Georgian Scientists/ . 7 N 4, 2025 409 Georgian Scientists Vol. 7 Issue 4, 2025 https://doi.org/10.52340/gs.2025.07.04.21 Invisible Editors: Impact of AI on Media Content Quality and Trust Arakhamia Mariam1, Mariam Gersamia2 1Researcher (lead author) 2Professor, Tbilisi State University, https://orcid.org/0000-0001- 9092-9232 Abstract: AI technologies such as machine learning, natural language processing, and automated journalism are rapidly transforming how media content is produced, distributed, and consumed. These tools promise greater efficiency, for example, automated news writing and personalized content recommendations, and enable real-time delivery of information. At the same time, their integration into media workflows has intensified concerns about the circulation of biased, low-quality, irrelevant information and disinformation. Because AI systems learn from existing data, they can reproduce and even amplify the biases embedded in that data, while accelerating the spread of disinformation and undermining trust in media institutions. This study examines how AI becomes an invisible gatekeeper, contributes to biased and irrelevant media content and explores the consequences for public trust and democratic discourse. Using a mixed-methods design, including: content analysis of AI-generated and AI-curated media, an online survey, and secondary data, the research shows that media practitioners and audiences recognize both the transformative potential of AI and its ethical risks. A strong majority of survey participants perceive AI’s impact on media as significant or very significant and associate AI algorithms with the spread of biased or low-quality information across news and social platforms. Respondents also express substantial concern about the opacity and limited accountability of AI systems in shaping what information people see. The findings point to an urgent need for strategies that reduce AI-induced bias and improve information quality, such as enhancing algorithmic transparency, diversifying training data, and developing clear regulatory and ethical frameworks. Drawing on media and communication theories, this article offers a critical analysis of AI’s role in contemporary media and outlines pathways for more responsible and accountable use of AI in the information ecosystem. Keywords: artificial intelligence, media bias, disinformation, AI ethics, algorithmic transparency, Gatekeeping Georgian Scientists/ . 7 N 4, 2025 410 Introduction In the digital era, artificial intelligence (AI) has become a central driver of change in the media ecosystem and industry. Newsrooms and online platforms increasingly rely on AI- driven tools for content creation, distribution, and audience analytics. Automated reporting systems generate routine stories in seconds, recommendation engines curate personalized news feeds, and moderation algorithms filter vast volumes of user-generated content. These developments can increase efficiency and engagement, allowing media organizations to operate at unprecedented speed and scale. However, alongside these benefits, the use of AI in media has raised serious questions about the quality, fairness, and reliability of information. Scholars and practitioners increasingly describe AI’s role in media as double-edged. On one side, AI can support more accurate reporting and tailor content to diverse audience needs; on the other, it can create “filter bubbles” and echo chambers that reinforce existing beliefs and limit exposure to alternative viewpoints (Allcott & Gentzkow, 2017). At the extreme, generative AI systems can produce deepfakes and other synthetic media that blur the boundary between fact and fabrication (Chesney & Citron, 2018). During major political events and public health crises, AI-driven recommendation and amplification systems have been implicated in spreading misleading narratives and conspiracy theories at scale (Allcott & Gentzkow, 2017; Howard & Kollanyi, 2016). A growing body of evidence shows that AI algorithms can inadvertently reproduce and amplify the biases present in their training data, while optimizing for engagement in ways that favor sensational, misleading, or polarizing content. In a context where public debates, elections, and collective decision-making are strongly shaped by algorithmic news feeds and recommendation systems, such biases and inaccuracies pose significant risks for media credibility and informed citizenship (Howard & Kollanyi, 2016). The purpose of this pilot study is to explore the ways in which AI influences the production and circulation of biased and irrelevant information in media, and to consider the broader implications for media credibility and public trust. The research is guided by four core questions: RQ1: How does AI contribute to the spread of biased information in media outlets? RQ2: Through what mechanisms do AI technologies introduce or amplify biases in media content? RQ3: What are the implications of AI-driven information dissemination for media credibility and public trust? RQ4: Which strategies are most promising for mitigating the negative impacts of AI on information quality and reliability? Addressing these questions is both theoretically and practically important. Theoretically, the study contributes to debates on the relationship between algorithmic systems and media effects tradition. Practically, the findings speak to media professionals, policymakers, and technology developers who seek to harness AI’s benefits without further eroding the Georgian Scientists/ . 