Artificial Intelligence and Journalistic Ethics: A Comparative Analysis of AI-Generated Content and Traditional Journalism

Abstract

This article presents a comparative study of content generated by artificial intelligence (AI) and articles authored by professional journalists, focusing on the perspective of a Kazakhstani audience. The analysis was conducted based on several key criteria, including the structure of the article, writing style, factual accuracy, citation of sources, and completeness of the information. The study spans a variety of topics, such as politics, economics, law, sports, education, and social issues. The results indicate that AI-generated articles tend to exhibit greater structural clarity and neutrality. On the other hand, articles written by journalists score higher in terms of factual accuracy, analytical depth, and the use of verified sources. Furthermore, the research explores the significance of journalistic ethics in ensuring transparency and information completeness in content production. Ultimately, the findings emphasize the importance of upholding rigorous journalistic standards when integrating AI into media practices.

1. Introduction

Every year, the use of artificial intelligence (AI) technologies is increasing in various fields, including journalism. As newsrooms strive to keep up with the fast pace of information production and dissemination, AI has become a practical tool for improving efficiency, generating content, automating tasks, and enhancing audience engagement (Yuan et al., 2024; Caswell, 2022). These technologies are now used for writing, editing, transcribing, translating, fact-checking, and even generating multimedia content (Banafi, 2024; Iaroshenko, 2024).

This automated content generation is made possible through two core technologies: Machine Learning (ML) and Natural Language Processing (NLP). ML allows systems to identify patterns and improve performance based on data input without being explicitly programmed (Solanki & Jain, 2020), while NLP enables machines to understand, interpret, and generate human language in a meaningful way (Carrasco Ramírez, 2024). These technologies lie at the heart of generative AI tools such as ChatGPT-3.5 model and Bard, which are increasingly utilized in modern journalistic workflows.

However, AI’s growing presence in journalism raises important questions beyond productivity. In particular, ethical challenges are becoming more prominent. Issues such as transparency, factual accuracy, source attribution, and the reliability of AI-generated content have sparked debate among media professionals and scholars (Gutiérrez-Caneda et al., 2024; Noain Sánchez, 2022; Sonni et al., 2024). While AI tools can improve speed and reduce workload, they may also distort facts, fabricate citations, or mislead audiences (Dhiman, 2023). Consequently, the role of human oversight remains critical in ensuring journalistic integrity.

This study explores how AI-generated news content compares to that written by professional journalists, with a particular focus on ethical considerations such as accuracy, transparency, and source attribution. This article also examines how Kazakhstani audiences perceive AI-generated texts and whether they find them as trustworthy and informative as human-written journalism.

According to the Global AI Readiness Index, Kazakhstan ranks 48th in the world, with a score of 0.55, making it the leading country in Central Asia in terms of readiness to adopt artificial intelligence, second only to Russia in the broader region (Caspian Post, 2024). This growing national interest highlights the urgency of understanding how AI may affect journalistic practices and public trust in news content.

The aim of this study is to compare articles written by artificial intelligence and traditional journalists within the context of a Kazakhstani audience.

The hypothesis of this study is that while AI-generated content demonstrates strong structure and a neutral, balanced tone, it often lacks depth in information and may present distorted facts.

RQ 1: In what ways do AI-generated articles differ from those written by professional journalists regarding structure, writing style, factual accuracy, citation of sources, and completeness of the information?

RQ 2: How are AI-generated texts and those written by professional journalists perceived by the Kazakhstani audience?

2. Literature Review

2.1. Evolution of AI in Journalism

At the initial stage, AI was used as an expert system to solve complex problems in such fields as engineering, business, science, and medicine (Pannu, 2015). Later on, the technology became more widespread in fields such as medical image classification, video games, accounting, and advertising (Kamble & Shah, 2018). During the 1960s and 1970s, AI technologies started to be incorporated into journalistic practices (Mari, 2024). In particular, work was carried out on spelling checks and automated spelling for sports and financial news. Since then, technology has increasingly become embedded in journalism, laying the groundwork for the rise of automated journalism. One of the pioneers in this area was the Associated Press, which was among the first to adopt AI technologies in 2014 to generate reports in the financial news sector (Blankespoor et al., 2018).

2.2. Capabilities of AI in Journalism

Today, AI technologies play a role in many different aspects of the journalistic workflow (Chan-Olmsted, 2019). One key area is writing and fact-checking, where many newsrooms have started using AI tools (Parratt-Fernández et al., 2021). For example, Banafi (2024) notes that AI is used in media organizations for content creation, transcribing speech to text, translating languages, and extracting and analyzing data.

In addition, AI offers journalists a variety of practical functions, such as writing code or editing PDF documents (Dhiman, 2023). In professional journalism, AI helps improve productivity by automating the writing of basic news stories (Ruiz & Sánchez, 2019). Several studies have shown that AI tools make tasks easier to complete, help monitor content, and improve the accuracy of fact-checking (Gutiérrez-Caneda & Vázquez-Herrero, 2024; Gonçalves et al., 2024; Sultan et al., 2024). For instance, the Factmata tool is used for fact-checking and generating automated reports (Sukhodolov et al., 2019), while ChatGPT has proven useful in writing journalistic texts and generating content ideas (Gutiérrez-Caneda et al., 2023). Tools like these can reduce journalists’ daily workload and help save time (Iaroshenko, 2024).

Another area where artificial intelligence is used in journalism is data analysis and visualization. To illustrate, Alimzhanova et al. (2024) highlight how visualizations produced with the help of AI can support journalists in developing more engaging and interactive content, which, in turn, makes complex information more accessible to readers. Verma (2024) also points out that AI can significantly improve data analysis processes, allowing for more tailored content, increased productivity, and a range of practical advantages for media professionals. Using AI tools to generate news headlines and write summaries can also significantly simplify the production process and increase audience engagement (Yuan et al., 2024). One more benefit is that tools like ChatGPT and Bard assist journalists with tasks such as language translation and data gathering (Dhiman, 2023).

