Editorial: Generative AI, human authorship and the transformation of scholarly communication
Year: 2026
Type: editorial
Abstract: The rapid diffusion of generative artificial intelligence (GenAI) tools into academic workflows represents a structural transformation in scholarly communication, rather than a transient technological trend (Dwivedi, Malik, Hughes, & Albashrawi, 2024). Similar concerns have been raised in recent editorials in management journals (Arsenyan & Piepenbrink, 2024), which highlight how digital transformation, AI and internationalization pressures are reshaping editorial standards, ethical expectations and scholarly practices (Dwivedi et al., 2023, 2024; Pinochet, Schnaider, & Carvalho, 2025). Unlike conventional software used for editing, formatting or reference management, large language models (LLMs) actively participate in text generation and rhetorical structuring (Brown et al., 2024), thereby challenging long-standing assumptions regarding authorship, intellectual responsibility and research integrity (Dwivedi et al., 2024). This phenomenon reflects what some scholars have characterized as an instant research paradigm, in which GenAI systems potentially conduct multiple steps of the research process simultaneously, raising fundamental questions about scholarly authenticity and human agency in knowledge production.Academic journals are urged to address this new scenario and develop new policies for the use of GenAI tools. Editorial responses to this transformation must engage with the ethical foundations that sustain trust, accountability and transparency in scientific knowledge production (Gatrell, Muzio, Post, & Wickert, 2024). This editorial examines how GenAI reshapes academic writing practices and discusses the ethical boundaries, responsibilities and governance principles required to preserve research integrity. At stake is not only compliance with emerging editorial policies but also the preservation of human authorship as the epistemic condition that sustains interpretation, accountability and legitimacy in scholarly communication.In this context, editorials play a distinctive role in scholarly governance as normative reference points that articulate values, frame ethical expectations and reduce interpretive ambiguity within the scholarly community. By clarifying principles rather than prescribing rigid rules, editorials contribute to a shared ethical understanding among authors, reviewers and editors, addressing the need for discipline-specific guidance, which has become increasingly evident across academic communities (Delios, Li, Schotter, & Vrontis, 2024; Delios, Tung, & Van Witteloostuijn, 2025).Recent reviews of research writing practices indicate that the ethical challenges associated with GenAI are intensified by the absence of harmonized policies across journals and publishers (Okina, 2024). Rather than operating within a coherent normative framework, authors are often confronted with fragmented, ambiguous or incomplete guidance regarding acceptable AI use in academic writing (Ngwenyama & Rowe, 2024). This lack of alignment creates ethical uncertainty and shifts responsibility from institutional governance to individual interpretation, increasing the likelihood of inconsistent disclosure practices and undermining shared standards of academic integrity (Bolaños, Salatino, Osborne, & Motta, 2024).Recent debates in publication ethics have highlighted that the central issue is not whether AI should be used in academic writing, but how, to what extent and under whose responsibility (Dwivedi & Helal, 2026). The presence of hallucinations and AI-generated confident but inaccurate assertions (Thomas, Romasanta, & Pujol Priego, 2026), particularly, threatens the epistemic integrity of academic research, and such errors are especially problematic when AI systems are applied to text analysis and literature synthesis without rigorous human oversight. Evidence from recent systematic and rapid reviews suggests that the ethical dilemma surrounding AI use in academic writing is not whether such tools should be employed but how ethical boundaries are interpreted and operationalized in practice. In the absence of clear editorial guidance, researchers often rely on personal judgment, disciplinary norms or perceived editorial tolerance, leading to uneven ethical reasoning and reinforcing the need for clearer governance structures that support the responsible and transparent use of AI technologies. In this sense, the ethical challenge is not exhausted by asking whether GenAI is present in academic writing. Still, by distinguishing acceptable, sensitive and unacceptable forms of use according to the research stage, the degree of human oversight and the risks posed to intellectual authorship.Viewing it through this lens, ethical engagement with GenAI in academic writing should not be understood as a binary choice between acceptance and prohibition, but as a process of progressive ethical maturation (Delios et al., 2024). What is required, therefore, is not merely procedural compliance but the preservation of cognitive responsibility: human authors must remain the primary agents of framing, interpretation, validation and judgment throughout the scholarly process (Tosi, 2025).As researchers integrate AI tools into scholarly workflows, the central challenge is to develop reflective, responsible and transparent practices that evolve with technological capabilities. Editorial guidance must support this trajectory by encouraging a conscious and proportionate use of AI, rather than treating ethical compliance as a static or purely procedural requirement. This approach acknowledges both the opportunities that AI offers for supporting researchers, particularly non-native English speakers who benefit from language refinement assistance, and the genuine epistemological risks posed by uncritical adoption (Barman, Wood, & Pawlowski, 2024).Good practice guidelines suggest that ethical parameters for AI should not be generic; rather, they should vary with technology and, above