7 N 4, 2025 411 integrity of public information. The article first reviews the literature on AI in media, bias, and misinformation; then outlines the theoretical framework; describes the mixed-methods design; presents the main empirical findings; and concludes with implications and recommendations for more responsible AI integration in media. Literature Review The application of AI in journalism and media has expanded rapidly. News organizations use AI to generate routine articles, edit video, optimize headlines, recommend stories, and analyze audience behavior. Natural language generation tools can produce financial or sports updates almost instantaneously, while recommendation systems tailor content to individual users based on their previous clicks and preferences. These tools have clear advantages: they reduce the time and cost of news production, support 24/7 coverage, and can help audiences navigate information overload. Yet researchers warn that AI systems frequently encode and reproduce patterns of bias contained in their training data, leading to skewed or unfair outcomes (Obermeyer et al., 2019; Datta et al., 2015). In the media context, this may appear as selective emphasis on particular groups, topics, or political orientations, or as stereotypical portrayals in AI- generated text and images. A recent analysis by MIT Technology Review highlights that widely used language models exhibit systematic political biases in their outputs, reflecting the imbalance and prejudices present in the online text on which they were trained (MIT Technology Review, 2023). If media organizations adopt such models for content creation or filtering, these built-in biases risk being imported into news coverage and editorial decisions. Beyond bias, AI is also deeply implicated in the production and spread of misinformation. Generative AI models can create synthetic text, images, and videos that closely mimic authentic material. Deepfakes, highly realistic AI-generated videos or audio recordings, have already been used to spread false claims about public figures, complicating efforts to verify information (Chesney & Citron, 2018). During the COVID-19 pandemic, automated accounts and AI-driven bots played a role in amplifying conspiracy theories and misleading content on social media (Howard & Kollanyi, 2016). Disinformation can be spread more quickly than fact-checked news on platforms where algorithms optimize for engagement, not accuracy (Allcott & Gentzkow, 2017). The 2016 U.S. presidential election and the Brexit referendum are often cited as early cases where algorithmically amplified disinformation shaped the information environment surrounding critical democratic decisions (Howard & Kollanyi, 2016; Allcott & Gentzkow, 2017). According to Georgian scholars (Gersamia et al., 2025), online media outlets in Georgia use AI capabilities in their daily work, particularly to create audio versions of texts, explain terms and laws, check and refine information, expand and clarify press release topics, and perform translations. Georgian Scientists/ . 7 N 4, 2025 412 At the same time, AI is also used to detect and counter disinformation and misinformation. For instance, machine learning models can flag suspicious news articles, identify deepfakes, and support fact-checking initiatives (Zellers et al., 2019). This dual role - both generating and detecting misleading content - underscores the complexity of AI’s influence on information quality. Journalists in Georgia consider information provided by AI to be unreliable and therefore triple-check the facts. The study highlights that there are no specific protocols for the use of AI in newsrooms; instead, journalists are guided by general rules and their own experience. A major challenge remains the spread of disinformation and misinformation, especially during political crises. Other examples illustrate the risks of deploying AI without robust safeguards. In recruitment, an experimental AI hiring tool developed by Amazon was found to systematically downgrade applications from women, apparently because it had learned from historical hiring patterns that favored men; the project was subsequently abandoned (Reuters, 2018). In policing, predictive algorithms trained on biased crime data have been criticized for reinforcing racial disparities in law enforcement (Lum & Isaac, 2016). In healthcare (Obermeyer et al., 2019) an algorithm used to guide care-management decisions underestimated the needs of Black patients because it used healthcare spending as a proxy for health status, a measure influenced by structural inequalities. These cases share a common mechanism: AI systems “inherit” past biases embedded in their data, and, if left unchecked, institutionalize them. In the media sector, similar dynamics may arise if news recommendation algorithms systematically favor certain topics, perspectives, or demographic representations. As public awareness of AI’s limitations grows, concerns about bias and fairness in media become intertwined with questions of credibility and business sustainability. Further Survey data reflect this growing skepticism. A 2023 Pew Research Center