small_720x360 (2).jpg

thumbnail_720x360 (2).jpg

720x360 (2).jpg

If a researcher consults AI before a mentor, colleague, a paper, or even before their own notes, what does that make AI?Recent <a href="https://cdn.openai.com/pdf/f4b4a5da-b2de-418d-9fcd-6b293e9dc157/oai_ai-as-a-scientific-collaborator_jan-2026.pdf" target="_blank" rel="noopener noreferrer">estimates</a> indicate that 1.3 million scientists generate over 8 million weekly queries on ChatGPT, many tied to core scientific reasoning and technical problem-solving. The scale is striking! But what is more notable is that AI is no longer peripheral to research workflows. It is becoming a part of the moment where knowledge is grasped, refined, and formed.That shift–from assistance to active participation, is what turns a tool into infrastructure.<h2>The Rise of Invisible Infrastructure</h2>Infrastructure is most powerful when it is silently embedded. Electricity, cloud computing, and the internet have all followed this trajectory, moving from visible systems to invisible backbones. Historically, research tools like statistical software, reference managers, or search engines were clearly bounded in function and supported discrete steps in the research lifecycle. AI systems, particularly large language models (LLMs), are fundamentally different, as they can perform a range of functions.This breadth has transformed AI into what many researchers now treat as a collaborator. The <a href="https://www.elsevier.com/insights/confidence-in-research/researcher-of-the-future" target="_blank" rel="noopener noreferrer">Researcher of the Future</a> report by Elsevier indicates that a significant proportion of researchers are already using AI tools across tasks like literature review (51%), writing (38%), and data analysis (38%). This indicates how strongly (but quietly) AI is being embedded into routine tasks and processes.Reporting from <a href="https://www.technologyreview.com/2026/01/27/1131793/openais-latest-product-lets-you-vibe-code-science/" target="_blank" rel="noopener noreferrer">MIT Technology Review</a> highlights how newer AI systems are collapsing multiple stages of the research process into a single interface. OpenAI’s newly introduced research workspace, designed to integrate conversational AI directly into writing, coding, and collaboration environments is a strong example.What emerges here is not just a more efficient workflow, but one where AI is continuously present for shaping decisions at multiple stages, rather than assisting at isolated points. This model, described as enabling scientists to “vibe code” their work, seems to be a transition toward AI systems that are not just assisting tasks, but orchestrating workflows, reducing friction between ideation and execution.This raises a critical question: When AI becomes always available, seamlessly integrated, and cognitively interactive; does it cease to feel like a tool and begin to function as an infrastructure for thought. This invisibility, however, is precisely what introduces and/or amplifies the associated risks.<h2>Hidden Risks of Invisible AI Dependence &nbsp;</h2><h3>1. Unseen Knowledge Mediation&nbsp;</h3>As a researcher, you may find yourself relying on AI-generated interpretations without fully interrogating underlying sources. For instance, researchers frequently publish papers with erroneous citations without checking, as seen in <a href="https://mental.jmir.org/2025/1/e80371" target="_blank" rel="noopener noreferrer">analyses</a> of 176 citations where 20% were wholly invented.&nbsp;Furthermore, a separate <a href="https://arxiv.org/abs/2602.06718?" target="_blank" rel="noopener noreferrer">study</a> identified a “verification gap” as 41.5% of researchers copy LLM-generated citations without checking and 76.7% of reviewers do not thoroughly verify references.&nbsp;Another particularly telling <a href="https://fortune.com/2026/01/21/neurips-ai-conferences-research-papers-hallucinations/" target="_blank" rel="noopener noreferrer">case</a> comes from the NeurIPS 2025 conference, where over 100 fabricated citations were identified across accepted papers, despite expert peer review. The fact that such errors passed through one rigorous review systems makes it deeply concerning.This indicates when AI mediates knowledge early in the process, errors are not corrected but inherited. Over time, this creates the risk of citation laundering, where fabricated or distorted references enter the scholarly record and are unknowingly reused.<h3>2. Silent Propagation of Bias and Hallucination</h3>Bias and hallucination in AI systems are well documented. What changes in an infrastructure context is not their existence, but their scale, persistence, and invisibility.&nbsp;A <a href="https://pmc.ncbi.nlm.nih.gov/articles/PMC10488525/" target="_blank" rel="noopener noreferrer">study</a> examining the validity of ChatGPT in identifying literature found that 68% of the links in the references were incorrect and 31% were fabricated. Even advanced models continue to <a href="https://mental.jmir.org/2025/1/e80371" target="_blank" rel="noopener noreferrer">fabricate</a> references, particularly in specialized domains where training data is sparse. Furthermore, LLM <a href="https://arxiv.org/abs/2510.25378" target="_blank" rel="noopener noreferrer">accuracy</a> is highly reliant on the how frequently the papers are cited; which implies that hallucinations increase when papers are less cited and domains are niche.&nbsp;Notably, these errors are not always random. They follow patterns like combining real author names with incorrect titles, fabricating DOIs that appear structurally valid, and blending multiple sources into a single “plausible” reference. This makes them exceptionally difficult to detect, especially at scale.When AI errors are embedded into workflows, they do not remain isolated, they propagate across documents, disciplines, and decision systems.<h3>3. Shifting Authority from Human Expertise to AI Reasoning</h3>As AI becomes more integrated into research workflows, it begins to influence&nbsp;how you validate, prioritize, and trust information.This shift manifests in subtle but significant ways, especially when you:&nbsp;<ul style="list-style-type:disc;"><li>use AI outputs as starting points for reasoning</li><li>treat AI-generated summaries as proxies for literature reviews</li><li>seek confirmation from AI rather than from primary sources or peers</li></ul>Over time, this can shift how you evaluate and trust information; subtly moving authority from primary sources toward AI-generated reasoning. Traditionally, authority in scholarly publishing was grounded in peer-reviewed literature, domain expertise, and methodological rigor. Perhaps in the future, it may increasingly be&nbsp;influenced by AI-generated reasoning pathways, which are opaque, non-verifiable, and not always reproducible.&nbsp;<h2>Why These Risks Matter More as AI Becomes “Invisible”</h2>Individually, none of these risks are new. What is new is how they interact under conditions of scale and integration, particularly when:&nbsp;<ul><li>Errors are introduced earlier in the workflow</li><li>Verification is inconsistent or absent</li><li>Outputs are highly persuasive</li><li>Systems are used repeatedly and routinely</li></ul>This creates a form of invisible dependency, and that is the core challenge. The more seamlessly AI integrates into research, the harder it becomes to distinguish between what we know, what we infer, and what the model has suggested.<h2>Why Researchers Still Continue to Rely on AI</h2>Despite the risks, AI adoption in research continues to accelerate for understandable reasons. At a practical level, AI addresses several long-standing pressures in academia like growing literature volume, interdisciplinary complexity, and tight timelines. AI/LLMs offer immediate and tangible outcomes. This prompts them to rely on AI for learning unfamiliar concepts/domains, creating starting points for thinking, and reducing the friction of early-stage ideation. The result is a fundamental trade-off between efficiency and epistemic control.The challenge is not whether you adopt AI or not, it is how well you can calibrate its role. On one side, it supports faster workflows, broader information access, and lesser burden. On the other end, it blurs the reasoning pathways, passes on undetected errors, and erodes critical evaluation.<h2>Practical Considerations for Researchers</h2>If AI is increasingly functioning as an embedded layer within research workflows, the question is no longer whether it should be used, but how its role can be calibrated without compromising epistemic rigor. This requires a shift from passive adoption to deliberate use.1. Maintain rigorous verification practices AI-generated citations, summaries, and interpretations should be treated as provisional. Independent validation against primary sources remains essential, particularly because many errors are plausible and therefore difficult to detect.2. Be intentional about when AI enters the workflow The stage at which AI is introduced matters. Early reliance can shape problem framing, source selection, and reasoning pathways. Engaging with primary materials before consulting AI helps preserve intellectual independence.3. Anchor analysis in primary literature AI outputs should not replace direct engagement with full texts. While summaries accelerate access, they often compress nuance, omit methodological detail, and obscure uncertainty. Sustained interaction with original sources remains critical.4. Treat AI as a support, not a substitute for reasoning AI can assist with structuring ideas or exploring alternatives, but it should not become the primary driver of interpretation or judgment. Maintaining ownership of reasoning processes is essential to preserving scholarly rigor.5. Periodically audit dependence on AI tools As AI becomes habitual, its influence becomes less visible. Reflecting on when and why it is used can help distinguish between productive augmentation and unexamined reliance.6. Ensure transparency and disclosure of AI use Clearly documenting where and how AI has contributed supports reproducibility and accountability. As AI becomes more integrated, structured disclosure practices will be increasingly important in making its role visible.Taken together, these strategies do not argue against the use of AI in research. Rather, they emphasize that as AI becomes more deeply embedded, maintaining rigor will depend less on limiting its use and more on making its role visible, bounded, and critically engaged.&nbsp;

1.3 Million Scientists Ask ChatGPT 8 Million Questions a Week: 6 Strategies for Researchers to Stay in Control

Over 1.3 million scientists ask ChatGPT 8 million questions weekly. Discover the hidden risks of AI in research and 6 practical strategies to maintain control, accuracy, and academic rigor.

