www.rsisinternational.org
Page 5817
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
The Role of AI in Shaping Political Discourse on Social Media
Mohammad Nurhafiz Hassim
*
, Nur Nasliza Arina Mohamad Nasir
Faculty of Communication and Media Studies, University Technology MARA, Malaysia
*
Corresponding Author
DOI: https://dx.doi.org/10.47772/IJRISS.2025.910000478
Received: 02 November 2025; Accepted: 08 November 2025; Published: 17 November 2025
ABSTRACT
In the contemporary landscape of politics, social media functions as the primary venue of engagement. The
algorithms of artificial intelligence (AI) systems curate political discourse, target advertising, moderate
discussions, and place generative media within social media feeds at marginal costs approaching zero.
However, the engagement-optimized closed systems distort content, create environments for unaccountable
microtargeting, spread misinformation, erode trust in the political system, and undermine the legitimacy of
elections. This research seeks to explain how AI on social media structures exposure, interpretation, and
perceived legitimacy of political discourse, and identifies interventions that retain benefits while reducing
harms. The analysis is based on secondary data situated in Google Scholar, Scopus, and Web of Sciences from
2020 to 2025. This study builds on peer-reviewed research articles as the foundation for assessing the AI-
political communication functions, strategies and AI technology. Six functions of AI systems were articulated:
algorithmic curation and ranking, political ad targeting and delivery optimization, automated moderation and
labeling, cross-cutting design choices, context dependence, and generative AI synthetic and synthetic political
content. In this context, the benefits of AI in political communication include improved relevance,
accessibility, safety, and efficiency of political campaigns, while the harms involve narrowed cross-cutting
exposure, high-arousal polarization, opaque segmented persuasion, uneven enforcement, and synthetic media
that undermines integrity.
Keywords: Artificial Intelligence, Political Discourse, Social Media; Content Generative AI, Secondary Data
INTRODUCTION
Media is a significant factor that impacts the lives of contemporary society. Media, including newspapers,
magazines, television, radio, the Internet, and social media, significantly contribute to the dissemination of
information, entertainment, and education throughout society. Social media is an interactive application of the
new millennium, characterised by engaging graphics, sound, and video features. It is unsurprising that social
media is becoming integrated into society due to its numerous advantages, including providing access to a
knowledge-rich environment and serving as a rapid conduit for information acquisition and dissemination.
Social media streamlines various facets of life, including communication, relationships, information access,
and community engagement (Nur Nasliza Arina & Mohammad Nurhafiz, 2023; Akmar Hayati, 2021).
Social media is a common place where people find current information, cultivate opinions, and find ways to
encourage participation in a cause. At the core of this environment are systems of artificial intelligence (AI)
that as a matter of routine decide (Liu, 2021) user invisibility, the audience of political communications, the
tagging, deletion of posts, and, most recently, the very presence of content itself (Dávila Cortés, 2025; Hartono
et al., 2025; Wang & Kim, 2023). In the real world, recommendation and ranking algorithms construct feeds
(Bartley et al., 2025), political ads are targeted and optimized for delivery through automated systems (Ali et
al., 2021), political text (Sinha et al., 2024), and voice (S. Kumar et al., 2024), as well as image and video (Liv
& Greenbaum, 2020) (Liv & Greenbaum, 2020) content, at a fraction of the cost of a human, instead of hand,
are produced by generative AI systems. It is also important to note that following the phase of training, these
models can be distributed at a zero incremental cost. The combination of low production cost and almost no
marginal cost for additional models of copies for sale enables the inexpensive and wide-ranging circulation of
www.rsisinternational.org
Page 5818
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
persuasive content (Lv, 2023). The ability of these electronic tools to influence the accessibility, framing, and
circulation of political thoughts changes the online political discourse profoundly.
The operational logic of AI systems on digital platforms poses distinct vulnerabilities to democratic processes
across four main areas. In the curation area, AI systems filter and rank information so that political information
environments become distorted, public trust diminishes, and polarization increases, ultimately undermining the
quality of democratic choices (Yazici, 2025). In the targeting area, the use of AI ‘raises ethical concerns about
privacy and the exploitation of psychological vulnerabilities (Kavoliūnaitė-Ragauskienė, 2024; Rubio Núñez,
2025) thus enabling public manipulation to go unchecked. In the area of modification and fabrication, AI
technologies disseminate misinformation at an unprecedented speed, obscure the boundaries of truth and lie,
and undermine the trust essential to democracy (Imran et al., 2025). Furthermore, the production of fake
content that may trigger disinformation flows is unprecedented and is a direct attack on the democratic order
(Cupać & Sienknecht, 2024). In summary, the operational logic of AI systems on digital platforms poses
distinct vulnerabilities to democracy in the areas of Curation, Targeting, Modification, and Fabrication. These
negative functionalities operate most destructively in high stakes, polarized settings such as during elections,
where they fundamentally shift the exposure, tone and trustworthiness of public discourse, thus influencing
electoral participation and undermining the legitimacy of institutions. This clear and present danger highlights
the need to develop and implement strong regulations and ethical standards to address the risks that artificial
intelligence poses to democracy.
This conceptual paper describes how AI applications in social media shape the political content individuals are
exposed to and the conversations surrounding this content, also addressing the advantages and disadvantages
these applications provide to political discussions online. By integrating the context, issues, and aims within
this introduction, the paper clearly articulates a structured, concept-driven exploration of the intersection
between AI and political discourse. It aims to provide students, educators, practitioners, and policymakers with
a rapid understanding of the primary functions of these systems, the implications of design choices, the
strongest and weakest points of available evidence, and the practical steps that reliably improve the quality of
discourse within expressive political alignment.
