Generative AI and the New Landscape of Automated Journalism: A Systematized Review of 185 Studies (2012–2024)

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This manuscript updates work from 2021 that focused on 2005-2020 research that included only 33 articles. The current review encompasses a much larger data set (185 studies) from between 2012-2024, which dates back to the first articles identified in the Danzon-Chambaud work.

Overall, the manuscript is well-organized and methodical. The writing is generally good, as well (with some noted exceptions below). 

I found a few areas that need to be re-examined for completion. For example, Figure 1 has missing "n"s (Reference from other sources and References removed). 

In Table 7, you include Prolific as an AI tool, but most consider this a recruitment tool similar to Mechanical Turk. Can you explain how you categorized these tools? Additionally, you refer to crowdsourcing sites but don't include them in the table. I suggest you add a section for that. 

In Table 7, I also wondered why you differentiated Google Gemini from platforms like ChatGPT. The models would be GPT 4.0 or something like that, but the platforms are very similar. 

Table 3 and Table 8 confuse me. What are the differences between these two tables? Table 3 is "keywords" and Table 8 is "key terms" but I don't see where you're getting "terms" from. I assumed keywords were those tagged in the databases? Please be explicit about where you getting these items from and how you are including them. 

For Table 12, you are showing "prevalence" numbers but I don't understand what you mean by that. You say it's not exact number of studies. Are you saying a study could be in more than one category? If so, please reword the description in the text and add a note in the table for more clarity. 

Can you give an example of what would qualify as a "general overview" and how you determined those studies would be excluded from this analysis? This has important implications since, as you note, it excludes works focused on several understudied parts of the world. 

In the literature review, you discuss how these terms are used to describe several aspects of journalism including the reporting process, the writing process, production and dissemination. However, that is not addressed in the rest of the paper. It would be helpful to understand how many of these studies are looking at outputs or use of automation in the writing process, versus how many are considering things like recommendation bots or data analysis assistance. The implications could be helpful on identifying gaps where research could do more to define and analyze relevant practices.

I’d like to see the authors do more to add implications and insights beyond those discussed in Danzon-Chambaud’s work. What has the influx of research done for the larger picture of study of AI and its impact on journalism? How can this work be used for future inquiry? I’d like to see more than “let’s all use the same language”.

For example, more work from female scholars and people from marginalized communities, more focus on mapping non-English research in this area to see how it changes the dynamics overall, A discussion of bias in research because of the overarching focus on Western contexts is merited here, as well. There’s not much by way of theory, either. Why are we not using something like technology acceptance theory or planned behavior theory to explain why people use/perceive the AI-generated content or AI-assisted production differently?

I think you’ve got a lot here but more could be done to make it influential for scholars in this area of research.

Errors in grammar/spelling/syntax: 

• On lines 229, 298, 340-342, 371, 378, 395, 399, 410, 416, 422, 437, 453, 459, 487, 492, 503, 522, 540, 586:  (Error! Reference source not found) is in the manuscript.
• On line 191, “Unlike previous AI-enabled tools, which made journalists’ able to do their work more” I think there’s a word wrong or missing here.
• On line 251, “This type of research synthesis is includes elements of systematic review, best suited to assess the quantity” – extra “is”
• On line 391, “While the top five journals in our 390 study fell under the discipline of communications within social sciences, we in fact studies 391 from 46 subjects within 14 disciplines,…” – “we in fact studies” doesn't make sense
• Line 514, “ibility were4 grouped together” Random 4 in the text
• Line 536, “one” should be “on” 
• On Line 588, “deluge of available terms to d3escribe automated journalism” – extra 3

 

Author Response

We would like to thank both reviewers for engaging with our work so thoroughly and for their valuable recommendations and positive feedback.

We have responded to all concerns raised in the table below and modified the manuscript accordingly, which we believe has now strengthened our manuscript.

Once again, many thanks!