2.3. Ethical Concerns in AI-Powered Journalism

Broadly speaking, AI is increasingly being used to automate a wide range of tasks within journalism—from generating and refining multimedia content to its delivery and dissemination to the public (Aissani et al., 2023). These technologies can ease the workload for journalists and bring certain advantages, especially when it comes to improving efficiency. That said, their use is not without drawbacks. One of the most pressing concerns is the lack of clear adherence to ethical principles.

Several studies have noted that the rise of the digital age—along with violations of ethical codes and the expansion of cybermedia—has posed new challenges with respect to traditional journalistic ethics (Suardana, 2020). One such challenge is the use of AI tools in journalism. For example, Amponsah and Atianashie (2024) highlight ethical concerns around the use of AI, including bias in data and breaches of confidentiality. Such concerns are often linked to algorithmic bias, which refers to systematic errors in AI decision-making processes that arise from biased or incomplete training data. This can lead to skewed or discriminatory outputs, even if the algorithm itself is technically neutral (Stinson, 2022; Cossette-Lefebvre & Maclure, 2023). In journalism, such bias may result in inaccurate representations, the omission of minority perspectives, or reinforcement of existing stereotypes.

According to Tolnaiová (2023), ensuring transparency and accountability while using AI is one of the most challenging tasks. Transparency, for instance, requires that journalists openly disclose whether and how AI tools were used during the production of news content, regardless of whether the software is commonly accepted. However, such disclosures are rarely standardized across media organizations. Moreover, accountability becomes problematic because AI systems do not possess moral agency—errors, biases, and misinformation generated by AI often lack clear attribution, making it difficult to assign responsibility (Mukta, 2025). In some cases, AI tools are prone to providing false data (Ayub, 2024). Moreover, there are challenges related to data accessibility, accuracy requirements, and conveying news value (Stray, 2019). Automating news production ensures objective, basic content (Carreira & Squirra, 2017). However, issues can arise in terms of textual structure. For example, Ruiz and Sánchez (2019) found that some AI tools lack the ability to generate complex texts.

Another notable concern is that AI systems can, in certain situations, fabricate citations, distort facts, and spread misinformation. According to Kidd and Birhane (2023), there is growing concern that AI models could mislead users by providing false information and distorting public trust. Therefore, editorial oversight and the involvement of professional journalists during the use of AI tools are seen as essential to ensuring high-quality and reliable information (Peña-Fernández et al., 2023; Dijkstra et al., 2024).

2.4. Comparative Studies: AI vs. Human Journalism

One of the most intriguing topics in scientific journalism today is the comparative analysis of AI-generated content versus human-created work. In recent years, several studies on this subject have been conducted and published, with results divided into two main perspectives. The first set of studies found no significant difference in terms of reliability and accuracy between articles written by AI and those written by journalists (Henestrosa et al., 2023; La-Rosa Barrolleta & Sandoval-Martín, 2024). In fact, some of these studies even concluded that AI-generated articles were more trustworthy than those written by human journalists.

On the other hand, the second set of findings showed that readers tend to value human-created content more highly, particularly due to the depth of analysis and interpretation provided in the articles (Kanasheva et al., 2024). A study conducted by Nah et al. (2024) reveals that machine-generated news tends to focus on a specific topic, whereas human-written articles cover a broader range of subjects. These studies highlight the potential for maintaining a balance between journalists and artificial intelligence, demonstrating opportunities for collaboration (El Nemr, 2024).

2.5. Research Gaps

So far, existing research has offered a broad look at both the benefits and the limitations of using artificial intelligence in journalism. In addition, some studies have analyzed the differences in texts written by AI and journalists, in addition to offering assessment from the perspective of the audience. Still, much of this work has centered on media environments in developed countries—mainly the United States and Western Europe—while regions like Kazakhstan and other parts of Central Asia remain largely overlooked. Despite Kazakhstan’s relatively strong position in the global AI readiness index (48th globally, with a score of 0.55), there is still a lack of empirical research on the ethical and communicative implications of AI-generated content in local journalism.

On top of that, there is a noticeable gap in studies comparing how AI interacts with a wide range of journalistic topics, such as politics, economics, law, sports, education, science, and social issues. Another missing piece is the lack of research into how well AI tools perform when working with texts in the Kazakh language. This study sets out to help fill those gaps by examining how AI technologies might be applied within journalism in less-studied regions, with a particular focus on Kazakhstan. It also examines how AI works with the Kazakh language, how the local audience perceives it, and how AI-generated texts compare with journalists’ work in terms of ethical standards, ultimately contributing to the field of knowledge.

3. Theoretical Framework

3.1. Journalistic Ethics Theory

In journalism, ethical principles are essential for creating reliable, authentic content for the general public (Roberts, 2019). The main ethical principles include respect for privacy, objectivity, confidentiality, and the fight against censorship (Jacquette, 2016). These standards are grounded in fairness and a commitment to truth and accuracy (Rezaee et al., 2024).

A central concept within journalistic ethics is objectivity. According to Ward (2011), objectivity is not about achieving perfect neutrality but following a method of “pragmatic objectivity”—a practical, disciplined process that strives for fairness and transparency in reporting. Similarly, Kovach and Rosenstiel (2001) define objectivity as a method of verification, emphasizing rigorous fact-checking and the clear presentation of sourced information. These approaches reinforce the idea that objectivity is a practice, not an absolute state, and rely on human judgment, context awareness, and accountability.