all, with the stage of the scientific process in which the tool is applied. This perspective shifts governance away from a binary logic of permit versus prohibit, toward defining proportional uses and safeguards for each phase.AI can contribute to specific tasks in scientific writing, such as identifying research gaps, assisting with content drafting and structuring, extracting data for evidence synthesis, supporting data analysis and interpretation and generating figures and tables (Dwivedi & Helal, 2026). A more precise editorial interpretation also requires recognizing that AI assistance is not ethically uniform. Technical assistance, organizational support and heuristic exploration may be acceptable when explicitly disclosed and critically reviewed. In contrast, substantive substitutions in hypothesis formulation, data interpretation or conclusion writing directly threaten the boundary between assistance and authorship. However, such support must not undermine intellectual authorship: the tool should assist rather than replace the researcher's work, and any AI-assisted text must undergo critical review and content validation, including verification of cited references (Dwivedi & Helal, 2026; Dwivedi et al., 2024).A critical ethical distinction that must be made explicit in editorial policies is the difference between AI-generated and AI-assisted writing. AI-generated writing refers to situations in which AI systems autonomously produce substantial portions of academic texts without meaningful human intellectual control. Such practices raise serious concerns about the attribution of intellectual agency, the erosion of accountability and the misrepresentation of scholarly contributions, particularly when left undisclosed (Gatrell et al., 2024). The fundamental issue is that autonomous AI text generation produces content that, while superficially authoritative, is susceptible to factual errors and misrepresentations (Kulkarni et al., 2024); in extreme cases, it may reference research papers that do not exist or provide incorrect identifiers of unrelated work, presenting what amounts to scholarly hallucinations (Dwivedi et al., 2024).Empirical studies indicate that this distinction is not always consciously perceived by authors (Brown et al., 2024). In many cases, researchers cognitively reframe AI-assisted contributions as extensions of their own thinking, particularly when generative systems are used simultaneously to structure arguments or refine expressions. This cognitive reframing reduces the perceived need for disclosure. It blurs the boundary between assistance and substitution, posing a subtle but significant challenge to traditional notions of authorship and intellectual ownership. Research has demonstrated that human reviewers struggle to reliably distinguish between human-authored and AI-generated abstracts (Kulkarni et al., 2024), further complicating editorial oversight.Ethical analyses consistently emphasize that authorship in scholarly publishing is inextricable from accountability (Hosseini, Rasmussen, & Resnik, 2024). From this perspective, the production of text alone does not constitute authorship unless accompanied by the capacity to assume moral, legal and epistemic responsibility for the content. Because AI systems lack agency, intentionality and accountability, their contributions, regardless of sophistication, cannot meet the normative criteria for scholarly authorship. Scholarly writing is not merely the production of grammatically coherent text; it is an interpretive and moral act through which authors assume responsibility for meaning, evidence and intellectual consequences. Major journal publishers have explicitly articulated this principle; for instance, Science journals have clarified that text generated by ChatGPT or other AI tools cannot be used in scholarly work, nor can such tools produce figures or graphics, and that violations constitute scientific misconduct (Lindebaum & Fleming, 2024).In contrast, AI-assisted writing involves the use of generative systems as supportive tools in human-led research processes (Pinochet, Moreira, Fávero, Santos, & Pardim, 2023). In such cases, AI may contribute to linguistic refinement, structural organization and preliminary drafting (Gatrell et al., 2024). Simultaneously, human authors retain full responsibility for conceptual development, theoretical coherence, methodological rigor and final editorial decisions. From an ethical standpoint, this distinction is fundamental: the acceptability of AI depends less on the technology itself and more on the preservation of human judgment, critical oversight and explicit disclosure (Thomas et al., 2026). This phenomenon has been described as the AI ghostwriter effect (Draxler et al., 2024), referring specifically to AI-assisted authorship rather than fully AI-generated content, in which generative systems support idea formulation or textual organization. At the same time, human authors retain intentionality, judgment and responsibility for the final text, remaining psychologically invisible as external agents. Unlike intentional misconduct, this effect reflects altered perceptions of authorship rather than strategic concealment, suggesting that ethical governance must address not only formal rules but also the cognitive dynamics that shape how AI involvement is interpreted and justified.Beyond individual instances of disclosure failure, a broader systemic risk emerges when GenAI becomes normalized and psychologically invisible in academic writing practices. As AI-mediated structuring, phrasing and ideation become routine, the risk of homogenization in scholarly argumentation, rhetorical patterns and epistemic styles increases. Such normalization may subtly reshape academic discourse, raising concerns not only about authorship and accountability but also about intellectual diversity, originality and the depth of scholarly reasoning. This process may reinforce automation bias and reduce the epistemic vigilance and reflective scrutiny required for rigorous scholarship, especially