LITERATURE REVIEW
The term Artificial Intelligence (AI) encompasses a branch of computer science that is concerned with the
development of machines and software that are capable of carrying out functions that entail a degree of human
intellect, including learning, reasoning, problem-solving, perception and decision making (Deng, 2018; Yadav
et al., 2025). Artificial Intelligence (AI) relies on the implementation of machine learning algorithms and
technologies that enable machines to perform tasks autonomously or semi-autonomously with a limited
cognitive component (Morandín-Ahuerma, 2022). One of the ways AI is shaping political discourse is through
AI-powered sentiment analysis which allows political campaigns to gauge public sentiment and adjust their
communicative strategies accordingly. This approach improves the precision with which campaigns focus on
voters and optimize campaign strategies, thereby enhancing the effectiveness of political engagement (Ayma
Quirita et al., 2024; Tembhurne et al., 2025).
The use of AI technologies such as deepfakes and personalized messaging can build custom narratives that can
sway opinion and deepening societal divides. This is particularly the case with populism where AI
technologies are utilized for the creation of neo-propaganda (Jain & Mitra, 2025). The employment of AI to
construct echo chambers and filter bubbles deepens biases and divides the polity and closes the gate on
deliberative democracy (Pinto-Bustamante et al., 2023). AI tools can skew the discourse of democracy and
fracture the discourse of democracy on imperfections. The resultant absences of transparency and the active
use of bias within AI tools in political discourse raises ethics of accountability within democracy (Ghose et al.,
2025). The opacity within algorithms and the exploitation of psychological weaknesses calls for the upholding
of democracy through ethics and regulation in order AI technologies to respond to the democratic demand
(Ghose et al., 2025; Jain & Mitra, 2025; Kashiramka et al., 2024).
Artificial Intelligence has been utilized to shape key political events such as the Brexit decision and the US
elections, through the use sophisticated analytics and tailored disinformation strategies (Long et al., 2020). The
www.rsisinternational.org
Page 5819
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
use of AI to propagate disinformation and fake news has emerged to be a problem, and the case of ChatGPT is
being examined as a potential distributor of false information (Vasist et al., 2025). The impact of AI on
political communication differs from one country to another and is determined by the specific national interests
and geopolitical situation. For example, the US situation is one of military preeminence, whereas the EU is
concerned with ethical governance. Comparative studies demonstrate the very different ways legislative bodies
across the world discuss and frame AI, and this highlights the discrepancies in regulation and ethics (Suter et
al., 2025).
Furthermore, AI is used fundamentally differently in authoritarian states versus democracies, with authoritarian
regimes leveraging technology for control while democracies focus more on service delivery. Authoritarian
states like China use AI primarily for surveillance and political control (Zeng, 2020). They employ facial
recognition, social media monitoring, and big data analytics to strengthen regime power and suppress dissent
(Bukhari & Anwar, 2025). In contrast, democratic governments predominantly use AI to improve public
services, with European Union studies showing AI applications mainly supporting service delivery and internal
management (Van Noordt & Misuraca, 2022). However, both systems face challenges, democracies risk
technological paternalism and polarization, while authoritarian states risk over-reliance on technological
control (Cupać et al., 2024). The key difference lies in governance approach, authoritarian states see AI as a
tool for political consolidation, while democracies view it as a potential mechanism for administrative
improvement.
Political discourse involves the intricate dynamics of verbal and written communication in each political
framework. It goes beyond merely communicating information (political communication); it involves a myriad
of ways through which rhetoric is deployed to enact, preserve, or contest forms of power, to create and
challenge ideologies, and to construct and alter (social) reality (Cherikbaeva et al., 2024). It includes the entire
range of politically oriented communication, in whatever form text, discourse, images, or symbols (Latif,
2017). Perhaps the most characteristic form of political communication is persuasion, primarily using
propaganda, which employs particular grammatical, morphological, and syntactic structures to perform
prescribed communicative functions (Shkvorchenko, 2020). In the field of political science, the most
innovative use of constructive discourse is the study of relational power. It enables political scientists to
theorize how politicians construct and sustain power over their audience (Latif, 2017; Sravani et al., 2021) The
design of social media platforms, especially Twitter and TikTok, democratizes political discourse by enabling
direct communication with constituents and the strategy of personalized, highly polarized, and emotionally
charged discourse (Berdón-Prieto et al., 2023; Diez-Gracia et al., 2023). The media also performs the crucial
task of framing the political discourse in such a way that it aims to achieve a particular outcome (Abdul Wahab
& Tew Abdullah, 2025).
Pauline Leong's (2015) research in Malaysia demonstrated that politicians utilise the media to influence and
secure voter support, particularly during electoral periods. Her investigation of the political process and
communication techniques in Malaysia revealed that new media, especially Web 2.0, has expanded the public
sphere and enhanced democratic involvement among Malaysians via information dissemination, mobilisation,
and crowdsourcing. New media has enabled the rise of varied viewpoints that challenge political dominance.