In addition, journalists are expected to maintain transparency in their writing, properly cite their sources, and work responsibly with their references—all of which are part of professional standards. As Phillips (2010) points out, transparency helps protect professional journalism and keeps it aligned with ethical guidelines.

Another strict requirement placed on the media is the proper handling of sources. According to Cushion et al. (2016), journalists often provide inaccurate references when citing statistical data. As a result, there is a growing tendency to approach statistics with more skepticism (Albright, 2017), which, in turn, creates room for the spread of misinformation.

In the context of artificial intelligence in journalism, these ethical standards face new challenges (Sigsgaard, 2024). AI tools lack moral judgment and decision-making capacity, which raises questions about how well AI-generated texts align with traditional journalistic norms. While AI can assist in generating news content quickly and efficiently, its inability to evaluate context, verify facts, or consider the societal impact of content may lead to ethical concerns, such as misinformation, biased reporting, or fabricated sources (Dhiman, 2023).

This study draws on journalistic ethics theory to assess and compare AI-generated and journalist-written texts in terms of their ethical quality. Special attention is given to aspects such as factual accuracy, proper source citation, objectivity, and adherence to professional norms. The goal is to determine whether AI-generated content meets the ethical expectations set by the journalistic profession or whether human oversight remains essential to maintain ethical integrity.

3.2. Media Credibility and Audience Perception

Media credibility plays a crucial role in how audiences engage with and evaluate journalistic content. It refers to the perceived trustworthiness, reliability, and professionalism of media messages and is shaped by various factors, such as writing quality, tone, sourcing, and the reputation of the content producer (Atish, 2024). In traditional journalism, credibility is often built through consistent adherence to professional norms, including accurate reporting, transparency, and ethical responsibility (Koliska, 2021). For example, an empirical study by Curry and Stroud (2021) found that higher transparency in journalistic content leads audiences to rate news stories as more trustworthy.

In the context of artificial intelligence, the question of media credibility becomes more complex. Unlike human journalists, AI lacks the capacity for moral judgment, contextual understanding, and editorial discretion. This has prompted researchers to examine how audiences perceive AI-generated news in comparison to content written by professional journalists. For instance, some studies suggest that when audiences are unaware of a text’s origin, they may judge AI-generated content to be equally as credible or even more credible due to its neutral tone and structured format (Henestrosa et al., 2023; Clerwall, 2014). However, once informed that a piece is machine-generated, audience trust tends to decline, especially in relation to depth of analysis, originality, and ethical standards (Kanasheva et al., 2024; Nah et al., 2024).

Similarly, some studies have found no significant difference in perceived credibility between AI- and human-authored texts when the content is simple and factual (Henestrosa et al., 2023). However, in more complex subject areas, AI authorship has been associated with lower perceived credibility of both the message and its source (Henestrosa & Kimmerle, 2024).

In this study, the credibility, transparency, and accountability of AI- and human-authored news texts are assessed across six topics—politics, economics, law, sports, education and science, and social issues—to explore how audience trust varies by content area and ethical standards.

4. Materials and Methods

This study employs a comparative analysis method to examine the differences and similarities between articles written by artificial intelligence and those created by traditional journalists. To achieve this, we selected articles generated using generative language models (GPT-4) and materials published by major news agencies in Kazakhstan, including Azattyq, Inform.kz, Sputnik.kz, Zhas Alash, Minber.kz, and Tengrinews.kz. These media outlets were selected because they are among the most widely recognized and influential news organizations in Kazakhstan. Each outlet covers a broad range of topics, including politics, economics, law, sports, education, and social issues, making them highly relevant for this study.

Furthermore, none of the selected outlets is state-owned. In Kazakhstan, state-run media tend to focus on brief official announcements and press releases, often lacking analytical depth. At the same time, the government employs various mechanisms to maintain control over mass media, which results in a lack of independent analysis (Utemissov & Koshkenov, 2021). In contrast, independent media organizations are known for publishing more in-depth, analytical, and critical reporting compared to state-run media (Atay, 2025). These outlets are often more trusted by the public due to their analytical depth, investigative reporting, and relative freedom from state control.

The comparison was based on several key criteria: article structure, writing style, factual accuracy, source citation presence, and the level of information coverage. Furthermore, the study utilized content analysis as a methodological approach, with a central aspect being the evaluation of adherence to journalistic ethical standards in the writing process, including the absence of bias, the balanced presentation of information, and the maintenance of objectivity and neutrality.

To demonstrate the significance of the study, we conducted an evaluation of articles written by both artificial intelligence and traditional journalists among a Kazakhstani audience using a survey method. The survey was distributed via a closed WhatsApp group titled “Liga doktorantov KazNU”, which consists of 841 members—primarily doctoral students and university faculty affiliated with Al-Farabi Kazakh National University. The group is used as a collaborative space where members support each other’s academic research and often participate in surveys voluntarily. It was completed by 97 respondents who volunteered between October and December 2024. As the data was being analyzed, it was noted that three individuals provided incomplete answers, so a total of 94 responses (96.9%) were used in the final analysis. The sample consisted of highly educated respondents, all with higher education degrees. Most were either doctoral students or university lecturers, making them a highly literate audience in terms of research and information evaluation. The respondents’ age ranged from 26 to 48 years old.

Participants were given articles with the authors’ names removed and asked to evaluate the texts based on criteria such as the reliability of the article structure, writing style, factual accuracy, citation of sources, and completeness of information. The questions were closed-ended and used a comparative format—participants were asked to choose which of the two articles (AI-generated or journalist-written) performed better for each criterion. These five core questions were replicated across all six topic areas, resulting in a total of 30 comparative responses per participant. A total of twelve articles were presented to the survey participants—six generated by artificial intelligence and six written by professional journalists. The survey was conducted online via Google Forms.

