Louisa Ha

School of Media and Communication, Bowling Green State University, Bowling Green, OH 43403, U.S.A.

Journal of Shanghai Institute of Socialism, 2026, No. 6 (overall issue 224), pp. 34–56. Online-first publication: September 28, 2026, 16:04:50.

Source: https://link.cnki.net/urlid/3.1903.C.20260928.1116.002

Abstract

Based on a roundtable discussion attended by editors of international journals from China, the United States and Pakistan, this paper systematically reviews the current artificial intelligence (AI) policies of major international English-language academic journal publishers and renowned academic associations, and further examines the interdependent relationships between reviewers and AI, and between publishers and AI. Centering on academic integrity and the originality of knowledge production, this paper explores the academic integrity crisis stemming from technological transgression and ethical issues in the use of AI, the academic self-awareness and ethical responsibility of scholars in human-AI collaboration, and the practical difficulties in distinguishing AI-assisted writing from rigorous scientific judgment in the implementation of AI policies. AI cannot assume the academic and ethical responsibilities that human authors should bear, and therefore should not be listed as an author of academic papers. The auxiliary application of AI in academic research should be incorporated into a tiered governance framework for regulation. At the first tier, routine auxiliary use may be exempt from disclosure or subject only to an advisory statement. At the second tier, any use involving generative content must be explicitly disclosed. At the third tier, any application of AI involving plagiarism, fabrication, or other forms of academic misconduct should be strictly prohibited. As AI transitions through multiple roles as assistant, collaborator, and mentor, AI policies should also continue to evolve, attending to the dynamic adjustment needs of AI academic governance and remaining vigilant against structural imbalances in the allocation of AI resources. The development of AI also offers the academic community an opportunity to re-examine “what defines excellent academic research”, prompting scholars to lead by example in demonstrating how to use AI in a manner consistent with publishing ethics and in a responsible way, so as to safeguard the fundamental values of academic research, including originality of thought and diversity of perspectives.

Keywords: artificial intelligence; AI policies; academic publishing; publishing ethics; academic research

Funding: This article reports an interim outcome of the major National Social Science Fund project “Research on the Design of an International Communication Effectiveness Evaluation System and the Development of Its Mechanisms” (24&ZD216).

Author biography: Louisa Ha is a Distinguished Research Professor in the School of Media and Communication at Bowling Green State University, U.S.A.; founding editor-in-chief of Shanghai International Studies University’s English-language academic journal Online Media and Global Communication; and former editor-in-chief of Journalism & Mass Communication Quarterly.

Suggested citation in the source: Ha, Louisa. “Publishing Ethics in the Age of Artificial Intelligence: Policies, Challenges, and Implications for Academic Research.” Journal of Shanghai Institute of Socialism, 2026 (6): –.

Chinese Library Classification: TP18; G239.1. Document identification code: A.

I. The Current State of AI Policies in Academic Publishing

As AI has become increasingly prevalent in academic research and manuscript writing, most international publishers and scholarly associations have successively issued relevant AI policies between late 2025 and 2026. Meanwhile, Chinese journals and academic communities have also proposed corresponding initiatives and standards. These policies and standards define the boundaries between acceptable and unacceptable AI use and specify requirements concerning academic ethics and research. To explore the new challenges and questions facing publishing ethics in the AI era, this article first systematically reviews the AI policies that major academic journal publishers have established for authors and reviewers. The foremost point of consensus is that AI cannot be an author of an academic paper.

A. AI Cannot Be an Author: Authors’ Responsibilities and Disclosure Requirements

Most international English-language academic journals are operated by commercial publishers. Four major publishers—De Gruyter Brill, SAGE, Taylor & Francis, and Springer Nature—explicitly reject AI authorship in their policies, primarily because AI cannot fulfill the responsibilities expected of human authors. AI can neither validly sign a publishing agreement consenting to publication nor provide contractually binding assurances that a work complies with publishing ethics. Accordingly, humans must lead and control the entire process of conducting research and preparing manuscripts.

Before incorporating any AI-generated content into a manuscript, authors are obliged to scrutinize it for possible errors, biases, and other ethical risks. AI may produce content that appears plausible but is inaccurate or misleading. Consequently, the named authors bear full and ultimate responsibility for every element of a published article, regardless of whether AI assisted in its production. The underlying rationale of these policies is to safeguard the integrity of academic research while acknowledging AI’s potential benefits as an auxiliary tool. Compliance helps journals ensure transparency in manuscript handling and legal compliance while sustaining the credibility of the scholarly record. Although all four publishers explicitly prohibit listing AI as an author, their disclosure requirements differ in specificity. Their policy choices reflect the combined influence of risk prevention, legal compliance, commercial interests, and academic norms.[1]

1. De Gruyter Brill’s AI Policy

The German publisher De Gruyter Brill has established an AI policy that provides clear and consistent guidance for all submitting authors. It welcomes responsible AI use as a tool supporting academic research but explicitly stipulates that manuscripts must not be wholly or predominantly AI-generated. Authorship and the associated accountability must remain with human scholars. Regarding visual materials, the policy prohibits using generative AI to create or manipulate images, charts, or graphical data; all visual materials must faithfully represent the original data. When AI tools are used in data analysis or image processing, its use must be explicitly disclosed, and the underlying original data must be presented in the manuscript or made publicly available as supplementary material. Every submission must include an Ethics and Legal Declaration stating whether AI tools were used and, if so, the nature and extent of their use. Although AI may assist with routine tasks such as grammar checking and language editing, such use must still be stated in the declaration. AI use for model construction, data generation, or substantive analysis requires a more detailed explanation. Authors are responsible for the accuracy, completeness, and fairness of all content, including any content produced with AI assistance.[2]

Online Media and Global Communication ( OMGC ), a communication journal currently ranked in Scopus’s first quartile, adopts De Gruyter Brill’s AI policy. Furthermore, because OMGC is published in Germany, it must comply fully with the European Union’s Artificial Intelligence Act (AI Act) and General Data Protection Regulation (GDPR); authors, editors, and reviewers are all subject to these legal frameworks. Reviewers must not upload copyrighted materials or manuscripts to publicly accessible AI systems unless they can ensure that doing so infringes neither third-party rights nor confidentiality obligations. The journal accepts no liability for copyright, privacy, or data protection infringements arising from authors’ use of AI tools.

Nevertheless, the current privacy and confidentiality provisions have certain limitations. The AI Act and GDPR compliance guidelines cited in De Gruyter Brill’s policy reflect a distinctly European regulatory orientation. Given OMGC’s international readership, authorship, and reviewer community, policymakers should also consider incorporating AI policies and regulatory frameworks from other jurisdictions, including the United States, the United Kingdom, China, and African countries, to make the policy more broadly applicable and inclusive.[3]

2. SAGE’s AI Policy

Headquartered in the United States, SAGE is one of the world’s leading independent academic publishers. It publishes more than a thousand journals in the humanities and social sciences, medicine, science, and engineering, and its publishing ethics and policies have long enjoyed considerable credibility in academia. In response to AI’s profound transformation of academic publishing, SAGE has issued and continually updated its AI policy in recent years. Its three-tier framework distinguishes assistive use, generative use, and prohibited improper use, providing systematic and clear operational guidance for authors, reviewers, and editors.

