This study seeks to analyze the impact of generative artificial intelligence tools, specifically ChatGPT and Midjourney, on the ecosystem of journalistic and creative content production in the Arab and international media environment. The study adopted a Mixed Methods approach combining quantitative analysis through an electronic questionnaire distributed to a sample of 412 journalists, editors, and digital content creators across six Arab countries, and qualitative analysis through semi-structured in-depth interviews with 28 experts in the fields of digital media, artificial intelligence, and media ethics. Theoretically, the study drew upon Marshall McLuhan’s Technological Determinism theory, Everett Rogers’ Diffusion of Innovations theory, and the Gatekeeping Theory in its contemporary digital formulation.
The results revealed that 78.4% of respondents use at least one generative AI tool across various stages of content production, and that 63.1% reported a noticeable improvement in production speed, while 54.6% expressed substantive concerns regarding accuracy, credibility, and originality. The study uncovered a statistically significant correlation between the journalist’s level of digital competence and the degree of generative AI tool adoption (r = 0.71, p < 0.001). The findings also highlighted five principal patterns of tool utilization, ranging from advanced integrative use to complete rejection. The study concluded that comprehensive regulatory and ethical frameworks governing the use of generative AI in media practice must be developed, emphasizing that these tools are reshaping the journalist’s role from “content producer” to “coordinator, verifier, and auditor of automatically generated content.”
The international media environment witnessed profound structural transformations during the second decade of the twenty-first century that reshaped concepts of media production, distribution, and consumption. However, the emergence of the new generation of Generative Artificial Intelligence tools in late 2022 and early 2023 produced what can be described as an “epistemological earthquake” in the media and digital content industry. OpenAI launched ChatGPT in November 2022, achieving the fastest growth rate in the history of digital applications by surpassing the one hundred million active users threshold within just two months of its launch (Hu, 2023). Concurrently, Midjourney, based on Diffusion Models, enabled the generation of high-quality images from simple textual inputs, sparking extensive debates about the future of visual creativity and journalistic design (Borji, 2023).
The intersection of generative artificial intelligence and media practice does not merely represent a marginal technical addition; rather, it constitutes a structural Paradigm Shift in the nature of the media process itself. These tools are capable of producing news texts, summarizing reports, generating compelling headlines, creating accompanying visual content, and even reformulating content in multiple languages—all of which raises fundamental questions about the identity of the media producer, the boundaries of originality and creativity, and the standards of credibility and verification in the age of automatically generated content (Pavlik, 2023; Túñez-López et al., 2024).
In the Arab context, these transformations acquire additional dimensions linked to the linguistic and cultural specificities of Arabic content, as Large Language Models (LLMs) face structural challenges in processing the Arabic language with sufficient accuracy owing to its morphological and syntactic complexities, in addition to the limited availability of high-quality Arabic training data compared to English (Abdelali et al., 2024). Furthermore, these tools raise profound issues related to Cultural and Informational Bias in automatically generated content and its impact on narrative diversity and pluralism in Arab media discourse.
Despite the rapidly increasing adoption of generative AI tools by media organizations, the academic literature still suffers from a clear knowledge gap in understanding the precise mechanisms through which these tools affect the journalistic and creative content production chain, particularly in the Arab context. Most existing studies either focused on the purely technical aspects of these tools without linking them to actual media practice (Cao et al., 2023), or were confined to theoretical explorations that were not empirically tested (Diakopoulos, 2023). Moreover, studies that specifically addressed the Arab media environment remain limited in number and restricted in geographic and methodological scope.
The research problem lies in the absence of an integrated analytical framework that connects the technical, professional, and ethical dimensions of employing generative AI tools in media production; the insufficiency of empirical evidence regarding actual usage patterns of these tools by Arab media practitioners; and the limited scientific understanding of the reciprocal effects between these tools and editorial decision-making processes and media gatekeeping mechanisms.
Identify and classify patterns of generative AI tool employment across the various stages of journalistic and creative production.
Analyze the impact of these tools on media content quality and production efficiency from the perspective of media practitioners.
Explore the ethical and professional challenges associated with integrating generative AI tools into media practice.