study found that a majority of Americans are more concerned than excited about the increasing use of AI in everyday life, including media, and many expect AI-generated content to reflect the biases of its data and developers (Faverio & Tyson, 2023). For media organizations, perceptions of bias are thus not only an ethical issue but also a reputational and strategic challenge. In response to these concerns, debates on AI governance have intensified. Media and technology scholars advocate for ethical design principles such as fairness, transparency, and human oversight to guide AI development and deployment (Chesney & Citron, 2018; Solaiman et al., 2019). Policymakers began to translate these ideas into legal frameworks. The European Union’s AI Act introduced extensive rules for high-risk AI systems, including transparency requirements and restrictions on certain harmful uses (Council of the EU, 2024). In the United States, the White House Office of Science and Technology Policy has proposed a “Blueprint for an AI Bill of Rights,” outlining principles to protect individuals from discriminatory and opaque automated systems (White House OSTP, 2022). These initiatives reflect a shared recognition that technical solutions alone are insufficient. In sectors such as media, where AI influences public discourse and democratic processes, Georgian Scientists/ . 7 N 4, 2025 413 formal safeguards and accountability mechanisms are needed. The present study builds on this literature, using it to frame the empirical investigation and to interpret the perceptions and experiences of media users and professionals. Theoretical Framework This research draws on three complementary theoretical perspectives: Social Learning Theory, Technological Determinism, and Media Ecology. Together they offer a multi-level lens on AI’s role in shaping media content, audience behavior, and the wider communication environment. Social Learning Theory (Bandura, 1977) posits that individuals acquire attitudes, norms, and behaviors by observing others and modeling what they see. In contemporary societies, media play a central role in this observational learning process. When AI-driven systems decide which news stories or posts appear in a person’s feed, they effectively structure the environment from which that person learns about the world. From this perspective, repeated exposure to AI-curated biased content may normalize particular viewpoints or stereotypes, subtly influencing public opinion and behavior. Conversely, if AI systems are designed to foreground reliable, diverse, and balanced content, they can support more informed and nuanced social learning. Social Learning Theory also highlights that audiences are not passive: awareness that content is AI-generated or AI-curated may shape how people interpret and trust that content. This study therefore considers not only what AI systems disseminate, but also how audiences perceive and respond to AI-mediated information. Technological Determinism views technology as a major driver of social and cultural change (McLuhan, 1964). Each new communication technology - from print to television to the internet - has reshaped how information flows, how institutions function, and how individuals relate to one another. Within this framework, AI can be seen as the latest transformative media technology. Applied to AI in media, Technological Determinism invites us to consider how algorithmic systems redistribute power and reshape journalistic practice. When AI tools influence which issues receive attention, how stories are framed, and even generate content independently, they shift parts of editorial control from human journalists to automated systems. This shift can have profound implications: it may enhance efficiency and personalization, but it can also entrench opaque forms of influence and magnify the reach of biased or misleading content. This study uses Technological Determinism to examine the extent to which AI is becoming an autonomous force in media, shaping agendas and narratives, rather than merely a neutral instrument used by human decision-makers. Georgian Scientists/ . 7 N 4, 2025 414 Media Ecology, associated with scholars such as McLuhan (1964) and Postman (1979), examines how media environments affect human perception, understanding, and social organization. It conceptualizes communication technologies, content, and audiences as parts of an interconnected ecosystem. From a Media Ecology perspective, AI constitutes a new structural element in the media environment that alters the circulation and visibility of information. Personalized recommendation systems mean that different individuals may inhabit qualitatively different informational “worlds,” potentially weakening the shared public sphere traditionally supported by mass media. The speed, volume, and personalization of AI-mediated content contribute to information overload and can encourage people to rely on cognitive shortcuts, such as confirmation bias, in processing news. Gatekeeping theory explains how information passes through a series of “gates” before reaching the public, with various actors deciding what is selected, shaped, or excluded. Originating in Kurt Lewin’s work on decision-making in social systems (1947) and adapted to journalism by David Manning White (1950), the theory initially focused on editors