The first category, assistive AI, includes basic writing support such as grammar and spelling checks, language editing, sentence improvement, and organization of the manuscript’s overall structure. These uses are generally considered legitimate aids to academic writing and do not substantially affect the originality of the research’s ideas. They are therefore generally acceptable and need not be specifically disclosed in the manuscript. SAGE nevertheless reminds authors that they retain full responsibility for the accuracy and rigor of the revised content.

The second category, generative AI, encompasses academic tasks involving more substantive intervention, including, but not limited to, automatically generating research images, initial modeling and processing of research data, and structured organization of literature searches and reviews. In such cases, AI goes beyond linguistic assistance and participates substantively in constructing knowledge content. Authors must therefore clearly and systematically disclose the AI tools used, their specific purposes, and the extent of their involvement, enabling reviewers and readers to distinguish human intellectual contributions from tool-based assistance.

The third category covers strictly prohibited improper uses, including, but not limited to: (1) using generative AI to fabricate inaccurate text or falsify substantive research findings; (2) using sequences of prompts to generate experimental data or entire passages for submission in bulk, avoiding the intellectual effort expected of authors; (3) substituting AI conversational systems for real participants in interviews or focus groups within qualitative research designs, undermining the authenticity and ethical legitimacy of the data; (4) delegating core tasks requiring researchers’ considered judgment—such as thematic coding, theory development, and empirical analysis—to AI; (5) plagiarizing existing scholarship or improperly attributing the sources and contributions of cited literature, blurring the boundaries of intellectual ownership; (6) presenting AI-generated images, tables, or diagrams as original research data or novel findings; and (7) inventing, altering, or fabricating references and research claims, systematically distorting the scholarly record.[4]

This tiered framework both accepts and guides AI’s supporting role in research and establishes firm ethical limits and boundaries of responsibility. SAGE emphasizes that, regardless of the type of AI tool used, named authors bear ultimate responsibility for the accuracy, originality, and academic integrity of all content. The framework helps maintain standards in academic publishing and offers a useful example of how publishing ethics can remain adaptable and forward-looking amid rapid technological change.

3. Taylor & Francis’s AI Policy

As one of the world’s leading academic publishing groups, the United Kingdom’s Taylor & Francis has an extensive portfolio in communication, sociology, and the humanities and social sciences. It publishes more communication journals than any other publisher and has broad academic influence in these disciplines. In response to AI’s new challenges for academic publishing, Taylor & Francis has developed and released a systematic policy identifying the principal risks, acceptable uses, and expressly prohibited activities associated with AI in research.

The policy first identifies four major risks: (1) inaccurate outputs and algorithmic bias; (2) inadequate citation of existing literature and insufficiently transparent tracing of knowledge sources; (3) breaches of confidentiality and intellectual property infringement; and (4) unintended uses of research materials, such as training AI models.

Regarding acceptable and responsible uses, the policy identifies the following activities: (1) generating and exploring preliminary research ideas with AI; (2) copyediting, language editing, and translation; (3) interactive online literature searches using search engines enhanced by large language models; (4) automated classification and organization of research literature; and (5) assistance with coding or data processing. Taylor & Francis also distinctively allows individual journals to establish different specific rules within its overarching framework. Some journals, for example, may prohibit generative AI entirely except for language editing and copyediting. The policy explicitly prohibits using generative AI to create or manipulate research data or clinical test results, protecting research findings against distortion through technological misuse.[5]

4. Springer Nature’s AI Policy

Springer Nature, a German–British publishing company, is one of the world’s largest academic publishers. Its portfolio includes the Nature family and many other high-impact journals covering natural sciences, engineering, medicine, and social sciences. Its publishing policies have an important exemplary influence throughout global academia. In response to AI’s far-reaching effects on academic publishing, Springer Nature has issued an explicit AI policy centered on a human-oriented approach, advocating the careful introduction of AI tools with responsible and transparent use and effective human oversight.

The policy states: “We believe that AI can provide powerful support for research and publishing when used responsibly, transparently, and with appropriate human oversight. AI can improve efficiency and assist with tasks throughout the research and publishing process, but it must never replace scholarly judgment, accountability, or editorial decision-making.” On this basis, it sets out four fundamental principles: first, human accountability for scholarly work cannot be transferred to AI systems; second, AI can only provide assistance and cannot replace scholars’ own academic judgment; third, transparent disclosure of AI use helps build trust and confidence within the scholarly community; and fourth, AI use must strictly protect data confidentiality and security against potential information leaks.[6] Together, these principles constitute the policy’s foundational values and behavioral standards.

5. AI Policies of Prominent International Scholarly Associations and Explorations in China

AI development has also prompted prominent international scholarly associations to explore and formulate policies. This article focuses on the frameworks of two such associations while also examining actions and developments in Chinese academia.

The first is the Association for Education in Journalism and Mass Communication (AEJMC). Founded in 1924, it is the oldest academic association in mass communication. Its Policy on Ethical Use of Artificial Intelligence provides systematic guidance for AI use in submissions to AEJMC journals, conference papers, and their review processes. It defines AI, specifies when and how authors must disclose its use, and establishes corresponding responsibilities. The policy identifies five core principles: disclosure, transparency, reproducibility, evaluability, and responsibility. Authors must adhere to these principles even when AI use has unintended consequences. Specifically: (1) Disclosure: authors must disclose AI use in any form, including its specific contributions to content development, data analysis, or interpretation of results. (2) Transparency: authors must be fully transparent about the tools used, their purposes, and the nature of their assistance, whether accessed through standalone platforms or embedded functions in other software. (3) Reproducibility: disclosures must contain sufficient detail and practical information for others to reproduce the relevant processes. (4) Evaluability: disclosures must be detailed enough for peers to assess the appropriateness, quality, and potential effects of AI use. (5) Responsibility: whether AI is used directly or indirectly through embedded software, authors are responsible for unintended or unforeseeable consequences, such as fabrication, falsification, or plagiarism.[7]

The policy expressly prohibits misappropriating intellectual property, including plagiarism and uploading another author’s unpublished work without prior permission. It also establishes consequences for violations, including manuscript rejection and disqualification from participation in AEJMC conferences or submission to its journals, as well as a two-year suspension of the offending authors’ right to submit work bearing their names to AEJMC conferences or journals.

AEJMC’s emphasis on disclosure, authors’ responsibility for unintended consequences, and explicit sanctions provides useful guidance. However, its classification is relatively simple, distinguishing only permitted and prohibited use. Although it lists specific circumstances requiring disclosure and its transparency principle is commendable, it does not provide sufficiently detailed guidance on acceptable AI use in ethical gray areas such as idea generation and literature reviews.