Examine the relationship between journalists’ demographic and professional variables and their degree of adoption and usage patterns of these tools.
Develop an integrative conceptual framework that explains the dynamics of interaction between generative AI and contemporary media practice.
Main Research Question (RQ): How do generative AI tools (ChatGPT, Midjourney) affect the ecosystem of journalistic and creative content production?
RQ1: What are the usage patterns of journalists and digital content creators with generative AI tools across the various stages of media production?
RQ2: What is the impact of generative AI tools on media content quality indicators (accuracy, originality, depth, diversity)?
RQ3: What are the most prominent ethical and professional challenges in employing these tools from the perspective of media practitioners and experts?
RQ4: Are there statistically significant differences in the patterns of generative AI tool adoption attributable to demographic and professional variables (age, experience, institution type, digital competence)?
RQ5: How do generative AI tools reshape the role of the media gatekeeper and editorial decision-making mechanisms?
H1: There is a positive and statistically significant correlation between journalists’ level of digital competence and their degree of generative AI tool adoption.
H2: There are statistically significant differences in the degree of generative AI tool adoption attributable to the age group variable.
H3: There is a negative and statistically significant correlation between years of professional experience and the degree of reliance on generative AI tools in media production.
H4: There is a statistically significant effect of the type of media institution (digital-native / digitally transformed legacy) on patterns of generative AI tool employment.
This study is delimited by the following boundaries: The topical boundaries focus on two principal tools—ChatGPT (text generation) and Midjourney (image generation)—with reference to other tools as appropriate. The human boundaries encompass journalists, editors, and digital content creators in six Arab countries: Egypt, the Kingdom of Saudi Arabia, the United Arab Emirates, Jordan, Lebanon, and Morocco. The temporal boundaries extend from March 2024 to January 2025.
Marshall McLuhan (1964), in his foundational work
In the context of generative AI, this theory acquires new dimensions. Tools such as ChatGPT do not merely alter the method by which content is produced; they redefine the essence of “content” and the criteria for its evaluation. When a news text becomes the product of interaction between a human input and an algorithmic output, the relationship between producer, text, and recipient undergoes a transformation (Broussard, 2023). This study adopts a critically modified version (Soft Technological Determinism, Winner, 1986), acknowledging human agency to adapt and redirect technology.
Everett Rogers (1962/2003) presented a framework for understanding how new ideas and technologies spread within social systems over time. Adopters are classified into Innovators, Early Adopters, Early Majority, Late Majority, and Laggards. Adoption is influenced by perceived Relative Advantage, Compatibility, Complexity, Trialability, and Observability. In the Arab media context, these are further influenced by digital infrastructure, institutional policies, professional culture, and the level of technical qualification of journalistic cadres (Weaver et al., 2023).
Gatekeeping theory, originating with Lewin (1947) and developed by White (1950), and later Shoemaker and Vos (2009), has evolved to include algorithmic dynamics (Bro & Wallberg, 2014; Wallace, 2018). With generative AI, an “Algorithmic Gatekeeper” emerges, making implicit inclusion/exclusion decisions based on training data and parameters, necessitating re-theorization of gatekeeping in the age of generative AI (Napoli, 2023; Diakopoulos, 2023). This study proposes “Hybrid Gatekeeping,” shaped by the interaction of the human journalist, the generative algorithm, and institutional policies.
Generative AI refers to systems capable of producing new content—text, images, audio, video—based on learned patterns (Goodfellow et al., 2014). Text-based tools like ChatGPT rely on the Transformer architecture (Vaswani et al., 2017). Diffusion Models (Ho et al., 2020; Rombach et al., 2022) advanced image generation (e.g., Midjourney), sparking debates on intellectual property and originality (Epstein et al., 2023).
AI use predates generative tools via “Automated Journalism” (Wordsmith, Quill) for formulaic reports (Graefe, 2016). With ChatGPT, discourse shifted to “comprehensive generation,” including analytical and creative content. Studies (Nishal & Textile, 2024; Latar, 2023; Wölker & Powell, 2024) show AI as an assistant, with human oversight critical.