AI in Research: 1.3M Scientists Use ChatGPT Weekly - Risks & 6 Control Strategies

Anagha Nair

large_New-Project.jpg

medium_New-Project.jpg

small_New-Project.jpg

thumbnail_New-Project.jpg

New-Project.jpg

With over 1.3 million scientists generating millions of weekly queries on ChatGPT, AI is rapidly becoming embedded in research workflows. While it accelerates discovery and simplifies complex tasks, it also introduces hidden risks like citation errors, bias, and reduced critical evaluation. This article explores how AI is evolving into research infrastructure and outlines six practical strategies to help researchers maintain control, accuracy, and intellectual independence.

thumbnail_340x200 (2).png

340x200 (2).png

small_720x360.jpg

thumbnail_720x360.jpg

720x360.jpg

In academic publishing, productivity has always been easy to measure. Count the papers, track the citations, and the numbers appear to tell a story of progress. But the rise of generative AI (GenAI) is forcing the research community to ask a more uncomfortable question: What if more papers do not necessarily mean better science?<a href="https://news.cornell.edu/stories/2025/12/ai-gives-scientists-boost-cost-too-many-mediocre-papers" target="_blank" rel="noopener noreferrer">Evidence</a> suggests that researchers using AI tools are publishing faster and in greater numbers than before. Yet alongside this surge in output, new concerns are emerging about narrowing research agendas, uneven benefits across researchers, and a growing strain on peer-review systems.&nbsp;<h2>Measured Productivity Gains but Reduced Thematic Diversity</h2>One of the clearest signals of AI’s impact on research productivity comes from a Cornell University <a href="https://news.cornell.edu/stories/2025/12/ai-gives-scientists-boost-cost-too-many-mediocre-papers" target="_blank" rel="noopener noreferrer">study</a> which found that scientists using large language models (LLMs) such as ChatGPT published up to 50% more papers than before.From one perspective, this is a positive development. AI tools can streamline time-intensive tasks like editing manuscripts, identifying relevant citations, or summarizing prior literature. However, increased output does not necessarily translate to deeper or more meaningful scientific progress.<a href="https://www.nature.com/articles/d41586-025-04092-3" target="_blank" rel="noopener noreferrer">Large-scale analyses</a> of scientific literature reveal a striking pattern. Researchers who adopt AI tools tend to publish more papers and receive more citations, but their work increasingly concentrates within narrower thematic areas. Instead of expanding the intellectual frontier, AI-assisted research often deepens existing trajectories.&nbsp;<h2>Unequal Gains in the AI Era</h2>Another dimension of the productivity paradox concerns who benefits from AI-driven productivity.A <a href="https://academic.oup.com/pnasnexus/article/4/2/pgae591/7996465" target="_blank" rel="noopener noreferrer">study</a> published in PNAS Nexus found that the productivity boost associated with AI adoption is 6.4% higher for male researchers compared with female researchers. This suggests that systemic disparities such as differences in access to resources, training, or institutional support may shape how effectively researchers can leverage AI tools.Without careful attention, AI could amplify existing inequities in academia, benefiting groups already advantaged in research ecosystems. But unequal access to productivity gains is only part of the picture. As AI-assisted research output grows, concerns are also emerging about the quality and rigor of that expanding volume of work.<h2>The Rise of “Productivity Without Rigor”</h2>Editors and conference organizers are beginning to confront a more practical consequence of AI-driven productivity. <a href="https://www.nature.com/articles/d41586-025-03967-9" target="_blank" rel="noopener noreferrer">Reports</a> from within the publishing industry describe a noticeable rise in low-quality or partially AI-generated submissions (sometimes referred to informally as “AI slop”). These manuscripts are often linguistically polished but lack methodological rigor or conceptual novelty, placing additional strain on peer-review systems already operating at capacity.This trend also creates significant pressure on peer-review infrastructure. As the volume of submissions rises, reviewers must spend more time filtering out weak or poorly verified work, potentially slowing the evaluation of genuinely novel research.<figure class="image"><img src="https://easydash.enago.com/uploads/image_48b38c5426.png"><figcaption>Disclaimer: The above image is generated using Notebook LM for illustration and understanding purposes.</figcaption></figure>The result is a growing tension: AI <a href="https://www.enago.com/responsible-ai-movement/resources/ai-assisted-writing-scientific-integrity-research" target="_blank" rel="noopener noreferrer">improves</a> writing fluency and productivity, but polished language can obscure weak scientific contributions, making it harder for editors and reviewers to distinguish meaningful research from superficial output.&nbsp;<h2>Rethinking What Productivity Means in Research</h2>The emerging evidence points to a paradox at the heart of AI-enabled research:<ul><li>Researchers can produce more papers faster</li><li>Yet the collective intellectual diversity of science may shrink</li><li>And the systems that evaluate research quality may become overwhelmed</li></ul>None of this suggests that AI is inherently harmful to research. On the contrary, AI tools can democratize access to scientific communication, reduce language barriers, and accelerate routine tasks that previously consumed researchers’ time.However, it strongly suggests that publication counts alone are becoming an increasingly unreliable measure of scientific productivity.If success continues to be measured primarily by publication counts, AI may simply fuel the parti pris in academia: producing more papers to meet institutional expectations. In such a system, AI becomes a powerful amplifier of the “publish-or-perish” culture already shaping research careers.<h2>What Should Researchers Do?</h2>Responsible AI use in science therefore requires a fundamental shift in how research contributions are valued. The question is not how many additional papers AI can help you produce, but whether it enables you to:<ul style="list-style-type:disc;"><li>Ask meaningful scientific questions</li></ul><ul><li>Add novel insights and expand knowledge</li><li>Explore diverse and emerging research areas</li><li>Strengthen methodological rigor and transparency</li><li>Enhance equity in global research participation</li></ul>The real risk is not that AI will change how research is produced, because it already has. The risk is that without clear standards, it will quietly redefine what counts as good science. But we understand that there is a gap in how these standards are perceived by researchers.&nbsp;This gap is precisely what <a href="https://www.enago.com/responsible-ai-movement" target="_blank" rel="noopener noreferrer">Enago’s Responsible Use of AI (RUAI)</a> initiative seeks to address. By collating clear guidelines, standardized disclosure practices, and guiding stronger human oversight, it brings much-needed structure to an ecosystem currently marked by inconsistency and uncertainty. It recognizes that responsible AI use cannot be left to individual judgment alone, rather requires coordinated action across researchers, publishers, and institutions.As AI continues to accelerate research output, the question is no longer whether we can produce more, but whether we can do so without compromising integrity, diversity, and trust. Responsible use of AI offers a path forward, not just to manage AI adoption, but to shape it deliberately, so that scientific progress is not only faster, but fundamentally better. <a href="https://www.enago.com/responsible-ai-movement/workshops" target="_blank" rel="noopener noreferrer">Let us know</a> if you need assistance in understanding publisher and university policies or in applying best practices for the responsible use of generative AI through the RUAI initiative.

50% More Papers, Less Science? What Does This Mean For You as a Researcher?

Explore why AI adoption boosts publication rates but narrows research diversity. Discover the productivity paradox, equity gaps, and responsible AI use in academia.

50% More Papers, Less Science? AI's Research Paradox

Generative AI tools are helping researchers publish 50% more papers—but is quantity driving down quality? This article explores the productivity paradox in academia, examining how AI adoption creates both opportunities and risks for research integrity, equity, and peer-review systems.