The study is based on two data sets. The first set includes news and information on political, economic, legal, sports, educational, scientific, and social topics, all generated by artificial intelligence. The second set consists of articles written by experienced journalists from major media outlets and published in Kazakhstani news sources.

The study analyzed articles across six key areas:

• Politics—The mass protests in Kazakhstan in January 2022. For analysis, the article “Complete Chronology of the January Events” from Azattyq, published on 28 February 2022, was selected.
• Economy—GDP growth. The article “Kazakhstan’s Economy in 2022: GDP Growth, New Industries, and a Record Grain Harvest”, published by Inform.kz on 10 January 2023, was chosen for analysis.
• Law—Domestic violence. The article “Bishimbayev Case: A Chronology of the Crime”, published by Sputnik.kz on 2 April 2024, was selected for analysis.
• Sports—The performance of Kazakhstani athletes at the Paris 2024 Olympics. The article “Paris-2024: 43rd Place Reflecting Kazakhstan’s Sports Results”, published by Zhas Alash on 13 August 2024, was used for analysis.
• Education and Science—University grants in 2024. The article “2024 Grants: 112,000 Scholarships for Higher Education Institutions”, published by Minber.kz, was selected for analysis.
• Social Issues—Flooding in Kazakhstan. The article “Flooding in Kazakhstan: What We Know So Far?”, published by Tengrinews.kz on 15 April 2024, was used for the analysis.

First, for each of the six thematic areas—politics, economy, law, sports, education and science, and social issues—a topic that had generated strong public resonance in Kazakhstan was identified. Then, for each selected topic, articles were searched across six different media outlets, and the article with the highest view count on its respective website was chosen for analysis.

The selected texts represent different journalistic genres, including analytical reporting and straight news. For example, the article from Azattyq (politics) provides an in-depth analytical review, while the ones from Inform.kz (economy) and Tengrinews.kz (social issues) are brief news items. Similarly, articles from Sputnik.kz (law), Zhas Alash (sports), and Minber.kz (education and science) vary in structure and tone—some being analytical and others informational. This variation helped assess AI’s ability to mirror different styles and genres of journalistic writing.

To generate an article using artificial intelligence, a specific query was submitted (see

Table 1. Queries used to generate articles with artificial intelligence.

Area	Topic	Query Format
Politics	The mass protests in Kazakhstan in January 2022	Write an article that provides a comprehensive overview of the protests in Kazakhstan in January 2022, maintaining neutrality and relying on verified data
Economy	GDP growth	Write an article that provides a comprehensive overview of Kazakhstan’s GDP growth over the first 11 months of 2022, maintaining neutrality and relying on verified data
Law	Domestic violence	Write an article about the case of former Minister Kuandyk Bishimbayev in 2024, maintaining neutrality and relying on verified data
Sports	The performance of Kazakhstani athletes at the Paris 2024 Olympics	Write an article about the achievements of Kazakhstani athletes at the 2024 Paris Olympics, highlighting their participation in various events, while maintaining neutrality and relying on verified data
Education and Science	University grants in 2024	Write an article about the grants allocated to universities in Kazakhstan in 2024, covering the specific areas and levels of funding, while maintaining neutrality and relying on verified data.
Social Issues	Flooding in Kazakhstan	Write an article about the flooding in Kazakhstan in 2024, maintaining neutrality and relying on verified data

). The query was made in Kazakh, as the research focuses on a Kazakh-language audience. This approach enabled an assessment of AI’s performance in Kazakh, as well as its adherence to ethical principles. Kazakhstan remains under-represented in global communication and media research, making it an important case for examining how emerging technologies such as AI interact with local journalistic practices, especially in the Kazakh language.

All six articles were selected due to their popularity among the Kazakhstani audience and their provision of comprehensive information on key events. The study found that AI struggles to effectively generate articles on brief, fact-based events. Additionally, the articles of varying length from Kazakhstani media were chosen (research papers, news reports, and reviews), allowing for a more nuanced evaluation of AI’s capabilities in journalism.

5. Results

5.1. News from Humans or Machines? A Comparative Study

5.1.1. News 1. Politics—The Mass Protests in Kazakhstan in January 2022

Structure of the article: The AI-generated article was titled “Protests in Kazakhstan: Causes, Developments, and Consequences” and included sections such as:

• Causes of the Protests;
• Development of the Protests;
• Government Response;
• Casualties and Damage;
• Consequences and Conclusions.

In comparison, no article from Kazakhstani media offers this level of structural detail. The Azattyq website features an article titled “Complete Chronology of the January Events”, which presents the events in chronological order but lacks the clear sectional divisions of the AI-generated article.

Writing style: The AI-generated article features a clear and coherent writing style. The sentences flow logically, making the text easy to understand. The information is presented in clear, straightforward Kazakh, without unnecessary complexity, and adheres to journalistic ethics, ensuring balance and a neutral tone, though this should not be equated with professional journalistic objectivity. In contrast, the Azattyq article emphasizes the protesters, while the government’s response is presented in a more emotionally charged tone. This style risks a one-sided interpretation and reduces the article’s objectivity.

Factual accuracy: The AI-generated article presented several factual inaccuracies:

• Timing of the protest: The AI-generated article indicated that the protest began on December 31. However, the Azattyq article correctly states that the protest started on January 2. After this error, we submitted the same query a second time, and the AI revised the start date to January 1.
• City name errors: The AI mentioned Nur-Sultan instead of Astana. It is important to note that Nur-Sultan was officially renamed Astana on 17 September 2022.
• Accuracy of the death toll: The AI broadly stated that more than 200 people died during the protests. In contrast, Azattyq, citing data from the General Prosecutor’s Office, reported that 225 people were killed. Additionally, Azattyq mentioned that 4578 individuals were injured.