Communication is more bilateral, with the public expecting greater participation and interactivity with their
elected officials. Her investigation indicated that the Internet and social media had injected unparalleled
complexity into the political communication process in Malaysia. External factors, such as the electoral system
and political institutions, affect the degree to which ideas propagated through social media can gain traction
within the political sphere, potentially resulting in political change. According to Mohd Faizal (2024), social
media heavily influences politics and democracy in developing democracies like Malaysia, particularly among
youth. Social media allows groups to voice their thoughts, expanding the public sphere in semi-democratic
countries like Malaysia. Social media allows members of diverse groups and cultures to participate
meaningfully in different registers or languages, unlike the public realm. His research on young voters and
social media reveals a passive interest in political information, with Twitter and Facebook being the most
influential channels for first-time voters seeking information. Additionally, his research indicates that first-time
voters are increasingly using private WhatsApp groups to engage in political discussions and debates due to
concerns about social media surveillance. WhatsApp and Instagram stories are becoming the primary venue for
disseminating political information.
www.rsisinternational.org
Page 5820
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
Combining these strands, AI utilizes social media affordances to shape political discourse. Sentiment analysis
and automated AI messaging tools, tools of AI, are widely used in political marketing to customize
communications to target specific subgroups within the electorate. This strategy amplifies polarization and
fabricates consensus in opinion manipulation (Das & Malaviya, 2025; Jain & Mitra, 2025; López Ponce et al.,
2024). AI systems, including deepfakes, are being used to make fake yet remarkably realistic material that can
alter perceptions and affect politics. This is especially the case with fake images of wars and other visual
disinformation (Bazarkina & Matyashova, 2022; García-Huete et al., 2025; Xiang, 2025). Direct
communication between political actors and the public has been made possible by social media platforms, first
and foremost Twitter and Facebook. They distribute political messages to the populace and serve as tools for
the organization of protests and other revolutionary actions (Ajaegbu & Ajaegbu, 2024; Jufri et al., 2024;
Stieglitz & Dang-Xuan, 2013). Social media users, especially through the mechanism of the homophily, can
connect and form communities around shared political ideas and are therefore able to create echo chambers.
This contributes to polarization and the entrenchment of beliefs (Matalon et al., 2021; Parmar et al., 2024; Wu
et al., 2025).
The interrelation of AI, political discourse, and social media is not only interwoven but is also evolving. Social
media and AI technologies can facilitate the evolution of politically targeted communication and engagement.
Yet, the technologies pose notable ethical and social problems. These problems will not be mitigated until there
is strong regulation and ethical social media use so that AI is integrated into democracy positively and social
media use is incorporated into democratic engagement.
METHODOLOGY
This study uses secondary data. It performs narrative analysis of literature on artificial intelligence (AI), social
media, and political discourse to describe the role of AI in shaping the political information individuals have
access to and the ways in which people talk about this information, as well as examining the reported
advantages and disadvantages in the recent literature. The period of study was from January 1, 2020, to
September 30, 2025. Data was collected using Google Scholar, Scopus, and the Web of Science (WoS). The
justification for the period of 2020 to 2025 is that it includes the period that the use of AI in politics evolved
quickly and became mainstream. Innovations in AI include the use of transformer-based models, large
language models, and diffusion-based generators which all greatly reduce the cost of creating political content.
Changes in AI for political discourse have also been made to automated moderation, recommendation systems,
and the policies of platforms regarding political content. Review of literature within this timeframe ensures
that the synthesis is based on contemporary technologies, practices, and governance issues.
The strategies for a search incorporated both general and function-specific terms customized to each database.
Central searches combined “social media” with “artificial intelligence,” “AI,” or “machine learning,” and
together with politically oriented terms like “political discourse,” “political communication,” or “election.”
The parameters for search results allowed for publications from 2020 to 2025 that were in English and
primarily journal articles and complete conference papers. Studies were included that focused on social media
and analysed at least one AI mediated function with results that were relevant to discourse at a political.
Contexts that were used for exclusion included purely non-social, politically non-relevant domains that did not
offer insights that were transferable, non-empirical or non-systematic opinion pieces, and works from 2020.
Background information may be cited for works prior to 2020, but core evidence should not include any
synthesis of these works.
FINDINGS AND DISCUSSION
Based on the narrative analysis of peer-reviewed research published between the years 2020 and 2025, the
analysis identifies six AI mediated functions on social media which include: algorithmic curation and ranking,
political ad targeting and delivery optimization, automated moderation and labeling, cross-cutting design
choices, context dependence and generative AI and synthetic political content. These functions collectively
shape the political information that users receive, the interpretative frameworks they apply during discussions,
and the legitimacy of the communicative context that surrounds them.
www.rsisinternational.org
Page 5821
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
Algorithmic curation and ranking
Concerning the targeting of political advertisements and the optimization of their delivery, artificial
intelligence (AI) systems have considerably reduced the cost of reaching audiences, thus enhancing the
efficiency of political campaigns for both established players and smaller civic organizations. The accessibility
features of political messages, such as automated captioning and translation, have democratized access for
audiences who have historically been sidelined by the mainstream media. However, the same optimization
algorithms, particularly those that automatically and dynamically identify cost-efficient audiences, can encode
and deepen demographic and ideological biases in exposure that may go well beyond explicit targeting (Ali et
al., 2021). As the delivery of political ads, as well as audience composition, remains opaque, users have little
understanding of who else is viewing a particular political message and the reasons for their selection, thereby
curtailing public accountability and enabling covert segmented persuasion that exploits psychological
manipulation (Kavoliūnaitė-Ragauskienė, 2024; Rubio Núñez, 2025). Citizens perceive the flow of political
communication as private and unaccountable when they suspect political messages are being circulated in this
manner. The analysis suggests that comprehensive political advertisement measures with thorough political ad
metadata, standardized “Why am I seeing this?” explanations, and “no sensitive attribute inference” controls
could contain the most severe efficiency-loss legitimacy trade-offs while still addressing the most severe gain-
loss efficiency trade-offs. The ability of independent auditors to access the data, particularly for researchers, is
valuable for identifying disparate impacts and for bounding optimization goals to civic integrity, rather than
just to cost and engagement.