thumbnail_340x200.png

340x200.png

small_Article_01 (4).jpg

thumbnail_Article_01 (4).jpg

Article_01 (4).jpg

Generative AI is no longer experimental in research. It is already embedded in everyday workflows. However, widespread adoption is only a part of the story. What matters more is how researchers are recalibrating expectations, confronting risks, and redefining what responsible AI use requires in practice.The latest <a href="https://www.wiley.com/en-in/about-us/ai-resources/ai-study/" target="_blank" rel="noopener noreferrer">ExplanAItions report</a> from Wiley, based on insights from more than 2,400 researchers globally, captures this shift in operational use and expectations. The findings show a research community that is not only adopting AI rapidly but are also becoming more aware of its limitations and the governance and oversight mechanisms required to use it responsibly.Three insights stand out. And together, they reveal a shift that goes far beyond adoption.<h2>1. AI Adoption has Increased, But the Expectations are Being Recalibrated</h2>The scale of AI-adoption is striking. AI usage among researchers increased significantly from 57% in 2024 to 84% in 2025, with 62% actively using AI in research and publication tasks.&nbsp;Yet the more important insight is what happened alongside this growth: expectations declined.In 2024, researchers believed AI outperformed humans in over half of tested research use cases. In 2025, that figure dropped to less than one-third. This signals a critical shift from hype-driven optimism to experience-driven realism.Interestingly, this pattern aligns with broader academic evidence. <a href="https://arxiv.org/abs/2510.17882" target="_blank" rel="noopener noreferrer">Large-scale analyses</a> of preprints show that GenAI is accelerating workflows and improving productivity, but acting more as a selective catalyst than a universal replacement for human expertise.&nbsp;This is a critical turning point. The responsible use of AI begins when users understand both its capabilities and limits. Unrealistic expectations lead to misuse; calibrated expectations enable meaningful integration.The research community appears to be reaching that inflection point.<h2>2. AI adoption is Outpacing Institutional Policy, Creating a Governance Gap</h2>The second finding reveals a structural gap. While researchers are integrating AI into their workflows, institutional policy infrastructure remains insufficient.According to the report, 57% (more than half) researchers cite lack of training and guidelines as a major barrier for AI use, and only 41% feel their organization provides adequate support. Most notably, 73% say publishers should provide clearer guidance on responsible AI use.&nbsp;This confirms what many in scholarly publishing are already observing: researchers are using AI, but struggling to swim the ethical waters due to missing frameworks. Without clarity on acceptable use, disclosure norms, and verification practices, AI integration risks research integrity rather than strengthening it.This is precisely where publishers, institutions, and research integrity leaders play a critical role.<h2>3. Researchers are Heavily Relying on General-Purpose AI Tools, Raising New Integrity and Transparency Challenges</h2>A structurally significant finding: researchers overwhelmingly rely on general-purpose AI tools rather than research-specific solutions.Nearly 80% use mainstream AI tools like ChatGPT, while only one in four use specialized research-focused AI solutions. Most rely on freely available tools rather than institutionally supported systems. &nbsp;This gap is also fueled by unawareness, as only 11% of researchers on average having heard of the specialized research tools surveyed.&nbsp;While accessibility has accelerated adoption, it has also amplified risks related to accuracy, security, and traceability.. Concerns about hallucinations and inaccuracies are being cited (64%) as a barrier to AI use, as researchers gain more experience with these tools. Moreover, growing concerns about information security and privacy (58%) is likely to hinder greater AI use.These risks are not theoretical as <a href="https://arxiv.org/abs/2512.06705" target="_blank" rel="noopener noreferrer">recent analyses</a> show that AI-assisted writing is increasing rapidly across research publications, while transparency remains extremely rare.<figure class="image image-style-align-left"><img src="https://easydash.enago.com/uploads/ruai_article_1_feb_2026_4b1e8ba8fd.png"><figcaption>Note: The above infographic is generated using NotebookLM for illustrative purposes.</figcaption></figure>Responsible AI adoption requires infrastructure designed specifically for research, not just convenience-driven adoption of general tools. This includes tools and workflows that support transparency, verifiability, and ethical integration into the research lifecycle. Emerging tools, such as <a href="https://www.enago.com/ai-disclosure-statement-generator" target="_blank" rel="noopener noreferrer">Enago’s AI Disclosure Statement Generator</a>, are addressing this need by helping researchers document AI use in a structured and transparent manner, aligned with evolving publisher and institutional expectations.As AI becomes embedded in everyday research workflows, transparency can no longer be treated as optional. It must become a standard part of the research process.<h2>What these Findings Mean for the Future of Responsible AI in Research</h2>Collectively, these findings reveal a deeper story.&nbsp;The research community has moved beyond initial experimentation with GenAI. Researchers now understand both its value and its risks. They are using AI extensively, but also recognizing the need for clear guidance, governance, and responsible implementation. The conversation is no longer about whether and where researchers will use AI. It is about how AI use will be governed, disclosed, and verified.This transition requires coordinated action across the research ecosystem:<ul><li>Publishers must define guidelines and develop standardized disclosure requirements</li><li>Institutions must provide governance frameworks and training</li><li>Researchers must apply human oversight and verify AI-assisted outputs</li></ul>And increasingly, researchers are looking to publishers, institutions, and research integrity leaders to help define that path.<h2>Enabling Responsible AI in Research: What happens next will define trust</h2>The research community has already crossed the adoption threshold. AI is no longer a future capability. It is a present-day force shaping how knowledge is created, written, and disseminated.But adoption alone does not determine impact. Responsibility does. The next phase of AI in research will not be defined by how powerful these tools become, but by how transparently and accountably they are integrated into the scholarly ecosystem. Clear disclosure standards, human oversight, and policy-aligned infrastructure will determine whether AI strengthens research integrity or introduces risks to research credibility.This is a defining moment for publishers, institutions, and research integrity leaders. The choices made now—around disclosure, governance, and verification will shape whether AI strengthens research credibility or erodes trust in scientific outputs.At Enago, we are committed to helping the global research community operationalize responsible AI use. Our <a href="https://www.enago.com/responsible-ai-movement" target="_blank" rel="noopener noreferrer">Responsible AI initiative</a> provides practical infrastructure—including standardized disclosure tools, governance frameworks, and expert-led workshops—to help researchers, publishers, and institutions integrate AI transparently and in alignment with emerging policy requirements.Our interactive Responsible AI workshops equip research teams with the knowledge and workflows needed to implement compliant AI disclosures, establish oversight mechanisms, and strengthen accountability across the research lifecycle. Partner with Enago or&nbsp;<a href="https://www.enago.com/responsible-ai-movement/workshops" target="_blank" rel="noopener noreferrer">book a Responsible AI workshop</a> to build the policies, skills, and infrastructure required to ensure AI strengthens, rather than compromises research integrity.

Frequently Asked Questions

Wiley Did a Study on GenAI Use. 3 Things Stood Out.

Wiley’s ExplanAItions report reveals how GenAI use in research surged in 2025, while expectations fell, policies lagged, and reliance on general AI tools raised integrity and transparency risks. Learn the 3 takeaways shaping responsible AI governance, disclosure, and verification in scholarly publishing.

Wiley Study on GenAI in Research: 3 Key Insights on Adoption, Governance, Integrity

Wiley surveyed 2,400+ researchers worldwide and found GenAI is now embedded in research workflows. But beyond rising adoption, three trends stand out: expectations are becoming more realistic, institutional policies are lagging behind usage, and heavy reliance on general-purpose tools is creating new risks for research integrity, privacy, and disclosure.