and journalists who chose which events became “news.” Later scholarship expanded the concept to include organizational routines, professional norms, owners, advertisers, and political actors as powerful gatekeepers, all of whom influence which issues are visible, how they are framed, and whose voices are amplified or marginalized. In this view, news is not a neutral reflection of reality but the outcome of continuous selection and filtering processes. More recent work extends gatekeeping theory to the digital environment and develops concepts such as digital. algorithmic and networked gatekeeping, where journalists, platforms, algorithms, and users jointly filter information (Meraz & Papacharissi, 2013; Wallace, 2018; Thorson & Wells, 2015; Shoemaker, 2020; van Dalen, 2023; Almakaty, 2025). Taken together, these theories allow for a layered analysis of AI in media. Social Learning Theory focuses on individual-level effects; Technological Determinism emphasizes the structural force of technology in shaping institutions and practices; and Media Ecology examines systemic changes in the communication environment. This integrated framework informs the research design and guides the interpretation of findings regarding AI-related bias, misinformation, and public trust. The gatekeeping theory helps to conceptualize AI systems, platforms, and algorithms as powerful new “gatekeepers” that shape which information reaches audiences, in what form, and with what prominence. This framework therefore prompts questions central to the present study and discussion: How is AI reconfiguring the conditions under which public discourse takes place? What happens to democratic deliberation when algorithmic curation fragments audiences into micro-publics with divergent information diets? Georgian Scientists/ . 7 N 4, 2025 415 Methodology To address the research questions, the study employed a mixed-methods design that combines qualitative and quantitative approaches. This combination makes it possible to capture both the concrete manifestations of AI-related bias in media content and the broader perceptions and attitudes of media users. On the qualitative side, the research included a qualitative content analysis of AI-generated or AI-curated media. The sample comprised news articles written by AI, social media feeds shaped by recommendation algorithms, and high-profile deepfake videos that had attracted public and journalistic attention. These materials were collected from online platforms and news archives using systematic criteria, such as domain diversity and documented AI involvement. A coding scheme was developed to assess indicators of bias (for example, partisan slant, demographic stereotypes, or exclusion of certain perspectives) and misinformation (such as factual inaccuracies or misleading framing). Two researchers independently coded the material and compared results to ensure inter-coder reliability; discrepancies were resolved through discussion. Selected case studies from existing reports and academic literature were also examined to contextualize the findings and connect them to known incidents of AI- related failures in media and adjacent sectors. On the quantitative side, online survey was conducted to explore how media consumers and professionals perceive AI’s influence on information quality. A total of 52 respondents participated, recruited through purposive sampling to ensure variation in age, gender, educational attainment, and professional background. Roughly one-third of respondents were under 25, about half were aged 26-34, and the remainder were over 35; slightly more women than men took part, and education levels ranged from secondary to doctoral. The questionnaire consisted of both closed and open-ended questions. It measured familiarity with AI technologies in media, perceived significance of AI’s impact on media production and dissemination, views on AI’s contribution to biased and irrelevant information, ethical concerns (e.g. privacy, accountability, and fairness), and support for potential mitigation strategies, including regulation. Many items used Likert-type scales, while open-ended questions provided space for more detailed comments. Descriptive statistics (frequencies and percentages) were used to summarize the quantitative responses, and thematic analysis was applied to the qualitative comments. Finally, secondary data from reputable sources such as the Pew Research Center and the Reuters Institute for the Study of Journalism were reviewed. These larger-scale surveys and reports served as benchmarks, enabling comparison between the present study’s findings and broader trends in public attitudes toward AI and news. The research adhered to established ethical standards. Participation in the survey was voluntary and anonymous; respondents were informed about the purpose of the study and Georgian Scientists/ . 