Automated Moderation and Labeling
Automation in moderation and labeling shows the most drastic scaling benefits: it removes or downranks clear
policy violations, adds contextual labels to misleading claims, and minimizes exposure to abusive, harassing,
and other categories of disinformation (Gorwa et al., 2020). Alongside these benefits, error patterns become
systematic rather than random. More false negatives occur in minority and low-resource languages, while
enforcement inconsistency, geo and topic skews risk claims of bias or political favoritism and chill legitimate
speech. Although labels can decrease reshares of some false claims, when labels are inconsistent or rationales
opaque, the “implied truth” effect of unlabeled content may create distrust and backfire. Best results come
from shaped layered, proportionate “friction” interventions, like nudges to read an article before reshare, mild
virality delays on new accounts, and clear, user-facing explanations and appeal pathways. Investment in
multilingual model parity and third-party evaluation minimizes uneven burdens to marginalized communities
and is repeatedly emphasized as essential to making the safety offered fair.
Cross cutting design choices
Multiple functions adopt a few selected design choices as robust, low regret work interventions that enhance
the quality of discussion without imposing blanket restrictions on political expression (Ghose et al., 2025;
Suter et al., 2025). Achieving user agency is possible through the enforcement of accountability norms that
advocate for the use of clear, plain language on how the decisions made in targeting, ranking, and enforcement
of any restrictions are made (Ghose et al., 2025; Kashiramka et al., 2024; Suter et al., 2025) Constraints on
how feeds and ads are controlled, that are surfaced in user flows, enabling users to refine their experiences and
enhance the diversity of their information to lessen the impact of engagement driven curation and echo
chamber dynamics (Ali et al., 2021; Bartley et al., 2025; Pinto-Bustamante et al., 2023; Yazici, 2025). Careful
jugged friction on reshares and replies during periods of heightened risk for the spread of low-quality claims
and in high periods of quality engagement functions as a proportionate complement to moderation and claims
labeling and captures genuine engagement with only limited costs (Gorwa et al., 2020; Imran et al., 2025;
Vasist et al., 2025). Transparent, searchable political ad libraries with standardized, stable metadata and APIs
for researchers enable external scrutiny that deters abuse and informs policy by making microtargeting
practices and delivery skews auditable (Ali et al., 2021; Kavoliūnaitė-Ragauskienė, 2024; Rubio Núñez, 2025;
Suter et al., 2025). Finally, visible provenance for AI generated content helps establish shared epistemic norms
about what is synthetic and what is not, even if such signals cannot guarantee authenticity in every case, given
the demonstrated ease of producing convincing synthetic text, audio, image, and video and the associated risks
of deceptive persuasion and visual disinformation (Bazarkina & Matyashova, 2022; García-Huete et al., 2025;
www.rsisinternational.org
Page 5822
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
Imran et al., 2025; V. Kumar et al., 2024; Liv & Greenbaum, 2020; Sinha et al., 2024; Vasist et al., 2025;
Xiang, 2025)
Context Dependence
Although useful, the effects of such interventions still hold relevance to context specific details: the behavioral
style, the potential risks to be managed, the languages utilized, platform affordances, and simultaneous
physical world happenings (Ajaegbu & Ajaegbu, 2024; Berdón-Prieto et al., 2023; Diez-Gracia et al., 2023;
Ghose et al., 2025; Jufri et al., 2024; Long et al., 2020; Matalon et al., 2021; Parmar et al., 2024; Wu et al.,
2025). Variability and divergence are, thus, central to the concern under discussion. The same intervention,
deployed cross products and rhetorical formats, can bring about varying impacts. Designed norms and rules of
turn-taking, along with the attention the social architecture of a platform pays to them, interlace with the
platform's algorithmic goals to steer attention and conversation in such a way that they achieve the same
outcome on a targeted intervention but different results on varying products (Berdón-Prieto et al., 2023; Diez-
Gracia et al., 2023). The asymmetries of linguistic resources are still systematically underestimated: low
resource languages are more likely to have poor quality moderation, and low quality recommendations, which
intensifies the inequitable conditions and puts the community at a higher level of vulnerability to abuse and
misinformation in a context that demands reliable, equitable, and transparent AI (Ghose et al., 2025;
Kashiramka et al., 2024; Yazici, 2025). High stakes political moments elections, referenda, crises amplify both
the benefits of rapid information access and the risks of harmful virality, as evidenced by platform mediated
mobilization and strategic communication in major political events (Abdul Wahab & Tew Abdullah, 2025;
Ajaegbu & Ajaegbu, 2024; Jufri et al., 2024; Long et al., 2020). Time bound, clearly communicated integrity
measures tailored to such moments for example, elevating authoritative information, temporarily increasing
friction on reshares, or tightening synthetic media policies appear to produce the most favorable tradeoffs
when they are transparently scoped and independently evaluated within ethical governance frameworks (Cupać
& Sienknecht, 2024; Gorwa et al., 2020; Imran et al., 2025; Suter et al., 2025).