thumbnail_Feb_2026 Articles_01.png

Feb_2026 Articles_01.png

small_Article_01 (3).jpg

thumbnail_Article_01 (3).jpg

Article_01 (3).jpg

Transparency around artificial intelligence use in research writing is now a near-universal expectation across scholarly publishing. Major publishers, journals, and ethics bodies have introduced Generative&nbsp;AI disclosure requirements, often framed as a cornerstone of responsible research practice. Yet, despite this apparent consensus, the ecosystem lacks agreement on a more fundamental question: what exactly counts as “AI use” in research writing?While much of the latest <a href="https://pmc.ncbi.nlm.nih.gov/articles/PMC12515416/" target="_blank" rel="noopener noreferrer">debate</a> focuses narrowly on manuscript writing and editing, this framing overlooks a broader reality. AI tools now influence foundational research decisions, including literature discovery, study design, data analysis, visualization, interpretation, and decision-making—often well before a manuscript is drafted. Treating AI disclosure as a writing-stage concern alone risks obscuring its deeper epistemic impact on research itself.<h2>Existing Gap Between AI Use and Disclosure</h2>&nbsp;A <a href="https://www.nature.com/articles/d41586-025-00343-5" target="_blank" rel="noopener noreferrer">2025 survey</a> of 5,000 researchers found more than 50% use Generative AI (GenAI) assistance for their research writing tasks. Additionally, detection <a href="https://arxiv.org/abs/2403.07183" target="_blank" rel="noopener noreferrer">studies</a> at AI conferences identified AI modifications in 6.5-16.9% of peer reviews via phrase frequency spikes, indicating hidden integration.However, a broad literature review of the available <a href="https://pubmed.ncbi.nlm.nih.gov/40675866/" target="_blank" rel="noopener noreferrer">sources</a> indicate a significant gap between AI use and its disclosure. <a href="https://pubmed.ncbi.nlm.nih.gov/40675866/" target="_blank" rel="noopener noreferrer">Bibliometric analysis</a> of 1,998 radiology manuscripts in Elsevier journals showed only 1.7% disclosed large language model use, mostly for minor edits and predominantly from non-English-speaking institutions.&nbsp;Collectively, these findings highlight a stark gap between AI adoption and disclosure, a gap shaped by multiple forces, including uncertainty over what types of AI assistance meaningfully count as “use” at different stages of the research lifecycle.<h2>A Patchwork of Definitions Across the Research Lifecycle</h2>Across policy reviews and <a href="https://www.bmj.com/content/384/bmj-2023-077192" target="_blank" rel="noopener noreferrer">comparative analyses</a>, a recurring theme emerges: AI disclosure requirements are fragmented and inconsistently defined. Most guidelines emphasize transparency in principle, but offer limited guidance on where in the research process AI involvement becomes ethically or scientifically material. Policies from major publishers show different interpretations of what qualifies as AI use requiring disclosure. For example:<ul><li><a href="https://www.sagepub.com/journals/publication-ethics-policies/artificial-intelligence-policy" target="_blank" rel="noopener noreferrer">SAGE</a> distinguishes assistive AI (no disclosure) from generative AI (disclosure), with specific but separate criteria across text, images, and analysis.</li><li><a href="https://www.nature.com/nature-portfolio/editorial-policies/ai" target="_blank" rel="noopener noreferrer">Nature Portfolio journals</a>, by contrast, require authors to disclose the use of large language models or generative AI tools (apart from their use in copy-editing), but do not provide a formal classification of AI use by function (e.g., assistive vs. generative, surface-level vs. substantive).</li></ul>While some publishers have begun to draw explicit functional distinctions between different types of AI use, while others retain broad, principle-based language that leaves substantial room for interpretation. Terms such as “substantial contribution,” “meaningful assistance,” or “content generation” appear frequently in publisher policies without thresholds, examples, or decision criteria. As a result, identical AI uses may be disclosed in one journal and omitted in another. This inconsistency extends beyond writing into upstream research activities, where AI support is often invisible to editors and readers alike.<h2>The Narrow Focus on Writing Misses the Larger Risk</h2>Focusing disclosure policies primarily on manuscript preparation creates a misleading boundary and misrepresents where AI influence is most consequential. AI systems increasingly assist with:<ul><li>Identifying relevant literature and synthesizing prior work</li><li>Generating or refining research questions</li><li>Supporting statistical analysis and modeling</li><li>Producing data visualizations and figures</li><li>Identifying patterns that influence interpretation</li></ul>Each of these activities can shape how knowledge is produced, framed, and interpreted. Yet many fall outside the scope of current disclosure norms, not because they are ethically neutral, but because definitions of AI use remain underdeveloped.This gap creates a paradox: visible AI use in writing is scrutinized, while less visible but potentially more influential AI use earlier in the research process often goes undisclosed.<h2>Editing, Automation, and Intellectual Influence: Where confusion begins</h2>The definitional challenge is compounded by long-standing norms around automation and support tools in research. Statistical software, reference managers, plagiarism detectors, and language editors have long been integrated into scholarly workflows without formal disclosure requirements.AI blurs these categories. When automation shifts from executing predefined instructions to generating suggestions, interpretations, or summaries, the boundary between support and intellectual influence becomes blurred. <a href="https://pmc.ncbi.nlm.nih.gov/articles/PMC12515416/" target="_blank" rel="noopener noreferrer">Studies</a> consistently show that researchers struggle to determine when AI assistance crosses this line—particularly when the output is reviewed and approved by a human.Without guidance that distinguishes between process efficiency and epistemic contribution, disclosure decisions are left to individual judgment, resulting in uneven reporting.<h2>Why Precision is Important to Responsible AI</h2>The absence of clear, lifecycle-wide definitions has implications that extend beyond compliance:<ul><li>Trust: Readers cannot reliably interpret disclosure statements if similar AI uses are reported differently. A recent <a href="https://arxiv.org/abs/2512.06705" target="_blank" rel="noopener noreferrer">analysis</a> of &gt;5.2 million papers showed a dramatic transparency gap. Only 0.1 % of 75,000 papers published since 2023 explicitly disclosed AI use, despite widespread policy adoption and actual use, suggesting trust issues when actual AI involvement is not known.</li><li>Comparability: Inconsistent and incomplete AI disclosure affects the reliability of meta-research that seeks to assess AI’s prevalence, impact, and risks in scholarly communication. A <a href="https://link.springer.com/article/10.1186/s12910-025-01239-9" target="_blank" rel="noopener noreferrer">comparative analysis</a> of AI policies in bioethics and humanities journals found 30% journals did not have a publicly available AI policy. This fragmentation makes it difficult to compare AI use across journals, weakening evidence-based policy development and obscuring the true role of AI in research practices.</li><li>Governance: Editors and publishers lack shared enforceable standards. Confusion about policy application risks uneven editorial decisions. The <a href="https://link.springer.com/article/10.1186/s12910-025-01239-9" target="_blank" rel="noopener noreferrer">BMC comparative study</a> discusses that lack of clear AI policies can confuse authors about whether to disclose AI use, and can result in different editors within the same journal applying different interpretations, weakening governance.</li><li>Equity: Definitional uncertainty and policy inconsistency create&nbsp;<a href="https://pmc.ncbi.nlm.nih.gov/articles/PMC12515416/" target="_blank" rel="noopener noreferrer">asymmetries</a> among researchers: those who disclose AI use may face perceived editorial risk, while others benefit from AI support without transparency. These dynamics disadvantage authors who seek to comply ethically, reinforcing inequities in scholarly recognition and trust.</li></ul>As a result, responsible AI governance in research depends not simply on disclosure mandates, but on definitions that align with how research is actually conducted. Rather than asking whether AI was used, the more meaningful question is: Did AI meaningfully influence research decisions, interpretations, or scholarly outputs?This shift enables proportional transparency, recognizing that not all AI use carries the same ethical or scientific weight.<h2>Toward a Holistic, Tiered Approach</h2>A lifecycle-aware disclosure framework distinguishes between:<figure class="table"><table><tbody><tr><td style="border:1.0pt solid windowtext;padding:0in 5.4pt;vertical-align:top;">Proposed Category of AI Use</td><td style="border-bottom-style:solid;border-color:windowtext;border-left-style:none;border-right-style:solid;border-top-style:solid;border-width:1.0pt;padding:0in 5.4pt;vertical-align:top;">Definition</td><td style="border-bottom-style:solid;border-color:windowtext;border-left-style:none;border-right-style:solid;border-top-style:solid;border-width:1.0pt;padding:0in 5.4pt;vertical-align:top;">Typical Examples to Consider</td><td style="border-bottom-style:solid;border-color:windowtext;border-left-style:none;border-right-style:solid;border-top-style:solid;border-width:1.0pt;padding:0in 5.4pt;vertical-align:top;">Proposed Disclosure Expectation</td><td style="border-bottom-style:solid;border-color:windowtext;border-left-style:none;border-right-style:solid;border-top-style:solid;border-width:1.0pt;padding:0in 5.4pt;vertical-align:top;">Expected Impact on Editorial Handling</td></tr><tr><td style="border-bottom-style:solid;border-color:windowtext;border-left-style:solid;border-right-style:solid;border-top-style:none;border-width:1.0pt;padding:0in 5.4pt;vertical-align:top;">Routine Automation</td><td style="border-bottom:1.0pt solid windowtext;border-left-style:none;border-right:1.0pt solid windowtext;border-top-style:none;padding:0in 5.4pt;vertical-align:top;">AI-enabled functions that support efficiency for mechanical tasks without shaping research content, reasoning, or credibility in any manner.</td><td style="border-bottom:1.0pt solid windowtext;border-left-style:none;border-right:1.0pt solid windowtext;border-top-style:none;padding:0in 5.4pt;vertical-align:top;">Spell-checking, citation formatting, transcription without summarization, plagiarism screening, file or figure formatting</td><td style="border-bottom:1.0pt solid windowtext;border-left-style:none;border-right:1.0pt solid windowtext;border-top-style:none;padding:0in 5.4pt;vertical-align:top;">Not required</td><td style="border-bottom:1.0pt solid windowtext;border-left-style:none;border-right:1.0pt solid windowtext;border-top-style:none;padding:0in 5.4pt;vertical-align:top;">Treated as standard digital infrastructure; no impact on peer review or authorship</td></tr><tr><td style="border-bottom-style:solid;border-color:windowtext;border-left-style:solid;border-right-style:solid;border-top-style:none;border-width:1.0pt;padding:0in 5.4pt;vertical-align:top;">Assistive Use</td><td style="border-bottom:1.0pt solid windowtext;border-left-style:none;border-right:1.0pt solid windowtext;border-top-style:none;padding:0in 5.4pt;vertical-align:top;">AI support that improves workflow or presentation by reshaping the expression, but does not directly or greatly alter scholarly claims or conclusions</td><td style="border-bottom:1.0pt solid windowtext;border-left-style:none;border-right:1.0pt solid windowtext;border-top-style:none;padding:0in 5.4pt;vertical-align:top;">Language editing, paraphrasing for clarity, translation verified by authors, structural suggestions</td><td style="border-bottom:1.0pt solid windowtext;border-left-style:none;border-right:1.0pt solid windowtext;border-top-style:none;padding:0in 5.4pt;vertical-align:top;">Recommended (light-touch)</td><td style="border-bottom:1.0pt solid windowtext;border-left-style:none;border-right:1.0pt solid windowtext;border-top-style:none;padding:0in 5.4pt;vertical-align:top;">Disclosure recorded for transparency; no editorial escalation</td></tr><tr><td style="border-bottom-style:solid;border-color:windowtext;border-left-style:solid;border-right-style:solid;border-top-style:none;border-width:1.0pt;padding:0in 5.4pt;vertical-align:top;">Substantive Contributions</td><td style="border-bottom:1.0pt solid windowtext;border-left-style:none;border-right:1.0pt solid windowtext;border-top-style:none;padding:0in 5.4pt;vertical-align:top;">AI use that contributes directly or indirectly to research outputs, while humans retain decision-making control</td><td style="border-bottom:1.0pt solid windowtext;border-left-style:none;border-right:1.0pt solid windowtext;border-top-style:none;padding:0in 5.4pt;vertical-align:top;">Drafting background sections, generating tables or figures, data classification, literature synthesis support</td><td style="border-bottom:1.0pt solid windowtext;border-left-style:none;border-right:1.0pt solid windowtext;border-top-style:none;padding:0in 5.4pt;vertical-align:top;">Required</td><td style="border-bottom:1.0pt solid windowtext;border-left-style:none;border-right:1.0pt solid windowtext;border-top-style:none;padding:0in 5.4pt;vertical-align:top;">Editors assess scope and appropriateness; reviewers may be informed</td></tr><tr><td style="border-bottom-style:solid;border-color:windowtext;border-left-style:solid;border-right-style:solid;border-top-style:none;border-width:1.0pt;padding:0in 5.4pt;vertical-align:top;">Interpretive Influence</td><td style="border-bottom:1.0pt solid windowtext;border-left-style:none;border-right:1.0pt solid windowtext;border-top-style:none;padding:0in 5.4pt;vertical-align:top;">AI use that affects reasoning, interpretation, or conclusions</td><td style="border-bottom:1.0pt solid windowtext;border-left-style:none;border-right:1.0pt solid windowtext;border-top-style:none;padding:0in 5.4pt;vertical-align:top;">AI-supported interpretation of results, predictive modeling, inference generation, scenario analysis</td><td style="border-bottom:1.0pt solid windowtext;border-left-style:none;border-right:1.0pt solid windowtext;border-top-style:none;padding:0in 5.4pt;vertical-align:top;">Mandatory with detailed description</td><td style="border-bottom:1.0pt solid windowtext;border-left-style:none;border-right:1.0pt solid windowtext;border-top-style:none;padding:0in 5.4pt;vertical-align:top;">Heightened editorial scrutiny; possible methodological justification or limitations required</td></tr></tbody></table></figure>Such an approach aligns disclosure requirements with function, influence, and accountability, rather than with specific technologies or brand names. It also preserves a foundational principle consistently emphasized across ethics guidelines: human researchers remain fully responsible for the integrity of their work, regardless of AI involvement.<h2>Defining AI Use Is a Governance Imperative</h2>The scholarly community has moved quickly to demand transparency around AI. What it has not yet resolved is the definitional foundation that makes transparency meaningful.As AI continues to shape research across the lifecycle, limiting disclosure discussions to writing alone risks misrepresenting its role in knowledge production. A holistic, impact-based understanding of AI use is essential—not only for consistent disclosure, but for sustaining trust in the scholarly record.For <a href="https://www.enago.com/responsible-ai-movement/publisher-ai-guidelines" target="_blank" rel="noopener noreferrer">Enago’s Responsible AI Movement</a>, this challenge conveys a central message: responsible AI adoption begins with clear definitions that reflect real research practice, not just policy intent.