7 N 4, 2025 416 consented to the use of their responses in aggregate form. No personally identifying information was collected. For the content analysis, only publicly available material was used, and care was taken not to inadvertently amplify harmful content. Deepfake videos and misleading posts, for example, were examined in a controlled research context and not redistributed. Secondary data were used in accordance with the terms specified by the original sources and are cited appropriately. To reduce the risk of researcher bias, multiple coders were involved in the qualitative analysis, and the study deliberately considered both the opportunities and the risks associated with AI in media. This reflexive approach supports the reliability and balance of the conclusions. Analysis The survey results indicate broad recognition that AI is reshaping media. When asked about the overall significance of AI’s influence on media production and dissemination, approximately 64% of respondents described it as “very significant” and a further 16% as “extremely significant.” Only a small minority saw AI’s impact as minimal. Respondents frequently referred to AI as a “game-changer,” noting its role in automating routine tasks, enabling continuous content generation, and tailoring news feeds to individual users. However, this recognition of AI’s utility is accompanied by substantial concern. Around 70% of respondents agreed that AI contributes to the spread of biased information in media to a moderate or great extent, and similar proportions associated AI systems with the dissemination of irrelevant or low-quality content. These perceptions echo scholarly arguments that data-driven recommendation engines can prioritize sensational or divisive content and inherit the biases present in their training data (Allcott & Gentzkow, 2017; MIT Technology Review, 2023). Ethical concerns were particularly pronounced. An overwhelming majority of respondents reported being “concerned” or “very concerned” about issues such as transparency and accountability. Several open-ended comments focused on the opacity of recommendation algorithms, respondents often did not know why particular stories appeared in their feeds, and on uncertainty about responsibility when AI-generated or AI-selected content turns out to be false or harmful. The qualitative content analysis provides insight into how these concerns materialize in practice. AI-written news articles tended to follow a neutral tone but revealed subtle patterns of bias in topic selection and examples. For instance, an AI-generated business story highlighted only male CEOs as examples of successful leadership, reflecting the gender imbalance in historical coverage. In another case, an AI-curated feed disproportionately emphasized crime stories associated with specific communities, echoing patterns of representation that have been criticized in discussions of algorithmic bias and racial Georgian Scientists/ . 7 N 4, 2025 417 stereotyping. The content analysis further illustrates how algorithmic selection and generation can privilege particular topics, actors, and frames (for example, repeatedly highlighting certain stereotypical representations, while marginalizing alternative perspectives). In this sense, AI does not merely accelerate existing editorial choices but reconfigures gatekeeping itself by embedding selection criteria into technical systems that optimize for engagement rather than public-interest values. The respondents call for human oversight, transparency, and clearer rules: they are, in effect, demanding more accountable gatekeeping in a media environment where the balance of control has shifted from human editors toward opaque algorithmic procedures. Instances of misinformation were also identified. One case involved an AI-generated sports report that incorrectly announced a team’s victory due to a misinterpretation of a statistical data feed. Deepfake case studies demonstrated how convincingly fabricated videos of public figures could circulate widely before being debunked, complicating efforts by journalists and fact-checkers to maintain informational integrity. These examples illustrate why many survey participants view AI as a source of low-quality or misleading content. AI systems lack contextual judgment and may confidently produce outputs that are factually wrong or inappropriate. Moreover, because their behavior is shaped by past data, existing social and cultural biases are easily reproduced and scaled up. Bringing together the survey and content analysis yields several overarching observations. Algorithmic selection and personalization play a central role in how AI shapes media. The more AI systems optimize for engagement, the greater the risk that they will serve users content that is emotionally charged, partisan, or sensational rather than accurate or diverse. According to the Reuters Institute’s 2024 Digital News Report, the lack of transparency in algorithmic processes undermines trust and audiences across many countries express lower trust in AI-generated news than in content produced and edited solely by humans, especially when the topic is politically sensitive or personally consequential. The same concerns emerge in our study. Respondents do not reject AI outright. 