The artificial intelligence provided some accurate data, including the increase in the price of liquefied gas from KZT 60 to 120 per liter, the timing of the request for assistance from the CSTO, the arrival timeline, and the cities where the protests took place.

Citation of sources: The AI-generated article did not provide citations of specific sources. In contrast, the Azattyq article relies on official data from the press service of the General Prosecutor’s Office.

Completeness of the information: The AI article omits details such as the protesters’ demands, businesses that were affected, and President K. Tokayev’s order to open fire without warning. These elements are included in the Azattyq article.

5.1.2. News 2. Economy—GDP Growth

Structure of the article: The AI-generated article proposed the title “Kazakhstan’s GDP Growth for the First 11 Months of 2022: Key Indicators and Causes”. It is divided into seven sections:

• Overall GDP Growth Indicators;
• Changes in GDP Structure;
• Industry and Energy;
• Agriculture;
• External and internal Factors;
• Challenges impacting GDP Growth.

On the other hand, the Inform.kz article titled “Kazakhstan’s Economy in 2022: GDP Growth, New Industries, and a Record Grain Harvest” is organized into six sections:

• GDP Growth;
• Opening New Industries;
• The Largest Grain Harvest in 10 Years;
• Fighting Inflation;
• Priorities for Small and Medium Businesses;
• Investments in Kazakhstan’s Economy.

Both articles feature a clear structure, making them easy to follow for the audience.

Writing style: The AI-generated article is easy to follow, thanks to short and clear sentences. Economic terms are used sparingly, and the abbreviation “GDP” is spelled out for clarity. The sentences are logically connected, though some phrases lack specific factual details, such as “This affects the service sector” and “This contributes to economic growth”. In contrast, the Inform.kz article uses more complex language and is aimed at professionals in the field of economics, featuring numerous terms and abbreviations without explanations. However, the author sticks strictly to factual information, avoiding unnecessary descriptive phrases.

Factual accuracy: The AI-generated article presented one fact: a 3.1% growth in Kazakhstan’s GDP for the first 11 months of 2022. However, according to official government data, the actual growth was 2.7%. Therefore, the AI’s information was inaccurate, violating the journalistic ethical principle of accuracy. After a second query, the AI revised the growth figure to 3.4%, which was also incorrect. The Inform.kz article, however, correctly reported the growth rate of 2.7%.

Citation of sources: The AI-generated article did not provide any references to sources in either query. In contrast, the Inform.kz article included references to the press services of the government and Akorda.

Completeness of the information: The AI mentioned agriculture and industry but did not provide specific data. As a result, it is difficult to categorize the article as belonging to the economic genre. On the other hand, the Inform.kz article presented growth percentages for each sector, accompanied by infographics for comparison.

5.1.3. News 3. Law—Domestic Violence

Structure of the article: The AI-generated article proposed the topic “The Case of Kuandyk Bishimbayev: The 2024 Verdict and Public Reaction”. It is split onto two key sections:

• The Trial and Verdict;
• Appeal and Subsequent Events.

This layout provides a concise summary of the case. On the other hand, the Sputnik.kz article, titled “Bishimbayev Case: A Chronology of the Crime”, is split into six sections:

• Jealousy as the Cause;
• Forced to Appear Naked;
• Torture at the Restaurant;
• The Murder Case;
• What Bishimbayev Says;
• I Admit My Guilt.

While the Sputnik.kz article provides a detailed account of the case, it can be challenging at times to grasp the main point.

Writing style: The AI-generated article used short, clear sentences, with the text being well-organized and logically presented. It is devoid of emotional tone. In contrast, the Sputnik.kz article maintains an emotional undertone throughout, with journalists compromising objectivity by using language that reflects hostility toward the defendant. It is essential to highlight that maintaining objectivity is a core ethical principle in journalism. Additionally, the Sputnik.kz article blends literary and journalistic styles.

Factual accuracy: The AI-generated article provided accurate details regarding the case of Kuandyk Bishimbayev. It included specific dates, such as the start of the trial (27 March 2024), the verdict announcement (13 May 2024), and the appellate court’s decision and subsequent legal actions. The article mentioned the following key details:

• The charges filed (Article 110, Part 2—“Torture”, Article 99, Part 2—“Murder with special cruelty”).
• The decision of the Interdistrict Court of Astana on 26 September 2024, ordering Bishimbayev to pay KZT 10 million in moral damages and KZT 6.88 million in court expenses to Aytbek Amangeldi.

In contrast, the Sputnik.kz article lacks full details of the legal proceedings. It delves more into the personal conflicts between Bishimbayev and his wife, but such a focus may lead to a subjective and emotionally charged portrayal. Sputnik.kz does not provide information about the trial’s start date or the verdict.

Citation of sources: The AI-generated article cited three sources: Tengrinews.kz, Massaget.kz, and Wikipedia.org. Upon verification, the information was found to be accurate. The Sputnik.kz article, on the other hand, relied on live coverage of the trial from Tengrinews.kz, with the sources properly credited.

Completeness of the information: Both articles omitted certain details. The AI-generated text did not mention personal aspects of the case, while Sputnik.kz failed to provide the precise timing of the trial proceedings.

5.1.4. News 4. Sports—The Performance of Kazakhstani Athletes at the Paris 2024 Olympics

Structure of the article: The AI-generated article was presented without a title and divided into sections labeled as follows:

• Gold Medal;
• Silver Medal;
• Bronze Medal.

This format allows readers to quickly access information about the athletes’ achievements. In contrast, the Zhas Alash newspaper (“Paris-2024: 43rd Place Reflecting Kazakhstan’s Sports Results”) article is more analytical in nature. It consists of three main sections:

• Kazakhstan: 7 Medals—Is it Too Few or Too Many?
• What Did We Fall Behind In?
• What Did We Surpass?