What Counts as “Generative AI Use” in Research Writing? The Definition Problem No One Has Solved!

Over 50% of researchers use GenAI, yet only 0.1% disclose it. Explore why AI disclosure policies fail without clear definitions of AI use across the research lifecycle, and a proposed tiered framework.

What Counts as Generative AI Use in Research Writing? The Definition Gap in Scholarly Publishing

Most journals now require AI disclosure but what actually counts as "AI use"? With over 50% of researchers using GenAI tools yet fewer than 2% disclosing it, the problem isn't transparency policies it's the lack of clear, lifecycle-wide definitions of AI involvement. This article examines why current disclosure norms fall short and proposes a tiered, impact-based framework for responsible AI governance in research.

thumbnail_Feb_2026 Articles.png

Feb_2026 Articles.png

small_Article_01_720x360 (2).jpg

thumbnail_Article_01_720x360 (2).jpg

Article_01_720x360 (2).jpg

Generative AI (Gen&nbsp;AI) tools such as ChatGPT are rapidly becoming embedded in scholarly workflows; assisting researchers with literature discovery, summarization, and reference list generation. While these systems offer undeniable efficiency gains, their use for reference and citation generation carries an often overlooked risk: hallucinated references.For researchers operating under intense publication pressures, this risk is no longer theoretical. It has direct implications for manuscript credibility, peer review outcomes, and long-term research integrity.<h2>What the Evidence Shows: Hallucinations at Scale</h2>A <a href="https://pmc.ncbi.nlm.nih.gov/articles/PMC12658395/" target="_blank" rel="noopener noreferrer">2025 study</a> published in PubMed Central (PMC) provides one of the most comprehensive evaluations to date of AI-generated references. The authors systematically assessed citations produced by GPT-based systems and reported alarming results:<ul><li>19.9% of AI-generated references were completely fabricated, with no traceable existence in the scholarly record.</li><li>Among the remaining citations, 45.4% contained serious bibliographic errors, including incorrect author names, journal titles, publication years, volumes, or DOIs.</li></ul>In effect, fewer than one in three references generated by the model were fully accurate and verifiable—a failure rate incompatible with scholarly standards. Crucially, the study also demonstrated that these errors were not random.&nbsp;Another <a href="https://pmc.ncbi.nlm.nih.gov/articles/PMC10488525/" target="_blank" rel="noopener noreferrer">PMC study</a> on identifying relevant nephrology papers using ChatGPT revealed that:<ul style="list-style-type:disc;"><li>68% of the links in the references were incorrect</li><li>31% were fabricated</li><li>7% of citations were incomplete references</li></ul>Fabricated citations often appeared plausible, mimicking real journals, legitimate-sounding article titles, and familiar author naming patterns. This makes hallucinations particularly difficult to detect without deliberate verification, especially for early-career researchers or interdisciplinary authors working outside their core domain.<h2>Publisher&nbsp;Credibility on the Line</h2>The theoretical concern about hallucinated citations has, in <a href="https://retractionwatch.com/2025/06/30/springer-nature-book-on-machine-learning-is-full-of-made-up-citations/" target="_blank" rel="noopener noreferrer">2025</a>, manifested in a high-profile scholarly publishing failure. The recent Springer Nature textbook retraction is often described as an “AI citation scandal.” But framing it this way obscures the more uncomfortable truth: this was a governance breakdown, not a technological anomaly. Independent checks of the machine learning textbook Mastering Machine Learning: From Basics to Advanced discovered that two-thirds of the sampled citations either did not exist or were materially inaccurate. Several researchers listed in the references confirmed they had never authored the works attributed to them—classic markers of AI-style fabrication.&nbsp;Following an integrity review, Springer Nature retracted the book and removed it from catalogs after recognizing that 25 out of 46 listed references could not be verified. What makes this case particularly instructive is not the presence of AI-style hallucinations—but the fact that they passed through multiple institutional checkpoints:<ul style="list-style-type:disc;"><li>Author submission</li><li>Editorial assessment</li><li>Peer review</li><li>Production and publication</li></ul>This is not an isolated anecdote. Similar problems have reportedly occurred (and are unfortunately expected to rise up in the future) in other academic books purporting to cover social, ethical, and legal aspects of AI, where unverifiable citations have sparked publisher investigations.&nbsp;These cases emphasize that hallucinated references are no longer restricted to early-career papers or student submissions: they have infiltrated mainstream academic publishing, even at major multinational publishers.<h2>Why Citation Hallucinations Expose Structural Weaknesses</h2>Hallucinations happen because AI systems, like GPT models, don’t have true access to real-time or live databases—they generate text by predicting what would be plausible based on patterns in training data. Therefore, references that sound legitimate can be fabricated, even though they don't exist in any research database.The problem is not that generative models occasionally fabricate references. This behavior is well <a href="https://www.sciencedirect.com/science/article/pii/S2949761223000366" target="_blank" rel="noopener noreferrer">documented</a> and technically predictable. The real issue is that academic systems continue to operate as if AI outputs are epistemically equivalent to human-authored content, without adapting policies, checks, or responsibilities accordingly.In many research workflows, reference accuracy is implicitly assumed to be verified “somewhere” in the process by co-authors, reviewers, or editorial teams; often under severe time constraints. In multi-author collaborations, responsibility for AI-assisted citations can become diffuse, allowing hallucinations to enter manuscripts unnoticed.Once published, fabricated or corrupted references can rapidly gain legitimacy through indexing, citation chaining, and reuse in secondary literature. This creates a silent integrity risk: citation errors that evade plagiarism detection and statistical checks, yet directly undermine reproducibility, attribution, and trust.<h2>Why This Matters for Your Next Paper</h2>References are not cosmetic elements of a manuscript. They form the scholarly infrastructure. They enable:<ul><li>Verification of claims</li><li>Reproducibility of methods</li><li>Proper attribution of intellectual labor</li><li>Cumulative knowledge building</li></ul>When fabricated or corrupted references perpetuate the scholarly record, they threaten each of these functions. Worse, they can persist undetected, being recited, indexed, and propagated by downstream authors, databases, and meta-analyses.This creates a feedback loop where false citations gain apparent legitimacy over time, eroding trust not only in individual papers, but in the reliability of the academic corpus itself.<h2>Potential Solutions</h2>Responsible AI adoption in research does not require abandoning generative tools—but it does demand clear boundaries, verification workflows, and institutional guidance. Organizations such as <a href="https://publicationethics.org/cope-focus/cope-focus-artificial-intelligence" target="_blank" rel="noopener noreferrer">COPE</a>, <a href="https://stm-assoc.org/document/recommendations-for-a-classification-of-ai-use-in-academic-manuscript-preparation/" target="_blank" rel="noopener noreferrer">STM</a>, and <a href="https://www.enago.com/responsible-ai-movement/publisher-ai-guidelines" target="_blank" rel="noopener noreferrer">major journal publishers</a> now emphasize that authors remain fully accountable for the accuracy of references, regardless of AI involvement.Instead of relying on AI to generate full citations, researchers should utilize AI tools for organizing and formatting references that have already been verified. This allows researchers to streamline workflows without risking incorrect or fabricated citations. Some best-practice safeguards increasingly recommended by publishers, integrity bodies, and AI governance frameworks include:<ul><li>Incorporating reference validation checks early in the writing process, using citation verification tools, and involving experienced co-authors in the final verification of citations.</li><li>Check with your institution or publisher for updated guidelines on AI usage in academic writing. If none exist, work with colleagues and peers to create and adopt internal protocols.</li><li>Training of&nbsp;researchers and reviewers&nbsp;to recognize hallucination patterns</li><li>Look into services or third-party tools that focus on AI-driven citation checking.</li></ul>As AI becomes embedded across the research life&nbsp;cycle, citation hallucination represents one of the clearest stress tests for responsible AI governance. It exposes a fundamental tension between speed and scholarly rigor and emphasizes why human oversight is not optional.Remember -&nbsp;Responsible AI in academia is not about rejecting innovation. It is about ensuring that efficiency never comes at the cost of truth and that progress is built on a foundation the scholarly community can rely upon.&nbsp;If left unchecked, AI-generated reference pollution risks weakening the very foundation of academic trust. If addressed responsibly, however, it can become a catalyst for better workflows, clearer policies, and more resilient research practices.For researchers, this moment presents a choice: to use AI passively and absorb its risks, or to engage actively in shaping how AI is governed within scholarly work. <a href="https://www.enago.com/responsible-ai-movement" target="_blank" rel="noopener noreferrer">Enago’s Responsible AI movement</a> invites researchers to be part of this stewardship—by adopting evidence-based best practices, contributing to community-driven standards, and embedding verification-first thinking into everyday research workflows. By engaging with RUAI, researchers can help ensure that AI strengthens, rather than erodes, the credibility of the academic record—and that innovation in research remains both powerful and trustworthy. Researchers interested in <a href="https://www.enago.com/responsible-ai-movement" target="_blank" rel="noopener noreferrer">contributing to or learning more</a> about responsible AI practices can connect with the initiative through <a href="https://www.enago.com/responsible-ai-movement" target="_blank" rel="noopener noreferrer">Enago’s inquiry form</a>, creating a direct channel to explore guidance, resources, and collaboration opportunities tailored to real-world research workflows.