The second association is the International Communication Association (ICA). It has established an AI policy task force consisting of eight scholars with extensive journal editing experience and relevant research backgrounds. ICA faces challenges closely resembling those confronting AEJMC. During the past year, AI use in conference submissions has attracted widespread attention and concern, prompting the association to create a task force to discuss and develop an appropriate policy. One of its greatest challenges is deciding what authors may and may not do with AI—a far more complex task than it initially appears. Academia has yet to reach a clear consensus. Some scholars, for example, believe AI may assist with language editing but not research idea generation; others maintain that AI-assisted brainstorming is entirely legitimate. This raises a further question: is brainstorming not itself integral to generating ideas? Where should the boundary lie? These questions are difficult and resist simple answers. Attempting to formulate highly detailed policies at this stage may therefore be unwise. The task force also cautions against framing AI in excessively negative terms, as though it inherently constitutes a problem. AI can provide many benefits to research. For scholars whose first language is not English, it can substantially improve academic writing and help bridge linguistic divides in international publishing. Alongside preventing misuse, a more constructive approach should help researchers use AI responsibly and transparently, both to improve language and to develop research ideas and optimize other stages of research.[8]

Some Chinese scholarly associations and researchers have likewise actively discussed and explored standards for AI use. On January 1, 2026, sixteen humanities and social science journal editors, primarily in communication, jointly signed an “Initiative on Regulating AI Use in Academic Papers.”[9] Addressed to authors, reviewers, and editors, it recognizes AI’s benefits while defining the boundaries of legitimate use. The initiative permits authors, but not reviewers, to use AI; reviewers must also refrain from discriminating against papers that use it appropriately. It further proposes a controversial possibility: establishing a journal that permits AI to be named as first author on topics concerning AI’s contribution to research, while human authors assume all responsibility.

Recognizing AI as first author has, in fact, provoked strong academic opposition. Most Chinese journals explicitly prohibit AI authorship. On July 15, 2026, China National Knowledge Infrastructure (CNKI), China’s most influential journal article database, publicly condemned the practice and announced that it would cease indexing and remove all articles naming AI as an author.[10] Notably, a Chinese journal attempted to publish AI-generated papers but subsequently deleted them all, on the grounds that the experiment set an undesirable precedent and could mislead readers and the scholarly community into treating AI-generated articles as original work written by human scholars. This raises an ethical issue deserving close attention. Although the journal has some standing in Chinese academia, its decision to experiment in this way warrants serious scrutiny.

These controversies raise a fundamental question: how should authorship be defined? There are multiple reasons to reject AI authorship. First is the origin of conceptualization: where do research ideas come from? If AI contributes ideas, it might theoretically be considered a contributor to some extent. Yet authorship entails much more than adding up contributions. We must ask what truly constitutes academic research. If machines can generate everything—including all analyses, text, and conclusions—what role remains for humans? If we stand aside and allow AI to do all the work, the result represents an AI product rather than human scholarship. Research fundamentally requires human intellectual effort, critical thinking, and independent judgment; machines, however sophisticated, lack these essential qualities.[11]

If responsibility is central to defining authorship, AI cannot qualify as an author, regardless of its contribution, because it cannot assume substantive academic responsibility or bear consequences for misconduct. Disclosing AI use does not diminish human authors’ obligations. Rather than naming AI as an author, researchers should disclose its use in activities such as idea generation, providing necessary transparency in the same way they report statistical software, literature sources, and other research tools and materials. When human creators declare AI use, their interaction with it in the research process is naturally made visible.

The concept of accountability also merits closer examination. Stating that authors are responsible is one thing; specifying what that responsibility means in practice is another. If readers discover factual errors in a published paper, how do existing mechanisms hold authors accountable operationally? An assertion of ultimate author responsibility may amount to a broad ethical declaration, while clear procedures for implementing that responsibility after errors occur remain lacking. A distinction is needed here: factual errors verifiable against public records, datasets, or established knowledge offer a comparatively straightforward route to accountability, at least in principle. The broader challenge, however, is to move from rhetorical declarations to concrete, workable institutional procedures ensuring that authors are held responsible for their academic conduct.[12]

B. Reviewers and AI: From Prohibition to Conditional Use

At present, major publishers and AEJMC explicitly prohibit reviewers from using AI to write peer-review reports. Their policies commonly state, for example, that “reviewers are strictly prohibited from using AI technology at any stage of the peer-review process,” or emphasize that “manuscripts submitted for review are confidential original documents; uploading them to AI systems violates confidentiality, infringes authors’ rights, and contravenes EU data protection obligations. Reviewers must rely on their own expertise when assessing manuscripts.”[13]

However, I do not support a blanket prohibition on AI tools in reviewing. OMGC, for example, does not currently require authors to disclose whether references were generated or organized with AI. Reviewers need to verify references, especially citations. As an important currency used to measure academic impact, citation accuracy directly affects the credibility of the scholarly record. Reviewers’ responsibilities have therefore increased substantially. Without AI assistance, they can hardly verify every cited source individually. One practical approach is to copy the reference text and use AI tools to check its authenticity and accuracy.

Rather than prohibiting AI in every circumstance, specific conditions should govern its use by reviewers. Reviewers could, for example, be authorized to use AI solely to verify citation accuracy and authenticity, randomly sampling references and cross-checking whether the original works actually contain the claims attributed to them. This would not require AI to examine the entire manuscript; intervention would remain strictly limited to targeted reference verification. The key consideration is that references are essentially publicly available bibliographic source data rather than the author’s proprietary intellectual property; the paper’s original and innovative arguments lie in the main text. AI-assisted reference checking could protect academic integrity without compromising the confidentiality of the manuscript’s core ideas. It would constitute a cautious, limited, and principled exception to the existing blanket prohibition.

Specifically, reviewers should be permitted to use AI to verify references but should not upload entire manuscripts. They must disclose to the editorial office which tools they use for this purpose and must not rely on AI to write the overall review. To facilitate this practice, editorial offices could provide the main text and references as separate files. This proposal does not advocate outsourcing evaluation of the entire article to AI; it focuses on the specific task of reference verification. AI is well known to generate erroneous citations or misrepresent the meaning of sources. Ironically, an effective way to identify these fabricated or inaccurate citations quickly often involves the very tools prohibited by current policies. We therefore face a dilemma: maintaining peer-review confidentiality and integrity while providing practical technical means for efficiently detecting this common form of misconduct. The challenge is to create a policy that confines AI use to citation checking, accommodates this narrowly defined need, and avoids introducing the other risks the prohibition was designed to prevent. Current policies do not adequately address this distinction, which deserves serious consideration in future revisions of peer-review guidelines.[14]

One possible solution is a shared verification platform that pools resources so reviewers can check references and information without relying on different personal AI tools. Different tools employ different underlying logic and algorithms, potentially producing inconsistent or unreliable verification results. A common platform could improve consistency and reliability, protecting review quality and reducing uncertainty arising from differences between tools.[15]

C. AI as a Reader: Distinguishing Types of Audiences

De Gruyter Brill’s data indicate that a substantial share of journal articles are now read by AI systems rather than humans. This fundamental shift raises profound questions about the purposes of academic publishing and its implications for journal editing: who is reading our articles?[16] Some journals have begun verifying whether online readers are humans or computers to better understand their audiences. Yet as AI becomes increasingly important in providing literature reviews and citation searches, journals inaccessible to AI will struggle to appear in its literature recommendations.

AI bots and web crawlers now dominate institutional repository traffic, accounting for as much as 89%.[17] During the first ten weeks of 2026, approximately half of inbound traffic to Silverchair, a major journal hosting platform serving publishers such as Oxford University Press and Emerald Publishing, came from bots, crawlers, and other automated systems. Their purposes varied: some indexed content for AI-assisted searches, some acted on behalf of chatbot users, and some attempted to collect accessible material for large language model training.[18]

Bots and humans should not, however, be treated as entirely separate audiences: a person may instruct an AI tool to retrieve and summarize an article. AI bots can search and read literature in response to user requests, whereas web crawlers such as Googlebot automatically and regularly crawl webpages without a direct request. Human use of Google Scholar illustrates how crawlers serve human authors. Nevertheless, distinguishing these two types of access is difficult unless journal websites are configured accordingly, and most have not completed such configurations. Thus an article’s nonhuman readers may include either web crawlers or AI bots. Although many publishers use bot and crawler management services such as Cloudflare and Akamai to distinguish traffic types, industry-wide systematic data remain unavailable on how many publishers use these services or obtain traffic reports broken down by crawler type.