Generative AI and synthetic political content
In this context, generative AI poses a different challenge as it collapses the marginal cost of creating
convincing written political communication, imagery, audio and video (Kumar et al., 2024; Liv & Greenbaum,
2020; Sinha et al., 2024) On the one hand, tools of this kind assist in translation, summarization, and
accessibility, increasing the volume and reach of civic information than previously possible in diverse formats
and languages. On the other hand, they enable the rapid production and distribution of deceptive synthetic
media and identity-based manipulations which exceed the current capacity of integrity systems (Cupać &
Sienknecht, 2024; Imran et al., 2025). While provenance and watermarking solutions are advancing and have
demonstrated value, they are, by themselves, insufficient. Adoption rates are lumpy and uneven, signals are
easily stripped, spoofed, or lost, and users do not know, care, or misunderstand provenance indicators. The
literature suggests that visible and consistent labeling of AI generated or AI edited content coupled with creator
side tools which attach content credentials instantly and coordinated detection and response across platforms
significantly improve resilience. These measures are most effective when combined with user training and
calibrated friction on the rapid dissemination of new or sensational claims, especially during periods of
heightened public interest such as election campaigns.
Finally, the results indicate that the impact of AI on political discourse is not strictly negative, nor is it wholly
liberating; it is, instead, conditional and can be mindful crafted. The same capabilities that underlie
personalization, efficiency, and accessibility can, with civically focused aims of the platforms, and with
independently resourced audits, be engineered to prioritize deliberative quality, diversity of views, legitimizing
discourse, and civically aligned investments in provenance, multilingual equity, and proportionate friction.
Policy choice can reinforce these decisions by bundling the preservation of space for deemed legitimate
political expression with foundational algorithmic transparency, accountability for political advertisements,
and protective measures against deceptive synthetic impersonation. The underspecified limitations within the
available evidence, particularly the constraining researcher access to platform data, and the pace with which
models and policies are evolving, accent the urgency for cross platform, multilingual, pre-registered studies
designed to evaluate not only engagement but also trust, polarization, and participation.
www.rsisinternational.org
Page 5823
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
Table 1: Six Functions Derived from the Analysis
Function
Core purpose
Typical inputs/logic
Primary outcome
emphasis
Algorithmic
Curation &
Ranking
Orders and surfaces
content in feeds
Engagement proxies
(click-through, watch
time, dwell), freshness,
social graph,
quality/credibility signals
Exposure (strong);
Interpretation (medium via
frames/cues); Legitimacy
(indirect via perceived
bias)
Political Ad
Targeting &
Delivery
Optimization
Finds cost-effective
audiences and
optimizes delivery
for political
messages
Advertiser parameters +
system delivery
optimization; inferred
interests/demographics;
lookalikes
Exposure (segmented);
Interpretation (segmented
narratives); Legitimacy
(opacity concerns)
Automated
Moderation &
Labeling
Enforces policies,
reduces harm, adds
context
ML classifiers; policy
rules; fact-
checks/knowledge bases
Exposure (reduce violative
content); Interpretation
(labels); Legitimacy
(depends on
consistency/fairness)
Cross-Cutting
Design Choices
Explanations, user
controls, calibrated
friction, ad
transparency,
provenance
UI controls; "Why am I
seeing this?"; reshare
prompts/delays; ad
libraries; content
credentials
Legitimacy (increase);
Exposure/Interpretation
(user-steered)
Context
Dependence
Modulates effects by
platform format,
language resources,
and events
Short-form vs threaded
formats; linguistic
resource availability;
offline salience and
timing
All three; effects vary by
context
Generative AI &
Synthetic Content
Produces
persuasive/synthetic
text, image, audio,
video at low
marginal cost
LLMs; diffusion models;
watermarking/provenance
signals
Exposure
(volume/novelty);
Interpretation
(ambiguity/credibility);
Legitimacy (risk without
provenance)
CONCLUSION
This conceptual paper aims to explore the role of artificial intelligence tools in social media in shaping political
discourse as well as the pieces of information people are exposed to and the benefits and disadvantages of such
artificial intelligence tools in online political discourse. By conducting a narrative synthesis of literature
published from 2020 to 2025, the analysis uncovered that six AI functions on social media are AI driven
algorithmic curation and ranking, automated political advertisement targeting and delivery optimization,
automated moderation and labeling, cross-cutting design choices, context dependence and the generative AI
and synthetic political content. From a more positive perspective, these systems improve relevance by
personalizing user feeds, rule violating content is removed or downranked resulting in improved safety,
accessibility is enhanced through translations and multimodal formats, and campaign audience targeting
precision increases with more automated tools. On the negative side, the same optimization logics narrow
information diets, privilege emotionally charged and polarizing material, accelerating the spread of low-quality
claims, produce uneven enforcement that can chill legitimate speech or exacerbate inequities across languages
and communities, and undermine perceived legitimacy through opaque microtargeting and imperfect
provenance signals for synthetic media. In conclusion, AI on social media is not simply a tool that uniformly
www.rsisinternational.org
Page 5824
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
degrades or enhances political discourse artificially. Rather, it is a set of configurable systems whose impacts
depend on design decisions. These decisions include decontextualized moderation systems, transparency
policies, and governance frameworks. In this paper, I argue that product designs featuring explanation and
control, calibrated friction, ad transparency and provenance, and independent evaluation combined with
proportionate regulation can immensely reduce relevance, safety, and accessibility. These product designs can
also lessen the harms of narrowed exposure, accessibility, polarization, misinformation diffusion, and erosion
of legitimacy. Platforms, policymakers, researchers, and civil society face a challenge to move beyond
simplistic narratives of technological determinism. They must also engage in the deliberate and evidence-based
engineering of AI systems to achieve democratic purposes. These stakes have become high. The online quality
of political discourse impacts participation in elections, public trust in democratic institutions, and the ability
to engage in informed, pluralistic deliberation. These stakes also drive the design of AI systems that engage in
diverse political discourse. These systems should be built AI's role in political discourse should be treated as
designable as opposed to inevitable, will ensure a communicative environment that fosters the amplification of
diverse voices and the elevation of credible, civil discourse.