“Looks Cited” Doesn’t Mean “is Cited.” Why 1 in 5 AI References Fail.

Discover how AI tools like ChatGPT are polluting scholarly research with fake citations. Learn why 19.9% of AI-generated references are fabricated and how to protect research integrity.

AI-Generated References: 1 in 5 Are Fake | Citation Integrity Crisis

Nearly 1 in 5 AI-generated references are completely fake. Learn how hallucinated citations threaten research integrity.

thumbnail_Article_01_720x360 (1).jpg

Article_01_720x360 (1).jpg

small_Article_03_720x360.jpg

thumbnail_Article_03_720x360.jpg

Article_03_720x360.jpg

The recent <a href="https://www.medrxiv.org/content/10.1101/2025.10.24.25338574v1" target="_blank" rel="noopener noreferrer">preprint</a> titled “Authors self-disclosed use of artificial intelligence in research submissions to 49 biomedical journals” reports a striking figure: only 5.7% of the 25,114 empirical biomedical manuscripts submitted between April and November 2024 openly disclosed any use of AI. This sharply contrasts with prior surveys, where between 28%&nbsp;and 76% of researchers admitted to using AI tools in preparing their manuscripts. This gap between estimated AI use and actual disclosure highlights a growing tension: AI is rapidly reshaping biomedical research and writing, but transparency is dangerously lagging behind.&nbsp;<h2>The 5.7% Disclosure Rate — and Why It Matters</h2>The study analyzed 25,114 empirical manuscripts submitted across 49 biomedical journals published by BMJ between April 8 and November 6, 2024. Of these, only 1,431 (5.7%) included a self-report of AI use. The most commonly disclosed AI tools were generative chatbots (e.g., LLM-based writing assistants), accounting for 56.7% of disclosures; writing-assistants more broadly accounted for 12.7%. The main declared purpose: improving writing quality (87% of disclosing authors). Other uses like translation, data generation, literature search, data analysis, image processing, code writing, reference management, were reported far less frequently. This means most disclosures reveal only low-risk uses while higher-risk AI tasks remain unreported. Furthermore, a separate <a href="https://pubmed.ncbi.nlm.nih.gov/40675866/" target="_blank" rel="noopener noreferrer">bibliometric analysis</a> of 1,998 radiology papers in Elsevier journals reported that only 1.7% disclosed the use of LLM’s (large language models).<h2>Why does this low rate of disclosure matter?</h2><ul style="list-style-type:disc;"><li>Scientific Integrity and Reproducibility:</li></ul>&nbsp;The use of AI for writing, analysis, data generation, or translation can affect meaning, nuance, and even compromise data integrity. Without transparency, peer reviewers and readers cannot properly evaluate the provenance of text, analyses, or interpretations.<ul style="list-style-type:disc;"><li>Erosion of Trust and Accountability:</li></ul>When AI use remains hidden, it fuels what some describe as a “hidden ecosystem” of AI-assisted writing and research. This puts accountability, authorship normativity, and the trustworthiness of the scientific record under radar.&nbsp;<h2>Is This Figure a Tip of the Iceberg?</h2>The answer remains “Yes!” Several independent studies suggest AI involvement in a much higher proportion of biomedical research, often without disclosure. A large-scale linguistic <a href="https://arxiv.org/abs/2406.07016?" target="_blank" rel="noopener noreferrer">analysis</a> of over 15 million biomedical abstracts (2010–2024) found a marked increase in so-called “style words,” which are adjectives and verbs that are typically found in LLM-generated text. Authors argue that these stylistic markers point to at least 13.5% of 2024 abstracts having been processed by large language models.&nbsp;These findings collectively suggest that the true prevalence of AI assistance in biomedical research is likely much higher than what disclosure statistics indicate. In other words: many papers may depend on AI for writing, editing, data handling, or even drafting, without any formal acknowledgment or structured disclosure.<h2>What are the Repercussions? &nbsp;</h2><h3>1. Accountability</h3>When authors use AI but do not disclose it, responsibility for accuracy, integrity, and originality becomes dangerously murky. If AI-generated content contains errors or biased phrasing, who is responsible? Undisclosed use blurs that line.<h3>2. Accuracy and Validity</h3>Generative AI tools, irrespective of their latest updates and the extent of training, are not flawless. They can hallucinate facts, misinterpret data, fabricate citations, overstate significance, or introduce subtle biases. Without a clear record of what was AI-assisted (e.g., text generation, data summarization, image processing), peers and readers cannot reliably assess and cross-verify methodological rigor, reproducibility, or interpret results in context.<h3>3. Peer Review and Publishing Systems</h3>Systematic non-disclosure can create a creeping “shadow literature,” where parts of academic output are AI-augmented and unknown to reviewers, editors, or readers. This threatens long-term trust in published evidence, especially in highly-sensitive fields like medicine, where accuracy matters for patient safety and policy decisions.<h2>Where Do the Existing Guidelines Fall Short?</h2>Many publishers and journals have issued guidance on AI use. Major academic publishers like <a href="https://www.elsevier.com/about/policies-and-standards/generative-ai-policies-for-journals" target="_blank" rel="noopener noreferrer">Elsevier</a>, <a href="https://authorservices.wiley.com/ethics-guidelines/index.html#22" target="_blank" rel="noopener noreferrer">Wiley</a>, <a href="https://taylorandfrancis.com/our-policies/ai-policy/" target="_blank" rel="noopener noreferrer">Taylor &amp; Francis</a>, and <a href="https://www.sagepub.com/journals/publication-ethics-policies/artificial-intelligence-policy" target="_blank" rel="noopener noreferrer">SAGE </a>have issued policies prohibiting AI authorship while allowing limited, disclosed use for tasks like language editing, with human oversight required. However, existing practices are fragmented, inconsistent, and often insufficient to ensure real transparency. Moreover, these guidelines often vary not only between publishers, but also between journals under the same publisher, leaving authors confused, and disclosure optional, inconsistent, or omitted. In addition, a lack of unified language and structure, i.e., where to disclose (methods section, acknowledgments, cover letter), how detailed to be (just tool name or detailed description of use), what types of AI use require disclosure (writing, data processing, analysis, translation, etc.) — adds to the opacity. This confusion and heterogeneity can dissuade authors from disclosing, even when they rely heavily on AI.<h2>What a Comprehensive Disclosure Guideline Could and Should Offer</h2>Given the stakes, the scholarly community needs a unified, standardized framework for AI disclosure. Here’s what such a guideline should include:<h3>1. Clear Definition of What Constitutes the “Right AI Use”</h3>The guideline should broadly define AI use to include: generative writing assistants (e.g., large language models), translation tools, code-generation, data analysis/processing tools, image-processing AI (e.g., for radiology), reference management AI, and any other AI-driven assistance. This helps avoid narrow definitions that only consider “writing” as AI use.<h3>2. Mandatory Disclosure With Structured Format</h3>Instead of leaving disclosure optional or hidden in cover letters, journals should require a dedicated section (e.g., “AI-Assistance Declaration”) in manuscripts where authors mention the AI tools used (by name or version), describe what they were used for (writing, translation, data analysis, image processing, reference management, etc.), clarify whether AI was used to generate new content (text/data), or only to assist (editing, translation, polishing).<h3>3. Alignment With Broader Open Science and Reproducibility Practices</h3>AI disclosure should be integrated into larger frameworks of transparency: data sharing, code availability, methodology reporting. Especially for studies using AI for data analysis or image processing, journals should encourage (or require) sharing of AI-generated code, models, and data, where permissible.<h2>Towards a Transparent, Trustworthy AI-Augmented Future</h2>The fact that only 5.7% of biomedical research manuscripts disclosed AI use, despite far higher estimates of actual AI involvement, is a wake-up call for the entire publishing ecosystem. AI tools are already reshaping how scientific research is conducted, written, and reviewed. But this transformation comes with risks: loss of transparency, dilution of accountability, compromised reproducibility, and potential erosion of trust in the scientific record.A unified, community-wide guideline for AI disclosure that is mandatory, structured, transparent, is no longer an optional nicety. It is essential. Supported by frameworks like <a href="https://arxiv.org/abs/2303.03886" target="_blank" rel="noopener noreferrer">AI Usage Cards</a>, and aligned with open-science and robust reporting practices, such guidelines can safeguard the integrity of biomedical research even as AI becomes deeply integrated.If journals, publishers, researchers, and funders adopt clear, consistent disclosure standards now, we can reap the productivity and accessibility benefits of AI without compromising the core ethical values in publishing. If you believe the scientific community deserves clarity and not guesswork, then join <a href="https://www.enago.com/responsible-ai-movement" target="_blank" rel="noopener noreferrer">Enago’s Responsible AI movement</a> to help set the standard and define what responsible AI looks like in global research!