when scholars begin to rely on fluent AI-generated outputs without sustained verification (Heersmink, 2024; Wong, 2024). Because language models rely on existing literature, heightened attention is needed to avoid the literal reproduction of passages from other authors. This concern is relevant even in the absence of intentional misconduct, as reliance on training corpora may induce textual reuse and unnoticed standardization (Ignjatović & Stevanović, 2023). Recent discussions in the information science and education literature suggest that excessive reliance on GenAI systems may weaken cognitive engagement, critical thinking and creative processes that are central to academic writing (Helal et al., 2025; Kumar & Gunn, 2025; Lindebaum & Fleming, 2024). While generative tools may accelerate drafting and ideation, sustained dependence risks bypassing the reflective, evaluative and iterative stages traditionally associated with scholarly reasoning, raising concerns that extend beyond authorship to the epistemic quality and intellectual depth of academic work.From an editorial standpoint, the ethical acceptability of AI-assisted writing depends on preserving meaningful human control throughout the research process (Thomas et al., 2026). Reviews of editorial policies have indicated that GenAI may be ethically integrated as a support mechanism for language refinement or structural organization, provided that human authors retain responsibility for conceptual framing, methodological decisions and interpretive judgment. This reinforces the principle that ethical evaluation should focus on responsibility and oversight rather than the mere presence of GenAI tools.In practice, transparency is strengthened when authors maintain a minimal provenance record of GenAI assistance (tool name, model/version, date of use and a brief description of prompts or interactions) so that the origins and reliability of AI-assisted passages can be evaluated if questions arise (Herhausen, Ludwig, Abedin, Haque, & De Jong, 2025). Such disclosure is strongest when it identifies not only the tool and model/version used, but also the purpose of use, the degree of intervention and the stage of the manuscript or research process affected, thereby creating conditions for traceability, ethical auditability and informed editorial judgment.This approach acknowledges that prompt engineering and the specific formulation of queries to LLMs significantly influence outputs, making the reproducibility of LLMs essential for scholarly verification (Wobst & Lueg, 2025). For editorial purposes, this differentiation provides a practical and principled framework for evaluating submissions, guiding authors and safeguarding research integrity without impeding legitimate methodological innovations (Herhausen et al., 2025).Within AI-mediated scholarly communication, peer reviewers play a pivotal yet often underarticulated role (Bechky & Davis, 2025). In this evolving landscape, reviewers should be understood not as technological detectors of hidden AI use but as evaluators of coherence, transparency, proportionality and human intellectual responsibility in manuscripts shaped by AI-mediated workflows. Beyond evaluating technical rigor and theoretical contributions, reviewers increasingly serve as interpreters of transparency disclosures and assessors of intellectual coherence in AI-assisted manuscripts. Their role is not to police authorship through technological suspicion but to exercise informed human judgment, contextual sensitivity and ethical discernment, reinforcing the centrality of human oversight within the publication process. This requires reviewers to develop new competencies in recognizing the variable performance profiles of LLMs across different tasks, understanding that AI systems exhibit jagged competencies, where they may perform reliably on certain tasks while failing significantly on others (Thomas et al., 2026).Editorial policies that attempt to ban the use of GenAI in academic writing face three structural limitations. First, such bans are largely unenforceable given the current and foreseeable limitations of AI detection technologies (Andrews, Fainshmidt, & Gaur, 2026). The reliability of detection tools remains contested, and researchers have raised significant concerns about false positives and methodological opacity in AI detection mechanisms. Second, they may inadvertently incentivize covert or undisclosed use, thereby weakening transparency and undermining ethical self-regulation. Research demonstrates that AI disclosure paradoxically reduces reader in regardless of potentially encouraging authors to AI use rather than it policies risk existing in scholarly communication, particularly for researchers in non-native (Ngwenyama & Rowe, emerging evidence that GenAI a meaningful role in supporting non-native English speakers and with by assisting with language refinement, structuring and From this perspective, editorial policies risk structural in scholarly particularly in academic where to linguistic and editorial support has been uneven et al., 2024). The is particularly given that academic publishing English and on AI assistance may inadvertently reinforce existing in a practical standpoint, meaningful human control a of critical validation of AI-generated outputs, verification of and an to where how outputs generated to support reproducibility and transparency, and a to avoid technological and to that may systematic errors or et al., 2024; et al., 2025; et al., 2024).A rapid review of evidence suggests that editorial policies are to be in AI in research writing. A more editorial is not but proportional forms of AI assistance should be according to their and epistemic rather than into a binary logic of or are to given the limitations of detection and may covert use ethical In contrast, policies in proportionality and responsibility are more with the principles of research ethics and to integrity in scholarly Recent editorials have articulated the for this rather than AI as an external and ethical should