REFERENCES
1. Abdul Wahab, S., & Tew Abdullah, M. H. (2025). Media framing on Bernama’s Twitter coverage during
Malaysia’s GE15: An exploratory factor analysis. SEARCH Journal Media and Communication
Research, 17(1), 41–57. https://doi.org/10.58946/search-17.1.P3
2. Ajaegbu, O. O., & Ajaegbu, C. (2024). The new democratisation: Social media impact on the political
process in Sub-Saharan Africa. Frontiers in Communication, 9, 1394949.
https://doi.org/10.3389/fcomm.2024.1394949
3. Ali, M., Sapiezynski, P., Korolova, A., Mislove, A., & Rieke, A. (2021). Ad Delivery Algorithms: The
Hidden Arbiters of Political Messaging. Proceedings of the 14th ACM International Conference on Web
Search and Data Mining, 13–21. https://doi.org/10.1145/3437963.3441801
4. Akmar Hayati, A.G. (2021, Januari 7). Pengguna setia media sosial. Harian Metro. Retrieved from
https://www.hmetro.com.my/WM/2021/01/661173/pengguna-setia-media-sosial
5. Ayma Quirita, V. A., David Cárdenas, J., Aliaga, W., Ayma Quirita, V. H., Palacios, A., & Sierra, R. B.
(2024). Peruvian Presidential Debates in the Elections of 2021 in Twitter/X: A Sentiment Analysis
Approach. IEEE Access, 12, 138386–138398. https://doi.org/10.1109/ACCESS.2024.3465503
6. Bartley, N., Burghardt, K., & Lerman, K. (2025). Impacts of Personalization on Social Network
Exposure. In L. M. Aiello, T. Chakraborty, & S. Gaito (Eds.), Social Networks Analysis and Mining (Vol.
15212, pp. 38–53). Springer Nature Switzerland. https://doi.org/10.1007/978-3-031-78538-2_3
7. Bazarkina, D., & Matyashova, D. (2022). “Smart” Psychological Operations in Social Media: Security
Challenges in China and Germany. 14–20. Scopus. https://www.scopus.com/inward/record.uri?eid=2-
s2.0-85147530339&partnerID=40&md5=5f9ca306159032f911d67b4bab9e5a6c
8. Berdón-Prieto, P., Herrero-Izquierdo, J., & Reguero-Sanz, I. (2023). Political polarization and
politainment: Methodology for analyzing crypto hate speech on TikTok. El Profesional de La
Información, e320601. https://doi.org/10.3145/epi.2023.nov.01
9. Bukhari, M., & Anwar, S. (2025). The Influence of AI on Models of Governance in Authoritarian and
Democratic Systems. Journal of Political Stability Archive, 3(1), 887–902.
https://doi.org/10.63468/jpsa.3.1.53
10. Cherikbaeva, N., Musaeva, T., & Kyrgyz-Turkish Manas University. (2024). Political Discourse And Its
Role In Shaping Public Opinion. Alatoo Academic Studies, 24(1), 277–290.
https://doi.org/10.17015/aas.2024.241.25
11. Cupać, J., & Sienknecht, M. (2024). Regulate against the machine: How the EU mitigates AI harm to
democracy. Democratization, 31(5), 1067–1090. https://doi.org/10.1080/13510347.2024.2353706
12. Das, A., & Malaviya, S. (2025). Artificial Intelligence and Sentiment Analysis in Political Campaigns. In
Cyber‐Physical Systems for Innovating and Transforming Society 5.0 (1st ed., pp. 193–209). Wiley.
https://doi.org/10.1002/9781394197750.ch9
13. Dávila Cortés, I. E. (2025). Los derechos político-electorales en México ante los desafíos de la
Inteligencia Artificial. Revista de Derecho Político, 123, 345–372.
https://doi.org/10.5944/rdp.123.2025.45516
www.rsisinternational.org
Page 5825
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
14. Deng, L. (2018). Artificial Intelligence in the Rising Wave of Deep Learning: The Historical Path and
Future Outlook [Perspectives]. IEEE Signal Processing Magazine, 35(1), 180–177.
https://doi.org/10.1109/MSP.2017.2762725
15. Diez-Gracia, A., Sánchez-García, P., & Martín-Román, J. (2023). Polarización y discurso emocional de
la agenda política en redes sociales: Desintermediación y engagement en campaña electoral. Revista
ICONO 14. Revista Científica de Comunicación y Tecnologías Emergentes, 21(1).
https://doi.org/10.7195/ri14.v21i1.1922
16. García-Huete, E., Ignacio-Cerrato, S., Pacios, D., Vázquez-Poletti, J. L., Pérez-Serrano, M. J., Donofrio,
A., Cesarano, C., Schetakis, N., & Di Iorio, A. (2025). Evaluating the role of generative AI and color
patterns in the dissemination of war imagery and disinformation on social media. Frontiers in Artificial
Intelligence, 7, 1457247. https://doi.org/10.3389/frai.2024.1457247
17. Ghose, A., Pallav, P., & Ali, S. M. A. (2025). AI and Political Polarization: Analyzing the Impact of
Algorithmic Bias on Governance in Diverse Political Systems. In A. Günar (Ed.), Advances in Finance,
Accounting, and Economics (pp. 135–160). IGI Global. https://doi.org/10.4018/979-8-3693-7036-
0.ch007
18. Gorwa, R., Binns, R., & Katzenbach, C. (2020). Algorithmic content moderation: Technical and political
challenges in the automation of platform governance. Big Data & Society, 7(1), 205395171989794.