Generative AI enhances efficiency, reduces errors, diversifies styles (Simon, 2024), but poses risks of hallucination (Ji et al., 2023; Huang et al., 2023) and bias (Zhou et al., 2023). Kreps et al. (2022) found readers often cannot distinguish AI vs human texts, raising transparency concerns.
Key ethical axes: Attribution/IP (Samuelson, 2023), Transparency/Disclosure (Montal & Reich, 2017), Bias/Fairness (Bender et al., 2021; Weidinger et al., 2022). Regulatory examples: AP guidelines (2023), EU AI Act (2024). Arab frameworks are nascent; UAE AI Strategy (2023) is noted.
Al-Shamri & Al-Otaibi (2024) found cautiously positive attitudes in Saudi Arabia with accuracy and cultural concerns. Hassan & Mohammed (2024) reported a gap in Egypt between awareness and use. Benabdallah (2024) in Morocco found linguistic barriers primary. This study distinguishes itself via cross-national scope (six countries), mixed-methods design, focus on two representative tools, and integrative theoretical framing.
The study adopted a Sequential Explanatory Mixed Methods Design (Creswell & Plano Clark, 2018): quantitative phase followed by qualitative phase.
Population: journalists, editors, digital content creators in six countries (Egypt, KSA, UAE, Jordan, Lebanon, Morocco). Multi-stage stratified sampling by organization type (digital-native, digitally transformed legacy, independent platforms). Final analyzable sample: 412 (response rate 68.7%). Gender: 56.3% male, 43.7% female. Age: 25–34 (38.1%), 35–44 (31.3%), 45–54 (18.9%), 55+ (11.7%). Experience: <5y (22.8%), 5–10 (33.5%), 11–20 (28.4%), >20 (15.3%). Institution type: digital-native (39.6%), digitally transformed legacy (42.2%), independent platforms (18.2%).
Structured electronic questionnaire (Qualtrics), five sections: demographics (12 items); usage patterns (18 items, 5-point Likert); perceived impact on quality (15 items); ethical/professional challenges (14 items); digital competence (10 items, adapted from DigComp 2.2). Content validity by 7 experts; pilot n=45. Cronbach’s Alpha overall 0.91 (subscales 0.84–0.93). CFA (AMOS v28): χ²/df = 2.14, CFI = 0.94, TLI = 0.92, RMSEA = 0.05, SRMR = 0.04.
SPSS v29 and AMOS v28. Descriptive statistics; Pearson correlations; One-Way ANOVA with Scheffé; Stepwise Multiple Regression; K-means Cluster Analysis. Significance α = 0.05.
Semi-structured interviews with 28 experts: 10 senior journalists/editors; 8 AI/media technology specialists; 5 academics; 5 policy/ethics officials. Diverse in geography, profession, gender.
Guide: 15 open questions + probes on impact perceptions, experiences, ethical challenges, future human-machine relations. Conducted via Zoom, May–July 2024, 45–90 minutes, recorded with consent.
Thematic Analysis (Braun & Clarke, 2006/2021): familiarization, coding, theme generation, review, definition, reporting. NVivo 14 used. Trustworthiness (Lincoln & Guba, 1985): credibility (triangulation, peer debriefing), transferability (thick description), dependability (audit trail), confirmability (reflexivity).
78.4% (n=323) use at least one tool; 21.6% do not. Tool usage among users: ChatGPT 89.2%; Google Gemini 41.5%; Midjourney 34.7%; DALL-E 28.3%; Claude 19.6%.
Stage means (Likert 1–5): brainstorming 3.87 (SD 0.92); research 3.71 (0.88); draft writing 3.54 (1.03); editing/rewriting 3.41 (0.95); headlines/summaries 3.38 (1.01); translation/multilingual 3.22 (1.12); visual content 2.94 (1.18); final publication without modification 1.86 (0.94). Highest usage in early stages; lowest in final publication.
Frequency: daily 23.2%; several times/week 35.6%; once/week 24.8%; monthly or less 16.4%.
Positive perceptions: speed 63.1%; linguistic/stylistic diversity 57.3%; structural organization 48.9%; overall productivity 44.2%.