65% of participants agreed that AI-generated content should be clearly labeled, and more than 80% considered human review an essential step before publication. Instead, they call for clearer rules and safeguards.These preferences mirror industry discussions on best practices, which emphasize “human-in-the-loop” approaches and transparency as key to responsible AI use in newsrooms. The survey results, where a majority of respondents associate AI with the spread of biased and low-quality content and express concern about opacity and accountability, suggest that audiences perceive these algorithmic gatekeepers as both powerful and insufficiently controlled. Respondents proposed several measures to reduce AI-related harms. These proposals closely align with ongoing regulatory and ethical debates at national and international levels (Council of the EU, 2024; White House OSTP, 2022). The most frequently mentioned include: Transparency and disclosure, Diverse and representative training data, Human oversight, Bias-aware design, Regulation and standards. In particular: Georgian Scientists/ . 7 N 4, 2025 418 1. Platforms should clearly explain why content is being recommended and label AI- generated or heavily AI-edited material. Simple explanations such as “You are seeing this because…” can help demystify algorithmic decisions. 2. Ensuring that AI systems learn from balanced and inclusive datasets can help reduce systemic bias in outputs. 3. Journalists and editors should remain central in the editorial process, using AI as a tool rather than a substitute. Human review is seen as crucial for checking facts, tone, and fairness. 4. AI systems could incorporate internal checks to detect and flag potentially biased or problematic content before publication. 5. Many respondents supported the development of legal and professional frameworks requiring transparency, fairness audits, and accountability mechanisms for AI used in media. Conclusion AI is now embedded in the everyday operations of media organizations, from automated story generation to algorithmic curation of news feeds. This study has examined how these technologies affect the circulation of biased and irrelevant information and what this means for media credibility and public trust. The findings suggest that AI’s role in media is ambivalent. On the one hand, AI offers powerful tools that can expand coverage, reduce costs, and personalize content. On the other hand, AI systems can reproduce existing biases and introduce new forms of misinformation, especially when optimized for engagement rather than accuracy. Survey respondents recognize both sides of this transformation: they perceive AI as highly influential but are also deeply concerned about opacity, bias, and responsibility. Content analysis and case studies show that these concerns are grounded in real-world examples, where AI has generated inaccurate stories, amplified skewed representations, or contributed to the spread of synthetic media. Through the lens of Social Learning Theory, the study underscores the risk that repeated exposure to AI-curated biased content can normalize distorted views of social and political realities. Technological Determinism highlights AI’s growing influence over editorial decisions and agenda-setting, while Media Ecology draws attention to the broader reshaping of the information environment into fragmented, personalized “micro-publics.” Taken together, these perspectives emphasize that AI does not simply add another tool to the media toolkit; it alters the conditions under which public discourse unfolds. AI does not merely add more content to the media environment but actively structures the flow and visibility of that content. By asking who (or what) makes decisions about selection, exclusion, and emphasis, gatekeeping theory provides a framework for assessing how AI- driven gates may introduce or amplify bias and distortions in the information landscape. Georgian Scientists/ . 7 N 4, 2025 419 Redistribution of gatekeeping power affects trust in the media. In combination with Social Learning Theory, Technological Determinism, and Media Ecology, gatekeeping theory thus helps to connect micro-level experiences of exposure (what appears in individual feeds) with macro-level concerns about media credibility, democratic discourse, and the legitimacy of AI-mediated information systems. The practical implications are clear. Media organizations and technology providers need to invest in bias mitigation, transparency, and human oversight. Diversifying training data, auditing algorithms, labeling AI-generated content, and maintaining robust editorial review processes are concrete steps that can help ensure AI supports rather than undermines journalistic values. Media literacy initiatives that help audiences understand how AI shapes their information environment are equally important. For policymakers, the results support the development of regulatory frameworks that require transparency and accountability for AI systems used in influential information channels. Such frameworks should aim to protect fundamental rights and democratic processes while still allowing for innovation in media practice and technology. This study has limitations: the survey sample is relatively small and not representative of any national population; the content analysis focuses on selected cases rather than a comprehensive dataset. Future research could extend the work by comparing perceptions across countries, conducting large-scale audits of AI-generated versus human-generated news, and experimentally testing the effectiveness of specific mitigation measures. Despite these limitations, the pilot research contributes to ongoing discussions about AI and media by highlighting both the risks and the opportunities associated with AI-driven information systems. Bias and irrelevance in media are longstanding