Only 5.7% Disclosed AI Use! Why Researchers Need Unified Guidelines Urgently

A major biomedical study reveals that just 5.7 percent of manuscripts disclosed AI use despite widespread adoption. Explore the risks of non-disclosure and the urgent need for unified AI guidelines.

Only 5.7 Percent Disclosed AI Use – Why Unified Guidelines Are Urgently Needed

New evidence shows that fewer than 6 percent of biomedical manuscripts disclose AI use, exposing critical gaps in transparency, accountability, and trust across scholarly publishing.

thumbnail_Article_03_340x200.png

Article_03_340x200.png

small_Article_01.jpg

thumbnail_Article_01.jpg

Article_01.jpg

In a stark wake-up call for scholarly communication, a recent <a href="https://www.science.org/content/article/one-fifth-computer-science-papers-may-include-ai-content" target="_blank" rel="noopener noreferrer">Science report</a> revealed: about one in five computer science papers may contain content generated by artificial intelligence (AI).&nbsp; What makes this even more alarming is the <a href="https://phys.org/news/2025-08-scientific-papers-written-chatgpt-science.html" target="_blank" rel="noopener noreferrer">speed of escalation</a>. In early 2022, detectable AI-written text in such papers was negligible; by mid-2023 it began to surface sporadically; and by late 2024 it had surged to over 22% of abstracts in some subfields. This stunning, two-year rise has sent shockwaves through the research ecosystem because it spotlights a broader and accelerating trend of AI use in research writing, raising urgent concerns about authorship, accountability, and trust.<h2>What did the Science Study Find?</h2>According to the Science article, researchers observed a sharp surge in AI-generated language across scientific manuscripts, particularly since November 2022, when OpenAI released ChatGPT. By analyzing patterns in phrasing, especially in abstracts and introductions in more than 1.1 million scientific articles and preprints (2020–2024), they found that up to 22.5% of computer science abstracts showed signs of large language model (LLM) involvement.&nbsp;Other fields also saw increasing AI influence: electrical engineering (~18%), statistics (~12.9%), physics (~9.8%), and even mathematics (~7.8%). Notably, these increases accelerated sharply after the release of ChatGPT, indicating a structural shift rather than a temporary trend.These numbers suggest that generative AI isn’t just a writing aid. What makes this trend especially concerning is the human detection gap. Reviewers, editors, and even experienced researchers are increasingly unable to differentiate between human-written and AI-written text. As a result, AI-modified writing is silently entering the literature, often without disclosure, and often without being noticed at all.<h2>The Other Side of the Coin: Predatory journals exposed by AI</h2>Meanwhile, a recent University of Colorado Boulder&nbsp;<a href="https://www.sciencedaily.com/releases/2025/08/250830001203.htm" target="_blank" rel="noopener noreferrer">study</a> revealed that a different AI tool has&nbsp;been working to identify potential threats to scientific integrity. The system screened ~15,200 open-access&nbsp;journals and initially&nbsp;flagged more than 1,400 journals as suspicious. After expert human review, over 1000 were confirmed to exhibit hallmarks of predatory publishing. This means that roughly 1 in every 14 open-access journals may be operating without credible scholarly standards. This figure is explosive, yet still dramatically under-communicated in the broader academic community.Some of the strongest red flags included missing or fake editorial boards, unusually high levels of self-citation, and authors with many institutional affiliations.<h2>Why These Two Threads Matter? And How They Connect</h2>Putting both trends together— AI quietly appearing in thousands of manuscripts and a different AI system exposing more than 1,000 suspect journals—reveals a landscape of both opportunity and peril for academic publishing.<h3>1. Scaling Risk &amp; Reward</h3>AI helps researchers write faster, polish text, and generate ideas. However, the same tools can produce authoritative-sounding yet misleading content (“hallucinations”), and may be used to inflate publication count with low-quality papers.<h3>2. Broken Transparency</h3>Many authors now generate substantial portions of their manuscripts using AI but do not disclose it. For example, the <a href="https://arxiv.org/pdf/2411.15218" target="_blank" rel="noopener noreferrer">Academ-AI dataset</a> shows hundreds of suspected instances of undeclared AI usage in peer-reviewed literature. Without clear acknowledgement, AI-assisted writing can affect trust: readers, reviewers, and institutions may not realize what “authorship” actually means in these cases.<h3>3. Predatory Publishing + AI = Double Threat</h3>Predatory journals exploit academic pressure for “publish or perish.” With AI writing, it becomes easier and cheaper to generate manuscripts that appear credible but lack scientific rigor. This creates a dangerous “supply chain” between AI-generated text and journals willing to publish anything for a fee.<h2>Why This Matters? Risks and Implications&nbsp;</h2><h3>1. Transparency &amp; Disclosure</h3>One of the most significant problems is that AI use is rarely being disclosed. Without clear acknowledgment, readers and reviewers may misinterpret AI-assisted sections as entirely human-written. This opacity affects trust. As some journals still lack concrete guidelines, many authors may not even realize the ethical dimensions of non-disclosure.<h3>2. Quality vs. Quantity Trade-off</h3>While AI can accelerate the writing process, there’s a risk that parts of a manuscript, especially pivotal sections like introductions, become formulaic or hollow. If generative tools are overused, we might see a loss of the subtle creativity, rigor, or critical insight that characterizes strong scientific writing.<h3>3. Peer-Review Challenges</h3>If peer reviewers are unaware of AI involvement, they may fail to catch errors, hallucinations, or misleading phrasing. Given how LLMs can hallucinate plausible but false claims, this is not a hypothetical risk. The more LLM-generated content slips through unchecked, the more credibility the academic record could lose.<h3>4. Erosion of Genuine Authorship and Credit</h3>When AI writes portions of a paper (or entire drafts), how do we assess contribution? Traditional norms of authorship like who did the experiments, who wrote, who edited may blur. This could challenge how credit, responsibility, and accountability are allocated in research teams.<h2>What Can Be Done?</h2>Given the scale and speed of AI’s infiltration into scientific writing, the academic community must act quickly. Here are some practical steps forward:<ol><li>Standardized Disclosure Statements Journals should require authors to clearly disclose where and how they used AI. The disclosure should include:</li></ol><ul><li>Which tool was used (e.g., ChatGPT, Claude)</li><li>The version or model (e.g., GPT-4)</li><li>When and how it was used (e.g., drafting introduction, polishing grammar)</li><li>The extent of use (percentage of text, sections)</li></ul><ol><li>AI-Usage Checklists for Submissions Just as authors fill out conflict-of-interest or data-sharing checklists, they should also complete an AI-usage checklist. This could help editors and reviewers understand potential AI contributions and decide where to scrutinize more deeply.</li><li>Training for Reviewers and Editors Peer reviewers and journal editors need to be educated on how to detect LLM-like writing. By learning common AI tropes — unusual phrasing, repetitive structures — they can better assess manuscript authenticity.</li><li>Open, Transparent Detection Tools Develop and support open-source tools for detecting AI-generated text. Transparency in the detection process will build trust; black-box detectors may raise fairness issues (e.g., bias against non-native English writers).</li><li>Institutional &amp; Funder Policies Universities, funding bodies, and research institutions should develop policies around acceptable AI usage. They should balance encouraging productivity with ensuring scientific integrity.</li></ol><h2>Why This Moment Matters</h2>The convergence of AI-assisted writing and AI-enabled journal policing represents a pivotal moment in the history of science. On one hand, generative tools can democratize writing, make knowledge creation more accessible, and speed up idea exchange. On the other hand, without guardrails, generative AI may erode the integrity of the research record, blur authorship, and bolster shady publishing practices. Yet there is hope. What we urgently need are community norms, shared policies, and transparency. With well-defined AI-disclosure guidelines, we can ensure that generative AI adds value to research, rather than becoming a reason for its degradation.Organizations across the ecosystem are beginning to define what “responsible AI” truly means in research, and one of the most proactive voices in this space is Enago’s Responsible AI Movement. By offering frameworks, training resources, and guidance for researchers worldwide, <a href="https://www.enago.com/responsible-ai-movement" target="_blank" rel="noopener noreferrer">Enago’s movement</a> reinforces the principle that AI should augment and not replace human scholarship. Ultimately, the future of science depends not only on technological tools but on the values that guide their use.