The second section breaks down the performance by sport, covering categories such as judo, boxing, and track and field. While the introduction briefly summarizes the athletes’ accomplishments, the article itself provides an in-depth analysis.

Writing style: The AI-generated article is composed of short, clear sentences that are easy to understand and logically connected. The language flows smoothly and is straightforward. In contrast, the Zhas Alash article frequently uses complex and outdated terminology, which can make it more difficult to read. Additionally, the article contains a degree of subjectivity and emotional coloring. The journalist expresses personal opinions about the athletes’ failures, which undermines the article’s objectivity. The AI-generated piece, on the other hand, maintains a neutral tone throughout, avoiding such biases.

Factual accuracy: The AI-generated article stated that Kazakhstan’s athletes secured 80 spots across 25 sports. However, according to the Ministry of Tourism and Sports of the Republic of Kazakhstan, the country received 92 spots, with 80 athletes ultimately competing at the Olympics. This represents a factual inaccuracy in the AI’s report. Similarly, the Zhas Alash article contained a similar error, reporting 79 athletes across 20 sports. Both articles, however, correctly identified the number of medals won.

Citation of sources: The AI article cited four sources: Egemen.kz, Olymp.kz, Stan.kz, and Wikipedia.org. On the other hand, the Zhas Alash article did not provide any sources.

Completeness of the information: The AI article mentioned seven medals but only referenced five athletes, omitting Gusman Kyrgyzbayev (bronze in judo) and Nurbek Oralbay (silver in boxing). Additionally, the AI article did not provide complete information on the 25 sports, focusing solely on the medalists. The article also failed to include Kazakhstan’s ranking for the Paris 2024 Olympics. In contrast, the Zhas Alash article included this missing information.

5.1.5. News 5. Education and Science—University Grants in 2024

Structure of the article: The AI-generated article titled “Distribution of Grants in Higher Education in Kazakhstan in 2024: An Analysis by Fields and Levels” is divided into six sections:

• Bachelor’s Degree;
• Master’s Degree;
• Doctoral Programs;
• Serpin Program;
• Targeted Grants;
• Additional Grants.

In comparison, the article on Minber.kz (2024 Grants: 112,000 Scholarships for Higher Education Institutions”) is organized into three sections:

• Bachelor’s Degree;
• Master’s Degree;
• Doctoral Programs.

The information provided by the AI aligns closely with the material on the Minber.kz website.

Writing style: The AI article is characterized by clear, straightforward sentences, though it often includes descriptive phrases. Similarly, the Minber.kz article is concise and easy to follow.

Factual accuracy: Both articles provide accurate information regarding the total number of grants. However, the AI-generated article presents distorted data when discussing specific fields. For example, it claims that “over 20,000 grants have been allocated to priority fields such as pedagogy, engineering, information technology, agriculture, and veterinary medicine”. However, according to the Ministry of Science and Higher Education of Kazakhstan’s press service, 19,344 grants were allocated to engineering and technical fields and 13,735 to pedagogy. The AI appears to have referenced official data from the government website, where it was stated that the figure of 20,000 refers to the increase in the number of applications compared to 2023. The article on Minber.kz, however, accurately reflects the official data.

Citation of sources: Both articles cite the official government information resource of Kazakhstan. Additionally, the Minber.kz article provides a reference to Testcenter.kz for specific data on the number of grants by specialty.

Completeness of the information: The AI-generated article lacks data on the number of grants allocated to certain specialties.

5.1.6. News 6. Social Issues—Flooding in Kazakhstan

Structure of the article: The AI-generated article is titled “Flooding in Kazakhstan: Causes and Consequences” and is divided into four sections:

• Causes of Flooding:
• Affected Areas and Residents;
• Rescue and Recovery Operations;
• Social Assistance.

In contrast, the article from Tengrinews.kz does not have distinct subsections, but it emphasizes key information by bolding important details.

Writing style: The AI-generated article employs clear and straightforward language, avoiding complex sentences. It is free from unnecessary or descriptive phrases. Similarly, Tengrinews.kz, as a platform for concise news, also utilizes compact, logically structured sentences. The two articles share a similar style, with well-organized and coherent sentences.

Factual accuracy: The information provided by the AI is accurate. It correctly states that the flooding affected 10 regions, and the recovery costs are accurate as well. This time, the AI adhered to journalistic ethics. Similarly, Tengrinews.kz relied solely on verified information. Both articles maintain ethical standards.

Citation of sources: The AI used five sources: Akorda.kz, Egemen.kz, Azattyq.org, Aikyn.kz, and Qazaqstanhalqyna.kz. However, not all cited data were relevant or used accurately. For instance, the reference to Akorda.kz in the context of “10 regions” did not contain the specified information. Nevertheless, the rest of the data provided by the AI is accurate. Tengrinews.kz cited the Ministry of Emergency Situations as a source.

Completeness of the information: The AI article did not include details about the flooded homes, affected individuals, or the funds allocated for recovery. In contrast, Tengrinews.kz provided comprehensive information on these topics.

5.2. What Do Kazakhstani Readers Really Think? Insights from Survey

To provide a comprehensive comparison between the AI-generated articles and those written by traditional journalists, a survey was conducted among a Kazakhstani audience.

The first assessed article was on a political topic (see Figure 1). A proportion of 65.9% of respondents preferred the AI article for its clarity and comprehensibility, while 34.1% favored the chronological structure of the Azattyq article. When evaluating the writing style, 63.8% of respondents found the AI article to be written in a neutral tone, with minimal emotional bias and simple sentence structures, while 36.2% selected the Azattyq article. Regarding factual accuracy, 72.3% of respondents identified inaccuracies in the AI article, preferring the Azattyq version. A proportion of 27.7% confirmed the accuracy of the information presented by the AI. The evaluation of source references showed that 88.3% favored the Azattyq article for citing official sources, while 11.7% selected the AI article. In terms of completeness, 87.2% of respondents noted that the Azattyq article provided all the necessary information, while 12.8% preferred the AI article.

