — Specification of search period:
Response:
We have added to the Methods section (p. 4): "The search was conducted on 26 May 2026, covering publications indexed from January 2020 to March 2026."
Comment 2.5 — Reference formatting errors:
Response:
We have conducted a thorough audit of all 181 bibliographic entries. Author name formatting has been standardised throughout: all multi-author entries now use "and" consistently (not semicolons), and missing punctuation has been corrected.
Comment 2.6 — Gender perspective:
Response:
We acknowledge that a systematic gender perspective analysis was not within the scope of the present review, as the bibliometric metadata available through Scopus and WoS does not reliably capture author gender information, and the included studies themselves rarely report gender-disaggregated findings. We have added a note to the Future Research Directions section (p. 20) recommending that subsequent work examine gender dimensions in AI journalism research, including both the gender composition of authorship and potential gender biases in AI-mediated news production and reception.
Reviewer 3 Report
Comments and Suggestions for AuthorsThis manuscript addresses a relevant topic and reflects a considerable amount of effort. However, I believe its academic contribution would be significantly strengthened by improving its conceptual coherence, particularly with regard to the scope of the review and the proposed thematic framework.
The term “AI/ML applications/methods in media research” is somewhat misleading, as it primarily suggests the use of AI/ML as methodological tools for conducting media research rather than research on AI/ML applications in journalism and media practice. This ambiguity is reflected throughout the whole manuscript.
While research objectives seem to focus more on AI/ML as research methodologies, the findings and discussion predominantly review studies on AI use within the media industry. The review would benefit from clearer wording and a more consistent conceptual focus.
Another important source of conceptual ambiguity concerns the four thematic categories proposed by the authors.
For example, audience perception and content analysis represent distinct research areas, yet they are combined into a single category. Moreover, Section 3.6 (l. 338–369) on this category focuses on audience perception and does not provide a substantive discussion of content analysis. It is therefore unclear why content analysis is included in this category.
Similarly, the rationale for combining meta-research and implementation studies into a single category is unclear. Meta-research examines the development of scholarship, whereas implementation studies investigate the organisational adoption of AI. These represent different levels of analysis and would be more appropriately presented as separate categories unless a stronger conceptual rationale for their integration is provided.
More broadly, grouping conceptually distinct topics within the same category reduces the explanatory value of the reported percentages, as the aggregated figures obscure rather than clarify the distribution of the reviewed literature. The authors state that their “proposed four-category framework provides a structured approach for understanding how AI/ML technologies are being integrated into media research methodologies.” However, the proposed categorisation requires stronger conceptual justification.
Finally, as a minor point, Figure 7 does not include foundational scholars such as Diakopoulos, Carlson, and Coddington, although they are cited earlier in the manuscript as key contributors to the field.
Overall, two key aspects require greater conceptual clarity: the definition of the review's scope and the rationale behind the proposed thematic categorisation. The manuscript would benefit from a clearer distinction between AI for media research (AI as a methodological tool) and AI in journalism/media (AI as the object of investigation). This conceptual ambiguity recurs in the objectives, the thematic framework, and the conclusion. For instance, the concluding statement referring to the “integration of AI and ML methodologies in media research” (ll. 659–660) implies the former, whereas much of the review actually addresses the latter. Clarifying these aspects throughout the manuscript would substantially strengthen its conceptual coherence and make the review's scope and contribution easier to understand.
Comments on the Quality of English LanguageThe overall standard of English is generally acceptable and the manuscript is readable. However, the recurring conceptual ambiguity mentioned earlier, results in reduced conceptual clarity. A careful revision of the terminology and phrasing to ensure consistency with the stated objectives and analytical framework would substantially improve the manuscript's clarity and readability.
Author Response
Reviewer Opening Comment:
This manuscript addresses a relevant topic and reflects a considerable amount of effort. However, I believe its academic contribution would be significantly strengthened by improving its conceptual coherence, particularly with regard to the scope of the review and the proposed thematic framework.
Response:
We sincerely thank Reviewer 3 for this incisive critique, which has prompted us to substantially strengthen the conceptual foundations of our manuscript. The comments have been invaluable in clarifying both the scope of our review and the rationale for our analytical framework. We address each concern below.
Comment 1 — Scope ambiguity: AI/ML as research method vs. AI in journalism:
"The term 'AI/ML applications/methods in media research' is somewhat misleading, as it primarily suggests the use of AI/ML as methodological tools for conducting media research rather than research on AI/ML applications in journalism and media practice. This ambiguity is reflected throughout the whole manuscript."
Response:
We fully acknowledge this ambiguity and have undertaken a systematic revision of the manuscript to maintain a clear and consistent distinction between "AI in journalism" (the object of investigation) and "AI/ML as a research methodology" (a secondary concern). Specific changes include:
(a) Title revised from:
"Artificial Intelligence and Machine Learning Methods in Media Research: A Systematic Review of Applications and Limitations"
to:
"Artificial Intelligence in Journalism and Media Practice: A Systematic Review of Applications, Limitations, and Research Approaches"
(b) Abstract revised: The opening sentence now reads "investigates the application of artificial intelligence (AI) and machine learning (ML) in journalism and media practice, and how these applications are examined within scholarly research" — clearly establishing AI in practice as the primary object of study.