Concerns: accuracy/credibility 54.6%; originality/innovation 51.8%; bias 47.3%; loss of distinctive voice 39.5%.
Composite dimensions (mean, SD): speed/efficiency 3.94 (0.87); linguistic/stylistic diversity 3.47 (0.94); organization/coherence 3.31 (0.99); accuracy/credibility 2.68 (1.08); originality/depth 2.43 (1.14). Pattern: efficiency gains, deep-quality concerns.
Severity means (1–5): misinformation risk 4.21 (0.83); source transparency 4.08 (0.91); IP violations 3.96 (0.97); cultural/informational bias 3.89 (0.94); erosion of skills 3.74 (1.02); employment threats 3.67 (1.11); lack of regulatory frameworks 3.58 (1.06).
H1: Digital competence vs adoption: r = 0.71, p < 0.001 (supported).
H2: Age group differences: ANOVA F(3, 408) = 14.73, p < 0.001, η² = 0.098. Means: 25–34 (3.89), 35–44 (3.52), 45–54 (3.11), 55+ (2.67). Scheffé: significant between 25–34 and 45–54, 55+ (supported).
H3: Experience vs reliance: r = -0.43, p < 0.001; partial (controlling digital competence) r = -0.19, p < 0.01 (partially supported).
H4: Institution type: ANOVA F(2, 409) = 11.28, p < 0.001, η² = 0.052. Means: digital-native 3.78; independent platforms 3.51; digitally transformed legacy 3.12 (supported).
Stepwise regression: R² = 0.583, F(4, 407) = 142.31, p < 0.001. Predictors (β): digital competence 0.44 (p < 0.001); age group -0.21 (p < 0.001); institution type 0.17 (p < 0.01); attitude toward technology 0.14 (p < 0.01).
K-means yielded five patterns: Advanced Adopters (14.3%); Selective Integrators (28.6%); Cautious Experimenters (22.1%); Passive Resisters (13.4%); Categorical Rejectors (21.6%).
The high usage rate indicates generative AI has entered mainstream professional practice in Arab media, but adoption is cautious and stage-specific (“mediated adoption”). This aligns with diffusion theory transitional phases.
Efficiency gains coexist with deep-quality risks, reflecting soft technological determinism: tool affordances shape outputs, but human agency can mitigate via verification and enrichment.
Digital competence is the strongest determinant of adoption, reframing perceived complexity as relational to user skill; narrowing competence gaps can reduce age/experience disparities.
Hybrid Gatekeeping emerges: algorithmic, human, and institutional layers interact; the opacity of the algorithmic layer introduces hidden gatekeeping risks (Diakopoulos, 2023).
Cultural bias is amplified in Arabic contexts due to training data scarcity and Western data dominance, raising risks of “digital colonialism.”
The “Triadic Interaction Model of Generative Media” (TIMGM) comprises: (1) Human Agent (competencies, judgment, ethics); (2) Algorithmic Agent (capabilities, biases, limits); (3) Institutional/Cultural Context (policies, professional culture, legal frameworks). Content quality from AI-assisted production is an interaction product of these components; deficits in any component degrade quality and reliability.
Conceptual framework diagram illustrating the interaction of Human Agent, Algorithmic Agent, and Institutional/Cultural Context.
Extends gatekeeping theory via “Hybrid Gatekeeping” incorporating the algorithmic agent.
Presents TIMGM as an integrative analytical framework for generative media dynamics.
Enriches diffusion of innovations theory with cross-national Arab empirical data, highlighting digital competence.
Surfaces embedded cultural bias as a structural issue in non-Western generative AI use.
Partial reliance on non-probability sampling due to frame access limits; self-report measures subject to social desirability and estimation biases; rapid evolution of AI tools may affect temporal validity; generalizability beyond studied contexts is constrained.
Generative AI tools are now part of Arab media practice, with cautious, selective adoption. They deliver efficiency and diversity gains but challenge accuracy, originality, bias, and transparency. They are reshaping the journalist’s role toward coordination, verification, and auditing of AI outputs, necessitating new competencies (prompt engineering, algorithmic literacy) and reconsideration of authorship, gatekeeping, and responsibility.