The results of the evaluation of the economic articles mirrored those of the previous assessment (see Figure 2). When it came to text structure, 50% of respondents favored the AI-generated article, while the other 50% preferred Inform.kz. In terms of readability, 87.2% of respondents found the AI article easy to understand, while 17.3% chose Inform.kz for its clarity. Regarding factual accuracy, 90.4% of respondents confirmed the accuracy of the data presented in the Inform.kz article, while 9.6% considered the information in the AI article to be credible. In terms of source citations, Inform.kz received 96.8% of the votes, while the AI article garnered only 9.2%. Finally, for completeness of information, Inform.kz scored 92.5%, while the AI article was selected by just 7.5%.

The results of the evaluation of the article on domestic violence are outlined as follows (see Figure 3). A proportion of 53.2% of respondents preferred the AI-generated article for its clear structure, while 48.8% favored the structure of the Sputnik.kz article. Regarding style, 86.2% of respondents considered the AI article to be written in a neutral tone and well-balanced, while 13.8% preferred the style of the journalists. All respondents confirmed the accuracy of the facts in both articles. When it came to source citations, the results were nearly equal: 51% favored the AI article, while 49% chose the Sputnik.kz article. In terms of completeness of information, the AI article received 75.5%, while the Sputnik.kz article earned 24.5%.

When evaluating the structure of the fourth news topic, 47.9% of respondents preferred the AI-generated article, while 52.1% chose the Zhas Alash article (Figure 4). In terms of writing style, the AI article received 79.8% of the votes, with 20.2% of respondents finding the style of the journalists to be more effective. Regarding data accuracy, 53.2% of respondents selected the AI article, while 46.8% believed the Zhas Alash article provided more accurate information. In terms of cited sources, the AI article was favored, with 87.2% of participants choosing it over the Zhas Alash article, which received 12.8% of the votes. Finally, for the completeness of information, the Zhas Alash article was rated more comprehensive by 88.3% of respondents, while 11.7% considered the AI article to be sufficiently informative.

The evaluation of articles focused on education and science yielded the following results (Figure 5). In terms of structure, the AI-generated article received 52.1% of the votes, while the Minber.kz article garnered 47.9%. All respondents found the style of both articles to be clear and understandable. When assessing factual accuracy, the journalists’ article was preferred by 73.4% of respondents, compared to 26.6% for the AI article. Regarding the sources cited, both articles received full approval, with 100% of respondents indicating they included the necessary references. In terms of information completeness, 92.6% of participants rated the journalists’ article as more comprehensive, while 7.4% considered the AI article to be sufficiently informative.

The evaluation of the sixth article on a social topic closely mirrored the assessment of the fifth article (Figure 6). In terms of structure, 86.2% of respondents preferred the AI-generated article, while 13.8% chose the Tengrinews.kz article. All participants found the style and accuracy of the data in both articles to be clear and correct. Regarding the sources cited, 50% of respondents favored the AI article, while the other 50% preferred the Tengrinews.kz article. When assessing the completeness of information, 81.9% of respondents considered the journalists’ article to be more detailed, while 18.1% found the AI article to be more informative.

The evaluation results revealed that AI-generated articles generally exhibit a clear structure and neutrality. However, AI occasionally incorporates inaccurate data, which violates journalistic ethics. These errors are more commonly found in articles on political and economic topics. Articles written by journalists received high marks for their completeness of information, factual accuracy, and transparency of sources. Nevertheless, journalists sometimes introduce emotional tones and subjective opinions, particularly in political, legal, and social subjects.

6. Discussion

This study aimed to explore the differences between AI-generated and journalist-written articles in terms of journalistic quality and to examine how audiences perceive these differences. The discussion is organized around the two research questions posed in the Introduction.

• RQ 1: In what ways do AI-generated articles differ from those written by professional journalists regarding structure, writing style, factual accuracy, citation of sources, and completeness of the information?

The results of the comparative analysis confirm that AI-generated articles generally demonstrate a strong structure and a neutral tone. This finding supports the initial hypothesis that AI excels in content organization and stylistic clarity. Across all six topics, AI articles maintained a logical structure with clear headings and straightforward language, making them easy to read and follow. These findings align with Clerwall’s (2014) study, which found that readers often perceive AI-generated content as objective and readable, although, at times, overly simplistic or descriptive.

However, it is important to clarify that the term objectivity in journalism refers not to an inherent quality of content but to a professional method applied by human journalists. As Ward (2011) explains, journalistic objectivity is a pragmatic approach aimed at minimizing bias through disciplined methodology. Similarly, Kovach and Rosenstiel (2001) refer to objectivity as a method of verification and a disciplined approach to gathering and presenting information. Therefore, while AI-generated content may appear neutral in tone, it does not meet the methodological criteria of journalistic objectivity.

According to Kreps et al. (2020), AI-generated texts can be indistinguishable from human-written content. We partially agree with this viewpoint, as AI-generated and journalist-written articles were stylistically similar on certain topics. Some research suggests that AI-generated content can be comparable to human-written news in terms of credibility and quality (La-Rosa Barrolleta & Sandoval-Martín, 2024).

The study found that AI tends to distort facts when writing articles. In the survey, respondents rated journalists’ articles higher in terms of factual accuracy. In contrast to journalist-written articles, which consistently cited government data, press services, and official documents, AI articles frequently included general references or failed to cite any sources at all. This finding aligns with other research. According to Dhiman (2023) and Gutiérrez-Caneda et al. (2023), AI tools can assist with various tasks but may produce inaccuracies or “hallucinations”—generating plausible but factually incorrect information.