(c) Research objectives (Section 1, enumerations) revised as follows:
- RO1: "...primary applications of AI and ML technologies in journalism and media practice" (was "methodologies in media research")
- RO2: "...studies investigating AI/ML applications in journalism" (was "AI/ML-based media research studies")
- RO3: "...applying AI/ML technologies in journalism practice" (was "applying AI/ML methods to media research")
- RO4: "...how AI/ML technologies are being studied across different research traditions in journalism" (was "evaluating the integration of AI/ML technologies in media research methodologies")
- RO5: "...transforming journalistic practice and media industry workflows" (was "transforming traditional media research practices")
(d) Full-text lexical audit: We systematically reviewed every occurrence of "AI/ML methods in media research" throughout the Introduction, Discussion, and Conclusions, replacing ambiguous phrasing with context-appropriate terms:
- "AI/ML applications in journalism" where referring to the phenomenon under study
- "methodological approaches used to study AI in journalism" where referring to research methods
- "AI technologies in news production" where referring to industry practice
(e) Concluding paragraph (Section 7): Rephrased from "integration of AI and ML methodologies in media research" to "integration of AI and ML technologies into journalism practice and their scholarly examination."
Comment 2 — Rationale for combining conceptually distinct categories:
"Audience perception and content analysis represent distinct research areas, yet they are combined into a single category. Moreover, Section 3.6 on this category focuses on audience perception and does not provide a substantive discussion of content analysis."
"Similarly, the rationale for combining meta-research and implementation studies into a single category is unclear. Meta-research examines the development of scholarship, whereas implementation studies investigate the organisational adoption of AI. These represent different levels of analysis."
Response:
We appreciate this observation and have strengthened the conceptual justification in two places:
(a) Added an explicit rationale paragraph in Methods (Section 2.3, after the framework description):
"The pairing of audience perception with content analysis reflects their complementary focus on the reception and interpretation of AI-mediated content. The former examines subjective human responses (trust, credibility, emotional reactions) through experimental and survey methods, while the latter deploys computational tools for systematic text and image analysis. Although methodologically distinct, these two sub-areas converge thematically: both investigate how AI-generated content is consumed, evaluated, and interpreted. Similarly, meta-research (studies examining the scholarly field itself) and implementation studies (studies investigating organisational adoption) are united by their secondary analytical position—both operate at a reflexive distance from primary AI applications, contributing to the field's self-understanding and its practical deployment conditions. The independent identification of K = 4 as the optimal topic solution by the STM provides empirical validation for the four-category structure."
(b) Strengthened the sixth limitation in Section 6 (Limitations):
"The classification framework, while empirically grounded and independently validated by STM (K = 4), organises studies across partially overlapping analytical dimensions. Audience perception and content analysis, though thematically convergent, employ different methodological traditions (experimental/survey vs. computational/text-analytic). Meta-research and implementation studies operate at different levels of analysis (field-level vs. organisational). The STM's independent identification of the four-topic structure provides empirical validation for the framework, but the quantitative proportions should be interpreted as indicative distributions that capture thematic affinities rather than rigid partitions."
Comment 3 — Section 3.6 (now 3.4): Content analysis discussion is missing:
Response:
We have revised Section 3.4 (Category 2) to include a substantive paragraph on AI/ML content analysis studies. The new paragraph (inserted after the audience perception discussion) covers:
- Studies using NLP for automated news framing detection
- Computer vision applications for visual bias detection in AI-generated news imagery
- Automated fact-checking and misinformation detection tools
- The relatively small number of content-analysis-focused studies in the corpus (n=3) and the corresponding need for further research in this area
This ensures that the category heading ("Audience Perception and Content Analysis") is fully substantiated in the body of the section.
Comment 4 — Figure 7 missing foundational scholars:
"Figure 7 does not include foundational scholars such as Diakopoulos, Carlson, and Coddington, although they are cited earlier in the manuscript as key contributors to the field."
Response:
We thank the reviewer for this careful observation. Upon investigation, we found that the top-10 author ranking (Figure 7) is based on first-author publication counts within the 70 included studies. Foundational scholars such as Diakopoulos, Carlson, and Coddington appear frequently as co-authors in the corpus—and their works are extensively cited throughout our manuscript—but their first-author contributions within the 2020–2026 window place them outside the top 10. We have added the following clarifying sentence in the Results section:
"While foundational scholars such as Diakopoulos, Carlson, and Coddington are frequently cited throughout the corpus and appear as co-authors on multiple included studies, their first-author publication count within the 2020–2026 window places them outside the top-10 ranking shown in Figure 7. This pattern reflects the collaborative nature of computational journalism research, where senior scholars often contribute as co-authors rather than sole or first authors."
Comment 5 — Overall conceptual coherence:
"The manuscript would benefit from a clearer distinction between AI for media research (AI as a methodological tool) and AI in journalism/media (AI as the object of investigation). This conceptual ambiguity recurs in the objectives, the thematic framework, and the conclusion. For instance, the concluding statement referring to the 'integration of AI and ML methodologies in media research' implies the former, whereas much of the review actually addresses the latter."
Response:
We agree completely. As detailed in our response to Comment 1 above, we have systematically revised the manuscript at every level—title, abstract, research objectives, thematic framework narrative, discussion, limitations, and conclusion—to maintain a consistent and clear distinction between:
- AI/ML as technologies deployed in journalism practice (primary focus)
- AI/ML as methodological tools for scholarly research (secondary, contextual focus)
The concluding paragraph now reads: "This systematic review has provided comprehensive insights into the integration of AI and ML technologies in journalism practice and their scholarly examination, analysing 121 peer-reviewed publications spanning 2020 to 2026."
We thank you for pushing us to articulate these conceptual foundations more clearly. The manuscript is substantially stronger as a result.