Section	Comment	Response
Method	Figure 1: Missing n numbers in the PRISMA flow chart where it says ‘references from other sources’, ‘citation searching’, and ‘grey literature’	Agreed. We have added these missing numbers in Figure 1.  In addition, we noticed that these numbers needed to be updated in the text, which has been done.
	Table 7: Prolific is mentioned as an AI tool but should be recruitment tool, crowdsourcing sites missing from the table, Google Gemini is listed as AI tool but should be language model like ChatGPT	Agreed. Prolific was mistakenly miscategorized. In our data set we grouped it with the crowdsourcing tools and renamed Recruitment, added this row to the table, and updated the numbers. Google Gemini was moved to the Language Model category, and specified as Google GeminiAI to distinguish it from the Chatbot tool.
	Table 8: Unclear how raw terms (here) differ from keywords (Table 3) – need to add a line here to explain the differentiation	Agreed. In order to distinguish from author-provided keywords, we changed key terms to “search terms,” which was the title of this section. We also adjusted RQ2 to reflect this. Additionally, we provided an example just before Table 8  to distinguish between author-provided keywords and the various terms authors were including in their studies to describe automated journalism.
Conclusion	Table 12: Unclear what ‘prevalence’ means in this context – need to add a line here to explain how the numbers come about	Agreed. We have added a note below Table 12 to explain these numbers.
Discussion	‘General overview’: Unclear why these studies were excluded – need to add a line here to justify exclusion	Agreed. We have added a line in the Discussion and further up in the manuscript to explain the ‘general overview’ criterion.
Literature Review	Suggestion to tease out how many studies reference AI tools for what stage of the news production process.	Partially agreed. While we agree that examining the use of AI tools at various stages and for various purposes would be an interesting and valuable pursuit, it is unfortunately beyond the scope of our analysis. We are planning a subsequent study that will examine how AI tools are used in particular.
Conclusion	Suggestion to tease out the impact of this study beyond the need for conceptual clarity, and to reference other theories.	Agreed. We have found both of these suggestions particularly valuable and have added a paragraph each in the Conclusion section.
Grammar / Spelling / Syntax	Suggestion to make changes according to the list provided.	Agreed. We have made all the requested changes in the manuscript.

Reviewer 2 Report

Comments and Suggestions for Authors

The study presented is of a significant relevance for the area as well as a rigorous and appropriate methodology, as are the respective proposals for future research. Concerning the text, although it presentes flaws that with a last reading would not appear, those flaws do not interfere in the global and final decision. 

The attached pdf signals the information bellow on commentary boxes.

Most important indications:

lines 178 + 186 = From what I understand, this reference should be (Shi & Sun, 2024) [from Shi, Y., & Sun, L. (2024). How Generative AI Is Transforming Journalism: Development, Application and Ethics. Journalism and Media, 5(2), 582-594. https://doi.org/10.3390/journalmedia5020039]

line 209 = "In 2021, Danzon-Chambaud (2021)" - doble dated

lines 216 + 268 + 375 + 557 + 626= " Danzon-Chambaud based his..." - advice to include date

lines 229 + 298  + 340 + 342 + 371 + 395 + 399 + 410 + 416 + 422 + 437 + 453 + 459 + 487 + 492 + 503 + 522 + 540 + 586  = "Error! Reference source not found." - substitute by the intended figure/table [This abscense is not understandable...]

line 426 = "Figure 4. Regions of individual study. " - figure 4 has no caption for color context

line 486 = "grouping, you also start to..." - should be read "grouping, we also start to..."

line 535 = "and so one..." - should be read "and so on..."

line 535 = " it is also. We built..." - the sentence 'it is also' is not finished.

line 535 = Shi, Y., Sun, L., Shi, Y., & Sun, L. (2024). - Correct to: 
Shi, Y., & Sun, L. (2024). How Generative AI Is Transforming Journalism: Development, Application and Ethics. Journalism and Media, 5(2), 582-594. https://doi.org/10.3390/journalmedia5020039

Comments for author File: Comments.pdf

Author Response

We would like to thank both reviewers for engaging with our work so thoroughly and for their valuable recommendations and positive feedback.

We have responded to all concerns raised in the table below and modified the manuscript accordingly, which we believe has now strengthened our manuscript.

Once again, many thanks!

Section	Comment	Response
Grammar / Spelling / Syntax	Suggestion to make changes according to the list provided.	Agreed. We have made all the requested changes in the manuscript.