When carrying out specific information searches, AI reading may resemble human reading in certain respects, but it differs fundamentally from reading for pleasure or exploratory interest. AI processes and reconstructs information, while human readers interpret and evaluate articles and situate them within broader intellectual and theoretical contexts. Humans read much more slowly. For large bodies of literature, AI reading and review tools can help locate and organize studies, extract information, and prepare initial summaries, but they cannot replace researchers’ critical reading and scholarly judgment.

Researchers hoping to save time by having AI read literature for them should recognize that tools such as Elicit, ResearchRabbit, and Connected Papers have no direct partnerships with journal publishers and therefore cannot access the full text of subscription journal articles. Before selecting a tool, researchers should determine which journals have partnerships with it and whether important journals in their field are covered, then supplement missing coverage with manual searches. Because AI can access open-access journals and abstracts from subscription journals, but not the full text of paywalled articles, users of general-purpose tools such as DeepSeek cannot obtain complete articles through chatbots. Much of the information they receive may derive from open-access journals of uneven quality. Only a small share of subscription journal articles are published open access; most remain behind publisher paywalls. Literature searches conducted by general-purpose AI tools are therefore likely to be substantially biased toward open-access journals and articles.[19]

The relationship between publishers and AI is also complex, perhaps comparable to the interdependence between social media and news media. Journal paywalls currently restrict AI access to the full text of approximately 50% of scholarly articles. Insufficient full-text access may cause lower-quality content to be repeatedly reused and reinforced in a vicious cycle, potentially threatening the integrity and credibility of academic publishing.[20] Yet journals inaccessible to AI will struggle to enter its literature lists as AI’s role in reviews and citation searches expands. This creates a dilemma: publishers need paywalls to sustain their business, but without AI-mediated readership, articles may receive fewer citations and less use. This tension is encouraging publishers to seek partnerships with AI companies.

AI companies and journal publishers are gradually becoming interdependent. AI access is an important prerequisite for journal content to be discovered through AI channels. Some major tools, including DeepSeek, have not established formal agreements with leading journal publishers. Meanwhile, certain publishers have reached agreements with particular AI companies allowing access to literature but prohibiting its use for model training. OpenAI’s ChatGPT and Anthropic’s Claude, for example, have established conditional content-access arrangements with Wiley, while Taylor & Francis has signed a partnership with Microsoft. As of now, some publishers, including Springer Nature, have not publicly established arrangements granting AI companies comprehensive full-text access. AI literature review tools may therefore offer incomplete or absent coverage of those publishers’ journals. Researchers should check whether their tools have appropriate partnerships with the journals they need, avoiding reviews that rely solely on abstracts, omit important studies, or fail to understand complete research content.

II. Rethinking AI’s Challenges for Academic Research

Before examining AI’s challenges, it is necessary to acknowledge its important benefits. AI can create fairer publishing opportunities for scholars from non-English-speaking countries and substantially increase research efficiency. However, efficiency must not come at the expense of quality. Every task performed by AI must undergo rigorous human scrutiny for accuracy and bias. Academic publishing should not resemble factory-style mass production; it should involve careful synthesis of knowledge and construction of systems of thought. Rethinking AI’s challenges is therefore a fundamental issue requiring collective consideration and response from all actors involved in producing and disseminating knowledge.

A. Technological Overreach and Ethical Violations: The Academic Integrity Crisis Caused by AI Misuse

1. Research Alienation Through the Excessive Use of Intelligent Technologies

The academic community’s foremost concern is that AI misuse may erode the foundations of academic integrity. Originality must therefore be established as an unwavering first principle before AI use is institutionalized through policy. Misuse may jeopardize the authenticity and originality of academic contributions; researchers’ intellectual achievements should not be improperly appropriated or diluted through technological intervention. Copyright represents the legal extension of this concern. Ethical and responsibility-related questions require thorough clarification at the beginning of policymaking. Only then can regulated AI use be legitimate and feasible.

Instead of simply regarding AI as a threat, we should ask a more fundamental question: what kind of scholarship do we wish to foster, and how can AI make legitimate and beneficial contributions? This requires clarifying the real concerns. Are we worried that authors’ distinctive voices and authentic scholarly contributions will be obscured or diminished? Or that the creative processes of problem-solving and critical analysis will be eroded? These underlying anxieties must be clearly articulated so that policy can respond thoughtfully rather than over-reaction.

The real risk arises when authors rely excessively on AI and outsource core scholarly judgments to algorithms. Literature reviews are a particularly vulnerable example. Reviews built entirely around AI systems may draw on structurally narrow or unbalanced sources, concentrate excessively on one scholarly tradition, or generate plausible-looking but inaccurate or entirely fabricated citations. The integrity and credibility of the scholarly record depend on systematic, careful, and critical engagement with existing literature. This core responsibility cannot and should not be outsourced. Good policy should target specific, identifiable risks capable of causing substantive harm, rather than rest on vague fears of technology. It should also support a discourse that broadens opportunities for academic participation.[21]

Internationally, academic communities differ significantly in how they understand and respond to AI. In Pakistan, scholars initially tended to view AI as a variant of plagiarism rather than a potentially useful research tool. During the early stages of policy implementation, the Higher Education Commission of Pakistan (HEC) concentrated on detection and compliance, treating AI-generated content much like plagiarism and using similarity thresholds, such as 18%, to determine acceptability. When detection suggested AI involvement exceeded permitted limits, students were simply told to revise or replace the content, without broader discussion of AI’s role in knowledge production.[22] Although this compliance-oriented approach can protect standards in the short term, it may inhibit researchers’ understanding and exploration of legitimate uses and intensify alienation in research practice.

2. The Crisis of Losing the Author’s Voice Under Algorithmic Mediation

Deep AI involvement in academic writing creates a less visible but far-reaching danger: the gradual erosion of authors’ own scholarly voices. When AI polishes language toward perfection, prose becomes fluent and orderly, yet authors may imperceptibly lose their distinctive ways of observing the world, identifying problems, and expressing ideas. This alienation is not overt plagiarism or fraud; it slowly undermines the scholarly self through everyday writing practices. As authors grow accustomed to AI-generated linguistic refinement, their own judgment and confidence may weaken, leading them to question whether unaided writing is adequate. A tool intended to assist can become a temptation or trap, encouraging authors to abandon authentic scholarly perspectives in favor of technically flawless, polished expression lacking individuality and human warmth.

This phenomenon is common in scholarly research. Bushra Hameedur Rahman recalled struggling when she read exceptionally polished literature reviews early in her career. Her attention centered more on achieving comparable stylistic excellence than on developing her distinctive analytical voice. AI now offers a similar but stronger temptation at lower cost and greater speed: superficial excellence becomes readily attainable, potentially discouraging the pursuit of intellectual depth and scholarly individuality.[23]

A comprehensive policy should therefore address more than dishonesty. It should clearly identify this deeper danger and advise authors to exercise restraint, both to protect academic integrity and to preserve the originality of their own perspectives. In an era of extensive algorithmic intervention, awareness of one’s own thinking and expression may be the final safeguard against losing one’s scholarly voice.