https://doi.org/10.1177/2053951719897945
19. Hartono, B., Suasana, I. S., Muthohir, M., Suresh, R., & Prihatmoko, S. (2025). Sociotechnical
Perspectives on Balancing Human Judgment and AI in Content Moderation. 2025 4th International
Conference on Creative Communication and Innovative Technology (ICCIT), 1–7.
https://doi.org/10.1109/ICCIT65724.2025.11167561
20. Imran, H., Maqsood, M., & Imran, Q. (2025). The deepfake crisis: Exploring new media literacy,
information integrity, and trust among Pakistani youth. Qualitative Research Reports in Communication,
1–12. https://doi.org/10.1080/17459435.2025.2541214
21. Jain, V., & Mitra, A. (2025). The Role of Social Media and AI in Shaping Neo-Propaganda for Populist
Movements: A Global Perspective. In A. T. Vieira, A. F. Joaquim, & A. Duarte (Eds.), Advances in
Wireless Technologies and Telecommunication (pp. 63–86). IGI Global Scientific Publishing.
https://doi.org/10.4018/979-8-3693-9999-6.ch004
22. Jufri, Rahman, A., & Suarga, H. (2024). Exploring the impact of social media on political discourse: A
case study of the Makassar mayoral election. International Journal of Advanced Technology and
Engineering Exploration, 11(114). https://doi.org/10.19101/IJATEE.2023.10102458
23. Kashiramka, S., Gupta, P., & Barman, S. (2024). Literature Review of Artificial Intelligence’s Impact on
Politics and Society. 2024 IEEE 11th Uttar Pradesh Section International Conference on Electrical,
Electronics and Computer Engineering (UPCON), 1–6.
https://doi.org/10.1109/UPCON62832.2024.10983037
24. Kavoliūnaitė-Ragauskienė, E. (2024). Artificial Intelligence in Manipulation: The Significance and
Strategies for Prevention. Baltic Journal of Law & Politics, 17(2), 116–141. https://doi.org/10.2478/bjlp-
2024-00018
25. Kumar, S., Musharaf, D., & Bhushan, B. (2024). Taming The Hate: Machine Learning Analysis Of Hate
Speech. 2024 2nd International Conference on Disruptive Technologies (ICDT), 269–273.
https://doi.org/10.1109/ICDT61202.2024.10489362
26. Kumar, V., Kapoor, A., Chaudhary, R. R., Gupta, L., & Khokhar, D. (2024). Preserving Integrity: A
Binary Classification Approach to Unmasking Artificially Generated Voices in the Age of Deepkakes.
2024 11th International Conference on Computing for Sustainable Global Development (INDIACom),
1449–1454. https://doi.org/10.23919/INDIACom61295.2024.10499177
27. Latif, E. A. (2017). The Routledge Handbook of Arabic Linguistics (1st ed.). Routledge.
https://doi.org/10.4324/9781315147062
28. Liu, B. (2021). In AI We Trust? Effects of Agency Locus and Transparency on Uncertainty Reduction in
Human–AI Interaction. Journal of Computer-Mediated Communication, 26(6), 384–402.
https://doi.org/10.1093/jcmc/zmab013
29. Liv, N., & Greenbaum, D. (2020). Deep Fakes and Memory Malleability: False Memories in the Service
of Fake News. AJOB Neuroscience, 11(2), 96–104. https://doi.org/10.1080/21507740.2020.1740351
www.rsisinternational.org
Page 5826
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
30. Long, Z., Chen, J., & Bao, T. (2020). Research on International Relations in the Age of Artificial
Intelligence. 2020 International Conference on Computer Engineering and Application (ICCEA), 436–
438. https://doi.org/10.1109/ICCEA50009.2020.00100
31. López Ponce, M., Barredo Ibañez, D., & Sánchez Gonzáles, H. (2024). Usos y riesgos de la Inteligencia
Artificial en las campañas electorales 2023: Encuesta Delphi a expertos estratégicos de Colombia.
Revista ICONO 14. Revista Científica de Comunicación y Tecnologías Emergentes, 22(1), e2078.
https://doi.org/10.7195/ri14.v22i2.2078
32. Lv, Z. (2023). Generative artificial intelligence in the metaverse era. Cognitive Robotics, 3, 208–217.
https://doi.org/10.1016/j.cogr.2023.06.001
33. Matalon, Y., Magdaci, O., Almozlino, A., & Yamin, D. (2021). Using sentiment analysis to predict
opinion inversion in Tweets of political communication. Scientific Reports, 11(1), 7250.
https://doi.org/10.1038/s41598-021-86510-w
34. Mohd Faizal, K. (2024). Social media as an online public sphere: A study among the first-time Malay
voters. Kajian Malaysia, 42(2), 29‒50.
35. Morandín-Ahuerma, F. (2022). What is Artificial Intelligence? International Journal of Research
Publication and Reviews, 03(12), 1947–1951. https://doi.org/10.55248/gengpi.2022.31261
36. Nur Nasliza Arina, M.N., & Mohammad Nurhafiz, H. (2023). Mass Media: Suitability and Efficacy in
Shaping Youth's Personality Towards Becoming Quality Human Capital. e-Journal of Media & Society,
6(2), 15-25.