AI as a tool whose legitimacy depends on transparent use within research processes et al., editorial control and ethical governance play essential and in research integrity 2023). While compliance provide the structure through editorial disclosure and procedural ethical governance beyond by human judgment and (Gatrell et al., 2024). In the of AI-assisted writing, compliance such as transparent disclosure of GenAI tools used each stage of the research process and clear authorship policies that that responsibility with human authors, should as of ethical rather than as integrate formal compliance with principled editorial judgment and contextual reinforcing accountability without human ethical principles into practice requires guidelines that can be consistently across editorial workflows. Editorial should clear that the of AI assistance acceptable within their disciplinary context, from language refinement and to structural organization and preliminary should explicit guidance on disclosure in or in disclosure in an AI or integrated into and the of on the and of AI By parameters journals can reduce interpretive ambiguity and provide authors with transparent expectations regarding acceptable practice, compliance through rather than through assumptions or editorial on institutional guidelines et al., 2026), uses of GenAI across research stages and the safeguards required to preserve human accountability and research of guidelines should training for both editorial and peer reviewers to that the distinction between AI-generated and AI-assisted and can reliably whether disclosure with the intellectual of human authors. disclosure that authors can to their specific research workflows, thereby the cognitive of disclosure while that essential tool name, specific and of use are consistently Such standardization does not but rather a of transparency that can be to disciplinary and reviewers to informed of authorship and intellectual formal disclosure journals should editorial that explicitly AI as a legitimate support tool within research while simultaneously reinforcing that human authors full responsibility for methodological rigor and intellectual This involves editorial communication, through guidelines and which AI governance not as or prohibition, but as a practice of and editorial can this by transparent about how editorial workflows that the academic is not from technological engagement but rather in critical and proportionate practice reinforce three directly points for AI tools should not be as as authorship the of authors should the specific editorial policies AI and when AI is used, it should be the purpose and the of the manuscript in which it an editorial perspective, a of responsible disclosure is more and than police authorship through detection et al., 2024). Research suggests that on AI detection technologies may in editorial about false methodological transparency and uneven disclosure and weaken the legitimacy of reinforce the that the ethical governance of AI-assisted writing is more when in proportionality and rather than technological studies have also highlighted that reliance on tools as a primary governance mechanism may ethical consequences. to false lack of methodological transparency and inconsistent can in editorial processes and disclosures (Wobst & Lueg, 2025). support a away from toward governance models that ethical judgment, accountability and AI detection tools should not be as for editorial judgment. The variable performance of LLMs across tasks suggests that they are to human analysis rather than it a principle that to editorial processes (Thomas et al., of transparent disclosure is the of the ethical use of GenAI in academic writing. should not be as a but as a substantive ethical practice that reviewers and to the and attribution of authorship, reliability and intellectual of scholarly (Gatrell et al., 2024). disclosure should be proportionate to the degree of AI distinguishing between linguistic assistance and substantive contributions to content generation or the structuring of the indicate that the ethical challenges associated with GenAI in academic writing cannot be fully through formal disclosure or detection tools they editorial of with the disciplinary and contextual that shape how GenAI use is perceived and in scholarly practice. such ethical guidelines risk but policies that articulate expectations regarding including how and to what extent AI use should be can integrity without the ethical of GenAI in academic writing requires editorial rather than responses (Delios et al., 2025). and among publishers and editorial can shared principles while disciplinary in As GenAI becomes increasingly in academic practice, the of scholarly less on technological detection and more on human ethical judgment and editorial In the integrity in AI-mediated not only on formal but also on institutional critical digital and the of shared ethical competencies across authors, reviewers and editorial is to that this editorial does not to disclosure or replace formal editorial purpose is to articulate ethical principles that support interpretive judgment, responsible practice and reinforce human accountability in AI-assisted academic writing. By the editorial to existing governance rather than broader is to support a of editorial governance in which technological remains with scientific interpretive responsibility and the of human arguments indicate that the ethical challenges associated with AI-assisted research writing cannot be through rules or technological (Bechky & Davis, 2025). they for integrated editorial that clear guidance, full disclosure and shared responsibilities among authors, and such ethical standards risk remaining but in the face of evolving generative technologies. in scholarly in the of GenAI is less of a technological challenge than an ethical and editorial
Source: Revista de Gestão
Language: English
Subfield: History and Philosophy of Science
Field: Arts and Humanities
Domain: Social Sciences
Open Access status: diamond
OpenAlex ID: W7160855111