B. Critical Human–AI Collaboration: Scholarly Self-Awareness in an Era of Algorithmic Assistance

1. Questioning and Reshaping Original Knowledge Production

Widespread AI involvement in academic writing forces us to revisit a fundamental question: what is academic research? Research is an intellectual challenge, but it also carries a deeper mission of responding to human concerns and engaging social issues. If AI is understood as a tool helping researchers solve problems and improve human life, such assistance need not depart from academic values; it can be an integral part of scholarly practice.[24]

Within this framework, scholars’ core responsibilities require clarification: humans must determine how research conclusions are formed and interpreted. Every author should ask whether writing and publishing a paper creates any new knowledge. If the answer is no, the manuscript has no academic value, whether produced entirely by humans or with AI assistance. Creating new knowledge is an irreplaceable measure of scholarly contribution.

The essence of research lies in originality and control over thinking. As long as human authors retain final authority over what to include, exclude, and argue, the work fundamentally belongs to them. In this sense, AI need not be treated as an author or collaborator. It is essentially a resource whose contributions should be transparently disclosed, like other forms of scholarly assistance. An author could simply state that AI helped compare alternative titles and select the most effective one. This is no more improper than seeking advice from a colleague or mentor. We do not prohibit asking another person which title is better; why prohibit asking a machine the same question? The broad acceptance of AI for grammar and language editing is instructive. Scholars have long hired copyeditors or asked fluent colleagues to polish articles without such assistance being considered a violation of integrity. Why should the same function become problematic when performed by a language model? Accepting human assistance with expression while rejecting AI assistance in surface choices such as titles, despite unchanged ownership of the underlying knowledge, lacks internal consistency.

The real challenge is therefore less about constructing artificial boundaries around every minor instance of AI assistance than about ensuring truthful and sufficient disclosure and continued human control. Original knowledge production is grounded in clear ethical awareness and careful practice that reshape research’s value boundaries amid technological change, rather than in excluding technology.

2. Ethical Disclosure and Quality Thresholds in Academic Publishing

Once the fundamental principles of research are clear, publication criteria also become clear: manuscripts meeting academic standards, following ethical principles, using authentic data, and advancing knowledge in their fields should be eligible for publication. Whether authors used AI assistance is not itself the central criterion. Producing equivalent or superior work without AI is commendable, but seeking limited help with titles, subheadings, grammar, or rhetoric does not invalidate scholarly work. The unacceptable boundary is clear: entirely relying on AI for core content, fabricating data, or presenting findings as one’s own without substantively understanding the article’s basic argument and results fundamentally violates academic ethics. The boundary concerns authors’ substantive control and responsibility over knowledge production.

An example from my teaching provides an analogy. When I present preliminary research ideas to students and debate them, I release those ideas into a smaller space that I still cannot fully control. I cannot predict where the discussion will lead or how the insights may later be reproduced or transformed elsewhere. I accept that risk as intrinsic to intellectual exchange. Why, then, treat AI platforms as fundamentally different and more dangerous spaces? We draw a sharp distinction between human and machine mediation, tending to trust the former and distrust the latter. Yet we allow human reviewers, who may retain, reuse, or be influenced by a manuscript’s ideas, to access unpublished work. We should therefore ask whether a blanket AI prohibition creates an unfair obstacle. It may even be a barrier we impose on ourselves, limiting AI’s potential to advance scholarship. The central issue is not only controlling AI use but also assessing whether controls are proportionate to actual risks or inadvertently suppress the intellectual openness and circulation of ideas that open-access publishing should promote.

C. Challenges in Implementing AI Policies: Identifying AI-Assisted Writing and Safeguarding Academic Integrity

1. Limitations in the Validity of AI Detection Tools

Implementing and enforcing AI policies presents a difficult practical challenge: determining whether manuscript content is AI-generated. Editors must exercise great caution, yet available detection tools remain insufficiently reliable to support decisions. Highly academic prose may be flagged as AI-generated even when written independently by a human. This may be the greatest current obstacle to policy enforcement. Editors often have to rely on experience and professional judgment, considering overall style, argument structure, and language patterns. No single characteristic constitutes conclusive evidence of AI use. Final judgments must therefore rest on careful professional assessment rather than an automated test supposedly delivering certainty.[25]

In Pakistan, Turnitin is the main detection tool available to most academic institutions, but its reliability in distinguishing AI-generated content from human writing is inconsistent, placing considerable pressure on editors and reviewers. Initial screening at the Journal of Media and Culture Studies begins with a Turnitin similarity check; only manuscripts below the set threshold of 18% proceed to peer review. Editors must also consider multiple textual characteristics before sending work for review, and reviewers must assess reference authenticity and accuracy. A typical AI problem is fabricated citations or nonexistent sources. Detection becomes even more difficult when AI-generated manuscripts are rewritten with tools such as Humata to resemble human prose. Experienced reviewers can often still notice anomalies: complex theoretical arguments may be oversimplified, depth reduced, and overall style inconsistent with substantive scholarly engagement.[26]

The most effective route to sound policy begins with identifying what we are truly concerned about and wish to prevent, then asking whether policy can avert the outcomes considered most harmful. Fair policy must also protect authors against unsupported misconduct accusations and account for reviewers’ practical circumstances, supplying clear guidance and usable verification methods. Only by balancing effective responses to real harm with procedural justice and individual fairness can a framework be both principled and workable.[27]

2. Editors’ Dilemmas of Judgment in AI-Mediated Publishing

Detection tools’ validity limitations place journal editors in an unprecedented dilemma. Suppose a manuscript is wholly or predominantly prepared with AI assistance, existing tools fail to identify it, and the author omits disclosure or makes a false statement. It may pass screening and enter peer review, where reviewers themselves may use AI to evaluate it. Editors then become central to huge burden of responsibility marked by multiple uncertainties: the manuscript may be AI-generated, the reviews AI-assisted, and the editorial workflow itself supported by AI. The central question is how to preserve research integrity, originality, and substantive contribution under such uncertainty. If the entire chain of knowledge production, evaluation, and administration depends on AI, the risk of inaccurate content, fabricated information, manipulated datasets, unreliable references, and superficial analysis entering academic publishing rises substantially. Ultimately, responsibility for quality and credibility falls unavoidably on editors, demanding exceptional agency, scholarly expertise, and technical competence.[28]

This challenge is particularly acute in developing countries. Editorial offices often lack funding and technical infrastructure for comprehensive verification and response mechanisms. AI may be needed to improve editorial efficiency, but the degree of dependence requires critical examination.[29]

A deeper question follows: if manuscripts are AI-generated, AI-reviewed, and administered through AI-assisted processes, how much are we truly advancing knowledge and scholarship? How can publishing remain an expression of human intelligence rather than become an entirely machine-driven process? The challenge is therefore to establish a responsible, transparent, and ethical institutional framework that uses AI’s efficiency while protecting human judgment, creativity, critical thinking, and academic integrity.[30]

III. AI’s Multidimensional Implications for Academic Research

AI’s profound effects extend beyond efficiency gains and require an epistemological reconsideration of research’s basic logic. As a principle of use, AI should be positioned as an auxiliary resource rather than an author, with its involvement predicated on substantive human control and ultimate responsibility. In policy development, frameworks must remain open and responsive to technological change. These intertwined dimensions suggest how research’s foundational values and ethical boundaries can be reshaped.