37. Parmar, S., Patel, P., & Farooqui, Y. (2024). Homophily and Sentiment Analysis for Political Issues
Related Discussions in Twitter. 2024 4th International Conference on Sustainable Expert Systems
(ICSES), 604–608. https://doi.org/10.1109/ICSES63445.2024.10762976
38. Pauline Leong. (2015). Political Communication in Malaysia: A Study on the Use of New Media in
Politics. E-journal of Democracy JeDEM 7(1), 46-71.
39. Pinto-Bustamante, B. J., Riaño-Moreno, J. C., Clavijo-Montoya, H. A., Cárdenas-Galindo, M. A., &
Campos-Figueredo, W. D. (2023). Bioethics and artificial intelligence: Between deliberation on values
and rational choice theory. Frontiers in Robotics and AI, 10, 1140901.
https://doi.org/10.3389/frobt.2023.1140901
40. Rubio Núñez, R. (2025). El uso de la inteligencia artificial en las campañas electorales y sus efectos
democráticos. Revista de Derecho Político, 122, 65–102. https://doi.org/10.5944/rdp.122.2025.44742
41. Shkvorchenko, N. (2020). Linguistic and gender peculiarities of english political discourse. Analele
Universitatii Din Craiova - Seria Stiinte Filologice, Lingvistica, 42(1–2), 398–416. Scopus.
42. Sinha, S., Datta, S., Kumar, R., Bhattacharya, S., Sarkar, A., & Das, K. (2024). Exploring Creativity: The
Development and Uses of Generative AI. In R. Kumar, S. Sahu, & S. Bhattacharya (Eds.), Advances in
Computational Intelligence and Robotics (pp. 167–198). IGI Global. https://doi.org/10.4018/979-8-3693-
3278-8.ch008
43. Sravani, D., Kameswari, L., & Mamidi, R. (2021). Political Discourse Analysis: A Case Study of Code
Mixing and Code Switching in Political Speeches. Workshop on Computational Approaches to
Linguistic Code-Switching, Stroudsburg, PA.
44. Stieglitz, S., & Dang-Xuan, L. (2013). Social media and political communication: A social media
analytics framework. Social Network Analysis and Mining, 3(4), 1277–1291.
https://doi.org/10.1007/s13278-012-0079-3
45. Suter, V., Ma, C., Pöhlmann, G., & Meckel, M. (2025). When Politicians Talk AI: Issue‐Frames in
Parliamentary Debates Before and After ChatGPT. Policy & Internet, 17(3), e70010.
https://doi.org/10.1002/poi3.70010
46. Tembhurne, J. V., Lakhotia, K., & Agrawal, A. (2025). Twitter sentiment analysis using ensemble of
multi-channel model based on machine learning and deep learning techniques. Knowledge and
Information Systems, 67(2), 1045–1071. https://doi.org/10.1007/s10115-024-02256-7
47. Van Noordt, C., & Misuraca, G. (2022). Artificial intelligence for the public sector: Results of
landscaping the use of AI in government across the European Union. Government Information Quarterly,
39(3), 101714. https://doi.org/10.1016/j.giq.2022.101714
48. Vasist, P. N., Krishnan, S., IIM Kozhikode, India & LUT School of Engineering Sciences, Teo, T.,
National University of Singapore, Azad, N., & LUT School of Engineering Sciences. (2025). Exploring
Concerns of Fake News on ChatGPT: A Network Analysis of Social Media Conversations.
www.rsisinternational.org
Page 5827
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
Communications of the Association for Information Systems, 56, 936–960.
https://doi.org/10.17705/1CAIS.05636
49. Wang, S., & Kim, K. J. (2023). Content Moderation on Social Media: Does It Matter Who and Why
Moderates Hate Speech? Cyberpsychology, Behavior, and Social Networking, 26(7), 527–534.
https://doi.org/10.1089/cyber.2022.0158
50. Wu, C., Jiang, S., Sun, J., & Liu, Y. (2025). Research on the influence mechanism of emotional
communication on Twitter (X) and the effect of spreading public anger. Acta Psychologica, 260, 105560.
https://doi.org/10.1016/j.actpsy.2025.105560
51. Xiang, Y. (2025). Framing the Fallout: AI-Generated Visual Disinformation and the Fukushima
Wastewater Narrative on Chinese Social Media. 406–410. Scopus.
https://www.scopus.com/inward/record.uri?eid=2-s2.0-
105011600010&partnerID=40&md5=7f71e357ea13eecf12962a79b73b1f57
52. Yadav, Dr. R. R., Gupta, Asst. Prof. A., Pandey, Asst. Prof. P. A., Pandey, Asst. Prof. A., & Wavhal, Asst.
Prof. S. (2025). A Study of Applications of Artificial Intelligence. International Journal of Research and
Scientific Innovation, XII(IV), 782–786. https://doi.org/10.51244/IJRSI.2025.12040067
53. Yazici, T. (2025). Artificial Intelligence Ethics in Communication: Challenges and Future Perspectives.
In A. Kalafat Çat, M. A. Yetgin, & J. I. Sarker (Eds.), Navigating Modern Digital Communication Ethics
and Law (pp. 33–74). IGI Global Scientific Publishing. https://doi.org/10.4018/979-8-3373-1702-
1.ch002
54. Zeng, J. (2020). Artificial intelligence and China’s authoritarian governance. International Affairs, 96(6),
1441–1459. https://doi.org/10.1093/ia/iiaa172