challenges, but AI has changed their scale and speed, searching for effective safeguards more urgent. A balanced approach, one that embraces AI’s potential while insisting on ethical design, transparency, and human responsibility, is essential for preserving the integrity of the information environment and supporting an informed and engaged public. Bibliography AI language models are rife with different political biases. (2023, August 7). MIT Technology Review. Retrieved from https://www.technologyreview.com/2023/08/07/1077324/ai-language-models-are-rife- with-political-biases/ (last seen: 20.11.2025) Almakaty, S. S. (2025). Gatekeeper theory in the digital media age: A comprehensive and critical literature review. Preprints. https://doi.org/10.20944/preprints202510.2484.v1 (last seen: 27.11.2025) Georgian Scientists/ . 7 N 4, 2025 420 Allcott, H., & Gentzkow, M. (2017). Social Media and Fake News in the 2016 Election. Journal of Economic Perspectives, 31(2), 211–236. https://doi.org/10.1257/jep.31.2.211 (last seen: 20.11.2025) Bandura, A. (1977). Social Learning Theory. Englewood Cliffs, NJ: Prentice-Hall. Chesney, R., & Citron, D. (2018). Deepfakes and the New Disinformation War. Foreign Affairs, 98(1), 147–155. Retrieved from https://www.foreignaffairs.com/articles/world/2018-12-11/deepfakes-and-new- disinformation-war (last seen: 20.11.2025) Council of the European Union. (2024, May 21). Artificial Intelligence Act: Council gives final green light to the first worldwide rules on AI [Press release]. Retrieved from https://www.consilium.europa.eu/en/press/press-releases/2024/05/21/artificial- intelligence-ai-act-council-gives-final-green-light-to-the-first-worldwide-rules-on-ai/ (last seen: 20.11.2025) Datta, A., Tschantz, M. C., & Datta, A. (2015). Automated Experiments on Ad Privacy Settings. Proceedings on Privacy Enhancing Technologies, 2015(1), 92–112. https://doi.org/10.1515/popets-2015-0007 (last seen: 20.11.2025) Faverio, M., & Tyson, A. (2023, November 21). What the data says about Americans’ views of artificial intelligence. Pew Research Center. Retrieved from https://www.pewresearch.org/short-reads/2023/11/21/what-the-data-says-about- americans-views-of-artificial-intelligence/ (last seen: 20.11.2025) Gersamia M., Gigauri E., Gersamia M., Media Environment 2025: Navigating During Media Capture in Georgia, Media and Communication Educational and Research Center - “Media Voice”, Media Voice - Europe, 2025. Howard, P. N., & Kollanyi, B. (2016). Bots, #StrongerIn, and #Brexit: Computational Propaganda during the UK–EU Referendum. SSRN Electronic Journal. http://dx.doi.org/10.2139/ssrn.2798311 (last seen: 20.11.2025) Lewin, K. (1947). Frontiers in group dynamics: II. Channels of group life; social planning and action research. Human Relations, 1(2), 143–153. https://doi.org/10.1177/001872674700100201 (last seen: 27.11.2025) McLuhan, M. (1964). Understanding Media: The Extensions of Man. New York: McGraw- Hill. Meraz, S., & Papacharissi, Z. (2013). Networked gatekeeping and networked framing on #Egypt. The International Journal of Press/Politics, 18(2), 138–166. https://doi.org/10.1177/1940161212474472 (last seen: 27.11.2025) Obermeyer, Z., Powers, B., Vogeli, C., & Mullainathan, S. (2019). Dissecting racial bias in an algorithm used to manage the health of populations. Science, 366(6464), 447–453. https://doi.org/10.1126/science.aax2342 (last seen: 20.11.2025) Georgian Scientists/ . 7 N 4, 2025 421 Postman, N. (1979). Teaching as a conserving activity. Delacorte Press. Reuters. (2018, October 10). Insight: Amazon scraps secret AI recruiting tool that showed bias against women. Reuters. Retrieved from https://www.reuters.com/article/worldNews/idUSKCN1MK0AH (last seen: 20.11.2025) Shoemaker, P. J. (2020). Gatekeeping and journalism. In J. Nussbaum (Ed.), Oxford research encyclopedia of communication. Oxford University Press. https://doi.org/10.1093/acrefore/9780190228613.013.819 (last seen: 27.11.2025) Thorson, K., & Wells, C. (2016). Curated flows: A framework for mapping media exposure in the digital age. Communication Theory, 26(3), 309–328. https://doi.org/10.1111/comt.12087 (last seen: 27.11.2025) van Dalen, A. (2023). Algorithmic gatekeeping for professional communicators: Power, trust, and legitimacy. Routledge. https://doi.org/10.4324/9781003375258 (last seen: 27.11.2025) White House Office of Science and Technology Policy (OSTP). (2022). Blueprint for an AI Bill of Rights: Making Automated Systems Work for the American People. Retrieved from https://bidenwhitehouse.archives.gov/ostp/ai-bill-of-rights/ (last seen: 20.11.2025) Wallace, J. (2018). Modelling contemporary gatekeeping: The rise of individuals, algorithms and platforms in digital news dissemination. Digital Journalism, 6(3), 274–293. https://doi.org/10.1080/21670811.2017.1343648 (last seen: 27.11.2025) White, D. M. (1950). The “gate keeper”: A case study in the selection of news. Journalism Quarterly, 27(4), 383–390. https://doi.org/10.1177/107769905002700403 (last seen: 27.11.2025) Zellers, R., Holtzman, A., Rashkin, H., Bisk, Y., Farhadi, A., Roesner, F., & Choi, Y. (2019). Defending against neural fake news. In Advances in Neural Information Processing Systems (Vol. 32). Retrieved from https://papers.nips.cc/paper_files/paper/2019/hash/3e9f0fc9b2f89e043bc6233994dfcf76- Abstract.html (last seen: 20.11.2025)