A. Principles for AI Use in Academic Research

1. Situating Use in Specific Contexts and Responding to Local Conditions

AI use cannot be discussed in abstraction from concrete scholarly practices and cultural contexts. An effective policy framework must be rooted in particular disciplinary traditions and cultural settings if it is to regulate and guide practice meaningfully.

First, disciplines differ substantially in their dependence on AI and their acceptance of it. In natural sciences and engineering, AI is widely used for data processing, model construction, and experimental simulation, with its role as a tool relatively clear. In the humanities and social sciences, its involvement more often concerns text generation, literature organization, and language editing, producing distinctly different effects and ethical implications. Principles must therefore respond to each field’s epistemological traditions and research practices. The discussion above emphasizes shared principles, while specific measures may vary by discipline. Regulated AI use must also respond closely to local conditions of knowledge production, aligning technological logic with local academic governance. Generative AI relies on global datasets to create generalized production models that can produce homogeneous, standardized knowledge while overlooking regional differences. In cross-border scholarly communication and publishing, national differences in AI regulation, data protection, and academic ethics form distinctive local contexts. Because Western publications dominate and associated biases persist, databases often favor the inclusion and prominence of Western research. Scholars must recognize these limitations. AI’s involvement in knowledge production should be grounded in local institutional environments, academic traditions, and governance rules to establish principles that are legally legitimate, locally appropriate, and scientifically sound.

2. Understanding Algorithmic Logic and Governing Tools Through Principles

AI’s appropriate position in research depends on researchers’ understanding of algorithmic logic and their deliberate control of it. Without basic understanding, they may trust outputs blindly and mistake probabilistic predictions for certain knowledge. Anthropomorphic conversational interfaces may also encourage the illusion that AI is an understanding, creative collaborator rather than an assistant.

Bushra Hameedur Rahman’s experience illustrates this. She asked an AI system about its use of first-person pronouns. When it replied, “If you need anything else, I can help you,” she asked what it meant by “I.” Its revealing response defined itself simply as a tool helping users develop ideas. This clarification shaped her understanding of the relationship: AI is only a “humble little helper”—a chota in her local terminology— organizing and extending intellectual work that remains fundamentally the user’s own. She explained:

The ideas belong to me. The tool only organizes and presents them, occasionally offering a constructively different perspective. This leads to a personal observation: interacting with AI early in developing research ideas not only refines existing ideas but often generates new thinking. When the system adds even a little to an emerging idea, it frequently triggers a series of new associations I might struggle to make alone. This interaction extends thinking outward, enriching and deepening an initial insight along paths I might otherwise not explore. As a result, I can accomplish more in a relatively short time.

This is, of course, a researcher’s subjective experience. Nevertheless, it has convinced Rahman that careful AI use need not threaten intellectual originality. When human thought remains firmly in control and machines are positioned as junior partners rather than authors, collaboration can strengthen rather than diminish creativity.[3]

Researchers who see beyond AI’s anthropomorphic appearance and recognize its algorithmic logic can reasonably treat it as a tool that extends cognition. The ultimate purpose of understanding that logic is to govern tools through principles, making technology serve research’s fundamental aims.

B. Directions in the Evolution of AI Policies

1. Maintaining Policy Openness and Institutional Flexibility

AI policymaking should retain a forward-looking, open perspective and sufficient institutional flexibility for rapid technological change. AI’s widespread penetration is unavoidable; it is gradually becoming indispensable infrastructure for research and knowledge production. Policy deliberation should therefore begin from the reality that its use cannot be avoided. Fairness, transparency, and creativity should be the central values. Within the requirements of fairness and justice, policies should maximize creativity and originality in human–AI systems. Achieving a dynamic balance among fairness, justice, and creativity is thus central to policy design. Appropriate disclosure of AI’s specific contributions is a basic requirement for integrity and institutional transparency.

Originality does not emerge from nothing. From a media-theoretical perspective, original human thinking always operates through particular tools and discursive systems: paper, pens, computers, and associated symbol systems all participate deeply in generating thought. Paper, for example, affects how information is organized and analyzed. AI is becoming a new kind of “paper,” helping humans understand the world and integrate knowledge in distinctive ways. If our fundamental principle is to preserve originality and institutionally ensure meaningful knowledge creation, we should recognize AI’s substantive involvement and give it fair institutional treatment.[32]

AI admittedly lacks agency and the ability to assume responsibility, and never exists or operates independently of other actors. Humans unquestionably bear primary responsibility for misuse, but accountability should extend beyond direct users to algorithm developers, model trainers, producers and distributors of training data, and actors shaping AI’s regulatory and institutional environment. Nor can we confront AI empty-handed. Detecting misuse requires responding to technology with technology, using digital tools to identify vulnerabilities. Reasonable suspicion that a manuscript used AI cannot justify holding a researcher accountable on conjecture alone. Policies must therefore prevent abuse while providing sufficient procedural protections for accused individuals, avoiding errors and injustices caused by uncertain detection tools.

2. Addressing the Need for Ongoing Adjustment in Academic AI Governance

AI’s transformative impact requires not only redefining research but also constructing forward-looking and adaptable governance. AI policymaking is inherently ongoing and must respond continually to new challenges, technologies, and applications. This does not mean adjustments should lack rhythm or stability. A rigid schedule in which one set of rules applies until a deadline and is then replaced by another can create substantial practical difficulties. Researchers need stable, predictable guidance to work responsibly within clear standards. Repeatedly changing deadlines and requirements may create confusion rather than provide meaningful assistance. The timing and manner of adjustments therefore require careful consideration.

One scholar offered an experimental proposal: publishers could establish a separate online journal devoted to fully disclosed studies with different degrees of AI-generated content and data, allowing comparison with entirely human-produced scholarship.[33] Such a platform could be instructive and educational, enabling direct observation of AI’s effects on quality and originality at different levels of involvement and providing an empirical foundation for policy.

A standardized framework is both feasible and beneficial. Its foremost advantage is consistency across journals: similar thresholds reduce the incentive for authors to abandon one journal for another with more permissive standards. Coordinated policies support fair competition in publishing. Standardization also has an educational role, helping mentors teach these principles effectively and enabling students to develop ethical awareness early in their training.