This lack of clear source citation in AI-generated texts directly undermines their credibility, as readers are unable to verify where the data originated. As AI-generated content becomes more common, some media organizations have begun requiring that any AI-assisted text include links to the sources from which information was derived (Kharchenko, 2023). Such source transparency is increasingly viewed as essential for maintaining trustworthiness in AI-powered journalism.

Moreover, AI-generated texts were often incomplete, omitting key contextual information, emotional nuances, or critical social details—especially in complex legal or political stories. While AI can efficiently produce news articles on topics like finance and sports (Ayapova, 2021), it struggles with more complex subjects requiring human insight and creativity (Aydın & İnce, 2024; Nandini et al., 2024).

• RQ 2: How are AI-generated texts and those written by professional journalists perceived by the Kazakhstani audience?

Survey results further illustrate how audiences respond to the two types of content. While AI articles were praised for their clarity, simplicity, and structured format—especially on law, education, and social topics—they were consistently rated lower in terms of trust, credibility, and factual reliability, particularly on complex or sensitive subjects such as politics and economics.

For instance, on the political topic, 62.9% of respondents preferred the clarity of the AI article, yet 88.3% trusted the journalist-written version more due to its verified sources and detailed information. A similar pattern was observed on the economic and legal topics. This duality echoes Henestrosa and Kimmerle (2024), who found that audiences often trust AI content stylistically but prefer human-authored texts when it comes to credibility and depth.

In several cases, the audience responses appeared to align closely with the factual and structural shortcomings identified in the AI-generated articles. For instance, in the political topic, the AI version incorrectly stated that the demonstrations began on December 31, while the verified timeline indicates they started on January 2. Additionally, the AI used the outdated name Nur-Sultan instead of Astana, despite the city’s renaming occurring in September 2022. Similarly, on the economic topic, the AI article lacked key contextual details about inflation indicators and did not cite any verifiable sources. This mistake was noted during the content analysis and likely influenced respondents’ trust levels.

Interestingly, on the law topic, respondents appreciated the neutrality of the AI article, with 86.2% acknowledging its balanced tone, but still expected more comprehensive details. On the legal topic, the AI-generated content omitted crucial information about the legal consequences of a court ruling, focusing instead on a procedural summary. This incompleteness corresponded with lower audience ratings for credibility, despite high ratings for neutrality. On the social topic, while respondents appreciated the structured format of the AI-generated article (86.2%), a significant majority (81.9%) rated the journalist-written piece as more complete. Similarly, on the sports and education topics, AI articles were found to be readable but incomplete, lacking in source depth or nuance (Szabo, 2023). These audience perceptions were consistent with the comparative content analysis, which also found the AI-generated articles on these topics to be structurally strong but lacking in comprehensive detail.

While Aydın and İnce (2024) found that AI-generated news lacked adherence to journalistic standards, other research suggests similarities between the two. La-Rosa Barrolleta and Sandoval-Martín (2024) reported no significant differences in credibility between AI- and human-written articles, though some biases in authorship perception were noted.

While AI enhances efficiency and personalization in news production (Verma, 2024), it also facilitates the spread of misinformation and fake news, eroding public trust in media (Uthman, 2024; Forja-Pena et al., 2024).

New Understandings and Implications

This study contributes new insights into how AI functions as a content creator within the Kazakhstani media context. It shows that while AI tools can generate readable, well-organized news, they are limited by factual inaccuracies and a lack of ethical safeguards. The results suggest that AI-generated journalism must be closely supervised by human editors to meet professional standards.

Moreover, the study reveals topic sensitivity: AI performs better on neutral subjects (education, sports) but less reliably on emotionally or politically charged topics. This insight is particularly relevant for media professionals considering the integration of AI into newsroom workflows.

7. Conclusions

The study identified key differences between AI-generated content and traditional articles written by journalists. While AI demonstrated a high level of structure and neutrality, it occasionally distorted facts. Among the Kazakhstani audience, journalists’ articles were highly rated for factual accuracy, completeness of information, and transparency of sources. However, subjectivity and emotional tones were also noted in journalistic materials.

These findings directly address the research questions posed in the study: RQ1 regarding content differences was explored through detailed comparative analysis, while RQ2 on audience perception was addressed via the survey results, revealing a consistent preference for journalist-written articles in terms of accuracy and completeness.

Importantly, the comparative findings between content analysis and the audience survey were aligned: articles that contained factual inaccuracies or lacked detailed sourcing were consistently rated lower in credibility and completeness by respondents. This reinforces the need for improved accuracy and source transparency in AI-generated content.

AI presents new opportunities for automating information gathering and news creation. However, for it to serve as a valuable tool in journalism, it is essential to enhance its ability in deep analysis and improve its fact-checking processes. The future of journalism lies in the synergy of these two approaches.

AI can serve as an additional tool for journalists, accelerating processes and handling large volumes of information while maintaining standards of neutrality, transparency, and ethics.

Despite AI’s structural clarity and seemingly balanced presentation, articles produced by professional journalists continue to have the edge in terms of accuracy, comprehensiveness, and adherence to journalistic ethics. This underscores the continued importance of human involvement in ensuring the quality of media.

However, this study has several limitations. First, it analyzed a limited number of articles across only six topics, which may not fully capture the diversity of AI and human-authored journalism. Second, the study focused exclusively on the Kazakhstani context and the Kazakh language, which may limit generalizability to other media environments. Lastly, while survey responses offer useful audience insights, the sample size and demographic characteristics may have influenced the findings.

Future research could expand on this study by exploring AI-generated journalism in other languages and cultural contexts, as well as examining collaboration models between journalists and AI systems. Additional research could also investigate the impact of disclosure (e.g., labeling content as AI-written) on audience trust and explore how journalists perceive the use of AI in their own professional practice.