Institutions are already exploring different governance approaches. In June 2026, the Chinese Academy of Social Sciences issued its Basic Standards for AI-Assisted Research at the Chinese Academy of Social Sciences (Trial) for staff and students. It specifies that research may be AI-assisted but must be human-led. It encourages AI use to improve quality and efficiency, promotes truthful disclosure, and identifies permissible involvement in major methods such as experiments, surveys, and archaeological research. It also prohibits AI use for classified data or activities violating national security.[34] However, computational analysis and qualitative research are not yet covered, demonstrating gaps in policy coverage. As noted earlier, blanket prohibitions on reviewer use may also remove an effective means of reference verification, creating new operational obstacles and substantial enforcement costs while seeking to prevent risk.[35] The Communication University of China prohibits submitting AI-generated papers, research outputs, or design proposals as independent academic achievements. All scholarly output must reflect researchers’ principal creative labor. AI may only assist specific stages, and researchers bear full responsibility for the final work’s originality and academic value.[36]

Internationally, Pakistan’s HEC is developing a comprehensive AI policy for higher education. Its latest document, the Framework for the Use of Generative AI Tools in Higher Education Institutions (Draft), remains under review, consultation, approval, and finalization. It explains why policy is necessary, identifying major challenges such as plagiarism, inaccurate or fabricated citations, misinformation, and entire datasets AI-generated. These concerns affect both journal editors and the teachers and researchers evaluating assignments, research papers, and scholarly work. HEC has therefore invited multiple stakeholders to participate in policymaking with the aim of establishing broadly agreed guidance.[37]

3. Guarding Against Structural Imbalances in the Allocation of AI Resources

AI development increasingly reflects privatization, corporate control, and specialization by domain. Development and deployment are becoming concentrated in closed environments controlled by commercial organizations and oriented toward profit rather than a shared public knowledge ecosystem. This shift changes access and conditions of use while more deeply reshaping research resource allocation. Widening resource gaps are a direct consequence. Researchers in different regions and institutions have increasingly unequal access: some can use the most advanced and expensive commercial tools for high-quality data processing and content generation, while others depend on free or inexpensive open-source alternatives with substantial limitations in functionality, accuracy, and applicability.

The Global North remains far ahead of the Global South in AI development capabilities, computing infrastructure, data resources, and training. Even if technical assistance reduces some linguistic barriers, deeper structural inequalities persist and may intensify as technological divides widen. Without deliberate institutional intervention, AI’s spread may entrench existing inequalities technologically rather than narrow North–South academic gaps. While embracing AI’s benefits, scholars and policymakers must remain alert to this imbalance and explore alternatives involving jointly developed resources, open access, and shared technology.[38]

Conclusion: AI’s Three Roles and a Three-Tier Policy Framework

AI can play three roles in research. First, as an assistant, it helps scholars complete tedious or repetitive basic tasks. Second, as a collaborator, especially for novice scholars and graduate students, it can stimulate thinking and provide feedback, serving an intellectually productive function similar to a reviewer. Third, when scholars lack sufficient knowledge of a topic, it can to some extent provide guidance as a mentor or coach. These roles also carry risks. Scholars may adopt suggestions without verification or critical thought, or may lack the disciplinary knowledge needed to distinguish true from false information. AI chatbots also lack human mood fluctuations and arrogance, potentially making users more comfortable seeking help from them than from people. This gives senior scholars an important responsibility: lead by example in showing students responsible AI use consistent with publishing ethics, while preserving research’s value orientation, originality of thought, and distinctive scholarly perspectives.[39]

At the policy level, the three-tier classification widely adopted by international publishers provides a useful institutional starting point. The first tier comprises assistance such as grammar checks and text editing that does not affect research content or accuracy; disclosure may be unnecessary or optional. The second covers uses affecting content accuracy and reproducibility, including image generation, literature review assistance, data analysis, and coding. These must be explicitly disclosed, with tool names, versions, and usage details. The third comprises uses undermining integrity, such as plagiarism and fabricated citations or data, which must be strictly prohibited.

Another issue requires careful consideration: should journals tell reviewers that AI was used in writing a manuscript? If reviewers view AI-assisted writing negatively, disclosure may expose the manuscript to discrimination and undermine the purpose of the requirement. Journals allowing AI-assisted manuscripts and requiring disclosure should therefore remind reviewers that the permission reflects careful deliberation and must not become a basis for prejudice. Conversely, journals that do not require disclosure or that prohibit AI use in writing should communicate their policies clearly and emphasize that judgments must rest on scholarly quality rather than whether AI was used.

Disclosure also carries risks of dishonesty. It should be treated as a binding scholarly declaration, comparable to sworn testimony, with authors held accountable for false statements. Submission systems can include checkboxes requiring authors to report AI use before submitting. Distinguishing routine assistance from uses requiring disclosure helps clarify and encourage legitimate, innovative applications, avoiding the simple stigmatization of AI as a threat to integrity. This supports healthy, ethical use while maintaining vigilance against misuse. AI is like fire: it can destroy your house, but it can also provide you with light, warmth, and cooked food. Its value and harm depend on how humans use it.

AI offers us an opportunity to reflect on excellent research. Meaningful scholarship lies in expressing the essence of human intelligence and humanistic values rather than accumulating vast datasets or elaborate prose. For researchers without research-assistant teams, AI can accelerate access to and understanding of literature, process large amounts of data, perform coding, and provide feedback when needed, allowing researchers to pay more attention to the original academic mission of developing or testing theories and methods. Although dishonest scholars may misuse these tools and pollute the research environment, AI itself should not be blamed. When scholars learn to use AI, journals establish reasonable frameworks, publishing integrity becomes a shared norm, and violations receive appropriate sanctions, AI can achieve its greatest value: combining human and artificial intelligence to produce better scholarship.

Responsible editor: Zhang Su.

Author’s Note

This article builds on the roundtable “Publishing Ethics in the Age of Artificial Intelligence: Challenges and Questions,” organized by Online Media and Global Communication (OMGC) on June 27, 2026, at Shanghai International Studies University’s Ninth Forum on Image Studies and Global Communication, and supplements the discussion with additional material. The event was part of OMGC’s fifth-anniversary activities. Participants included editors of English-language academic journals from China, the United States, and Pakistan, and members of OMGC’s editorial board:

• Louisa Ha: founding editor-in-chief, OMGC; Distinguished Research Professor, Bowling Green State University, U.S.A.

• Ke Guo: co-editor-in-chief, OMGC; professor, Shanghai International Studies University, China.

• Peiqin Chen: co-editor-in-chief, OMGC; professor, Shanghai International Studies University, China.

• Ji Pan: associate editor, OMGC; professor, Fudan University, China.

• Xiaomeng Li: managing editor, OMGC; assistant research fellow, China Center for Global Public Opinion Research, Shanghai International Studies University, China.

• Abida Ashraf: editor, Journal of Media & Culture Studies; director, Institute of Communication Studies, University of the Punjab, Pakistan.

• Bushra Hameedur Rahman: editorial board member, OMGC; dean, School of Media and Mass Communication, Beaconhouse National University, Pakistan.

• Zhi Li: editorial board member, OMGC; professor, School of Television, Communication University of China.

• Lei Guo: associate editor, Communication and Change; professor, School of Journalism, Fudan University, China.

• Siyue Li: associate editor, Communication and the Public; associate dean, College of Media and International Culture, Zhejiang University, China.

Notes

[1] ZHOU Lingyan and TIAN Zhengzheng, “A Study of International Academic Publishers’ AIGC Governance Policies,” Publishing & Printing, online-first publication, August 2026.

[2] De Gruyter Brill, “AI-Policy for Authors,” https://www.degruyterbrill.com/publishing/for-authors/author-policies/artificial-intelligence, accessed August 15, 2026.

[3] Remarks by Ke Guo at the roundtable “Publishing Ethics in the Age of Artificial Intelligence: Challenges and Questions,” Ninth Forum on Image Studies and Global Communication, Shanghai International Studies University, June 27, 2026.

[4] SAGE, “Artificial Intelligence Policy,” https://www.sagepub.com/journals/publication-ethics-policies/artificial-intelligence-policy, accessed August 15, 2026.

[5] Taylor & Francis, “AI Policy,” https://taylorandfrancis.com/our-policies/ai-policy/, accessed August 15, 2026.

[6] Springer Nature, “Editorial Policies,” https://www.springernature.com/gp/policies/editorial-policies/, accessed August 15, 2026.

[7] AEJMC, “Paper Competition—AEJMC Policy on Ethical Use of Artificial Intelligence (AI),” https://www.aejmc.org/aejmc-events/conference/paper-competition, accessed August 15, 2026.

[8] Remarks by Lei Guo at the roundtable “Publishing Ethics in the Age of Artificial Intelligence: Challenges and Questions,” Ninth Forum on Image Studies and Global Communication, Shanghai International Studies University, June 27, 2026.

[9] “Initiative on Regulating AI Use in Academic Papers,” Game Studies, February 4, 2026, https://mp.weixin.qq.com/s/unL_CXIvltSSOvVXoNqR6A, accessed August 15, 2026.

[10] “CNKI Statement on the Handling of Papers Naming Artificial Intelligence (AI) as an Author,” CNKI, July 15, 2026, https://mp.weixin.qq.com/s/eQ9QEai1HsDzlQOfolZWHg, accessed August 15, 2026.

[11] Remarks by Ke Guo at the roundtable “Publishing Ethics in the Age of Artificial Intelligence: Challenges and Questions,” Ninth Forum on Image Studies and Global Communication, Shanghai International Studies University, June 27, 2026.

[12] Remarks by Bushra Hameedur Rahman at the roundtable “Publishing Ethics in the Age of Artificial Intelligence: Challenges and Questions,” Ninth Forum on Image Studies and Global Communication, Shanghai International Studies University, June 27, 2026.

[13] De Gruyter Brill, “AI-Policy for Authors,” https://www.degruyterbrill.com/publishing/for-authors/author-policies/artificial-intelligence, accessed August 15, 2026.

[14] Remarks by Bushra Hameedur Rahman at the roundtable “Publishing Ethics in the Age of Artificial Intelligence: Challenges and Questions,” Ninth Forum on Image Studies and Global Communication, Shanghai International Studies University, June 27, 2026.

[15] Remarks by Ke Guo at the same roundtable, June 27, 2026.

[16] Remarks by Ke Guo at the same roundtable, June 27, 2026.

[17] Jhean Aguilar-Bueno, Robert A. Alvarado-Lugo, and Jean P. Lujan-Leon, “Evaluating the Influence of Human and Non-Human Web Traffic on Academic Repository Analytics,” paper presented at the 2025 IEEE Colombian Caribbean Conference (C3), Santa Marta, Colombia, September 17–20, 2025.

[18] Hannah Heckner Swain, “Bots, Crawlers, And What They Mean For Your Platform,” Silverchair, March 24, 2026, https://www.silverchair.com/news/understanding-ai-traffic-recap/, accessed August 15, 2026.

[19] Zheng, H., and H. Zhan, “Science Behind a Paywall: Restricted Access Limits the Promise of Artificial Intelligence,” Learned Publishing, Vol. 39, No. e2059, https://doi.org/10.1002/leap.2059.

[20] Stephanie Lovegrove Hansen, “The User Has Changed. Has Scholarly Publishing?” The Scholarly Kitchen, May 27, 2026, https://scholarlykitchen.sspnet.org/2026/05/27/the-user-has-changed-has-scholarly-publishing/, accessed August 15, 2026.

[21] Remarks by Bushra Hameedur Rahman at the roundtable “Publishing Ethics in the Age of Artificial Intelligence: Challenges and Questions,” Ninth Forum on Image Studies and Global Communication, Shanghai International Studies University, June 27, 2026.

[22] Remarks by Abida Ashraf at the same roundtable, June 27, 2026.

[23] Remarks by Bushra Hameedur Rahman at the same roundtable, June 27, 2026.

[24] Remarks by Bushra Hameedur Rahman at the roundtable “Publishing Ethics in the Age of Artificial Intelligence: Challenges and Questions,” Ninth Forum on Image Studies and Global Communication, Shanghai International Studies University, June 27, 2026.

[25] Remarks by Siyue Li at the roundtable “Publishing Ethics in the Age of Artificial Intelligence: Challenges and Questions,” Ninth Forum on Image Studies and Global Communication, Shanghai International Studies University, June 27, 2026.

[26] Remarks by Abida Ashraf at the roundtable “Publishing Ethics in the Age of Artificial Intelligence: Challenges and Questions,” Ninth Forum on Image Studies and Global Communication, Shanghai International Studies University, June 27, 2026.

[27] Remarks by Bushra Hameedur Rahman at the same roundtable, June 27, 2026.

[28] PU Lifang, “The Value, Dilemmas, and Implementation Pathways of AI-Empowered High-Quality Development of Academic Journals,” Journal of Henan University (Social Sciences), 2026, No. 4.

[29] Remarks by Abida Ashraf at the roundtable “Publishing Ethics in the Age of Artificial Intelligence: Challenges and Questions,” Ninth Forum on Image Studies and Global Communication, Shanghai International Studies University, June 27, 2026.

[30] Remarks by Bushra Hameedur Rahman at the same roundtable, June 27, 2026.

[31] Remarks by Bushra Hameedur Rahman at the roundtable “Publishing Ethics in the Age of Artificial Intelligence: Challenges and Questions,” Ninth Forum on Image Studies and Global Communication, Shanghai International Studies University, June 27, 2026.

[32] Remarks by Ji Pan at the roundtable “Publishing Ethics in the Age of Artificial Intelligence: Challenges and Questions,” Ninth Forum on Image Studies and Global Communication, Shanghai International Studies University, June 27, 2026.

[33] Remarks by Ji Pan at the same roundtable, June 27, 2026.

[34] Chinese Academy of Social Sciences, “Basic Standards for AI-Assisted Research at the Chinese Academy of Social Sciences (Trial),” official website of the Chinese Academy of Social Sciences, July 9, 2026, http://www.cass.net.cn/tupianxinwen/202607/t20260709_6058415.shtml, accessed July 12, 2026.

[35] WANG Dandan, DONG Yaqi, and YANG Shan, “A Comparison of the ‘Checklist’ Control and ‘Instruction Manual’ Guidance Models in Generative AI Use Disclosure Policies and Their Implications,” Acta Editologica, 2026, No. 4.

[36] Remarks by Zhi Li at the roundtable “Publishing Ethics in the Age of Artificial Intelligence: Challenges and Questions,” Ninth Forum on Image Studies and Global Communication, Shanghai International Studies University, June 27, 2026.

[37] Remarks by Abida Ashraf at the same roundtable, June 27, 2026.

[38] Remarks by Ke Guo at the roundtable “Publishing Ethics in the Age of Artificial Intelligence: Challenges and Questions,” Ninth Forum on Image Studies and Global Communication, Shanghai International Studies University, June 27, 2026.

[39] Colin Campbell, “Did you write this? Rethinking authorship in the age of AI,” Journal of Advertising Research, Vol. 66, No. 2, 2026, pp. 185–189.

Publishing_Ethics_AI_English_Translation.pdf