Researcher Alert! AI Writing Surges in Research: 1 in 5 Computer Science Papers and Over 1,000 Journals at Risk

A Science study reveals 1 in 5 computer science papers contain AI-generated content, while detection tools exposed 1,000+ predatory journals. Explore the transparency crisis threatening academic integrity and solutions for responsible AI use in research.

AI in Academic Publishing: 22% of Papers Use AI & 1000+ Predatory Journals Exposed

Discover how AI is reshaping academic publishing: from 22% of computer science papers containing AI-generated content to over 1,000 predatory journals exposed by detection tools. Learn about the transparency crisis and what the research community must do now.

thumbnail_Dec_2025_Article_01.png

Dec_2025_Article_01.png

small_article 1.jpg

thumbnail_article 1.jpg

article 1.jpg

With the rise of Generative AI (GenAI) tools, researchers now have access to assistants that can refine text, rephrase sentences, and even suggest new ideas. But what happens when AI’s role in the writing process becomes so significant that it blurs the line between assistance and authorship?Without proper oversight, AI can unknowingly take on the role of a “co-author” by influencing or altering the content and key arguments of a <a href="https://www.enago.com/submission-preparation.htm">research paper</a>, without any formal acknowledgement. The <a href="https://retractionwatch.com/2025/02/10/as-springer-nature-journal-clears-ai-papers-one-universitys-retractions-rise-drastically/">recent surge</a> in AI-generated manuscripts and subsequent retractions in Neurosurgical Review highlights the evolving challenges in research integrity. Furthermore, at least <a href="https://www.aapsnewsmagazine.org/aapsnewsmagazine/articles/new-page3/feb25/meetings-feb25">10% of scientific abstracts</a> are believed to likely undergo review by an LLM, based on linguistic analysis of overused words like ‘delve.’ This highlights the growing influence of AI in academic writing.Table of Contents<ul><li><a href="https://www.enago.com/academy/ai-unintentional-co-author-ethical-risks/#How_AI_Can_Unknowingly_Become_a_Co-Author">How AI Can Unknowingly Become a Co-Author</a></li><li><a href="https://www.enago.com/academy/ai-unintentional-co-author-ethical-risks/#Ethical_Implications">Ethical Implications</a></li><li><a href="https://www.enago.com/academy/ai-unintentional-co-author-ethical-risks/#Preventing_AI_from_Unknowingly_Becoming_a_Co-Author">Preventing AI from Unknowingly Becoming a Co-Author</a></li><li><a href="https://www.enago.com/academy/ai-unintentional-co-author-ethical-risks/#Additional_References">Additional References:</a></li></ul><h2>How AI Can Unknowingly Become a Co-Author</h2><figure class="image"><img src="https://www.enago.com/academy/wp-content/uploads/2025/04/Medium.png" alt="How AI becomes a co-author unintentionally"></figure><h3>1. Generative AI Tools Rewriting or Altering Content</h3>Generative AI tools (like ChatGPT or others) and LLMs can do more than just <a href="https://www.trinka.ai/grammar-checker/?utm_source=academy&amp;utm_medium=referral&amp;utm_campaign=article">fix grammar</a>. The tool might:<ul><li>Generate or rephrase key arguments or conclusions in ways that change their original meaning.</li><li>Paraphrase original findings or suggest edits that affects scientific accuracy.</li><li>Generate new insights or edits weren’t a part of the researcher’s initial thought process.</li></ul>If researchers rely too much on AI’s output without critically reviewing it, they may end up incorporating AI-generated content as their own. This gives AI an uncredited and significant intellectual ownership of portions of the paper.<h3>2. Lack of Transparency and Acknowledgment of AI’s Role</h3>Currently, AI and/or LLMs cannot be listed as an author, since it doesn’t take accountability or responsibility for its contributions. However, when researchers fail to disclose AI’s involvement, it might appear that AI played a bigger role than intended. This can happen when:<ul><li>AI-generated text is included without proper acknowledgment (such as in the “Methods” section or other areas).</li><li>The line between human-written and AI-generated content becomes unclear.</li><li>Researchers unknowingly rely on AI’s suggestions or edits, allowing it structure the paper, without acknowledging its role.</li></ul>This could lead to ethical concerns about academic integrity, as the AI is indirectly taking part in the research process without being properly credited.<h3>3. Unintended Creation of New Ideas</h3>Beyond editing, LLMs can suggest alternative hypotheses, interpretations, or conclusions. If a researcher adopts these suggestions without careful evaluation, the LLM is effectively creating new intellectual contributions, a role traditionally reserved for human authors.For example, AI may suggest changes that alter the hypothesis or offer new ways to interpret data. If a researcher then incorporates these changes into their final manuscript without evaluating their validity or accuracy, the AI’s output could become a central intellectual contribution—effectively making it a “ghost contributor.”<h3>4. Influencing the Final Version Without Accountability</h3>As the use of AI grows, there may be instances where researchers rely so heavily on AI tools that the final version of the paper is more reflective of the AI’s suggestions than the researcher’s original thoughts. This can happen if:<ul><li>AI makes substantial structural changes (e.g., reorganizing paragraphs, rewording sections, or altering argument flow).</li><li>Large portions of text are rewritten by AI with minimal human intervention.</li><li>The researcher does not critically assess the changes or its impact</li></ul>At this point, AI isn’t just assisting, it’s shaping the paper in ways that could be seen as authorship-level contributions.<h2>Ethical Implications</h2>Academic authorship comes with responsibility. Authors must stand by their work, ensure its accuracy, and be accountable for ethical concerns like plagiarism or misrepresentation. AI, however:<figure class="image"><img src="https://www.enago.com/academy/wp-content/uploads/2025/06/Medium-1.png" alt="What AI cannot do"></figure>In long run, the use of AI-generated text without proper disclosure can lead to more retractions. For <a href="https://www.researchprofessionalnews.com/rr-news-europe-innovation-2024-10-ai-scientific-publishing-is-not-prepared-for-this/">instance</a>, of more than 10,000 papers retracted in 2023, almost 100 papers were seen to be written using GenAI. <a href="https://pmc.ncbi.nlm.nih.gov/articles/PMC10828852/">Several publishers</a> have emphasized that AI should not be considered a co-author because it lacks accountability for the research, and human authors must take full responsibility for the content. They should also ensure full disclosure of AI use in academic papers, specifically for AI-assisted <a href="https://www.enago.com/copy-editing">copy editing</a>.<h2>Preventing AI from Unknowingly Becoming a Co-Author</h2>To maintain research integrity, researchers should take the following precautions:<ol><li>Transparent Declaration: Researchers must disclose the use of AI tools in the Methods or relevant sections of their paper, as outlined by the ethical guidelines.</li><li>Critical Evaluation: Researchers should critically assess AI-generated content before accepting it, ensuring that it does not alter the original research’s meaning.</li><li>Maintain Human Control: AI should only be used structural editing and text clarification upon proper evaluation of the generated output.</li><li>Authorship Guidelines: Clear guidelines should be followed, ensuring that only human contributors who are accountable for the research are listed as authors. AI tools should be acknowledged for their technical assistance, but not as co-authors.</li><li>Ownership of Ideas: Researchers should maintain ownership of the ideas, analysis, and intellectual contributions in their paper, using AI only to support their work without substituting or overshadowing it.</li></ol>AI is undoubtedly a powerful asset in academic writing, but its role must be carefully managed. As we foresee the future, journals and institutions are expected to update their policies to define AI’s acceptable level of involvement and mandate explicit declarations in research papers. While disclosure norms are still evolving, researchers are encouraged to establish transparency, critical evaluation, and clear authorship practices to ensure that AI remains a tool—not an unintended co-author.<h2>Additional References:</h2><a href="https://publicationethics.org/guidance/cope-position/authorship-and-ai-tools">https://publicationethics.org/guidance/cope-position/authorship-and-ai-tools</a><a href="https://link.springer.com/journal/10462/submission-guidelines?srsltid=AfmBOophhMQb-im36jZ7qRIW6VyvJgBN7m8wEqoxnQYqYOXhWv1hS6R4">https://link.springer.com/journal/10462/submission-guidelines?srsltid=AfmBOophhMQb-im36jZ7qRIW6VyvJgBN7m8wEqoxnQYqYOXhWv1hS6R4</a><a href="https://www.elsevier.com/about/policies-and-standards/the-use-of-generative-ai-and-ai-assisted-technologies-in-writing-for-elsevier">https://www.elsevier.com/about/policies-and-standards/the-use-of-generative-ai-and-ai-assisted-technologies-in-writing-for-elsevier</a><a href="https://pmc.ncbi.nlm.nih.gov/articles/PMC10828852/">https://pmc.ncbi.nlm.nih.gov/articles/PMC10828852/</a>

Is AI Becoming an Unintentional Co-Author? Understanding the Risks and Ethical Concerns

What happens when AI’s role in the writing process becomes so significant that it blurs the line between assistance and authorship? Read now!

AI as Unintentional Co-Author: Risks Ethics | Responsible AI

Explore the latest articles, podcasts, webinars, and lectures on ethical AI use in manuscript preparation to stay informed in this fast-evolving field.

Anagha Nair

Articles Published:

Anagha Nair

1.3 Million Scientists Ask ChatGPT 8 Million Questions a Week: 6 Strategies for Researchers to Stay in Control

Anagha Nair

50% More Papers, Less Science? What Does This Mean For You as a Researcher?

Anagha Nair

Wiley Did a Study on GenAI Use. 3 Things Stood Out.

Anagha Nair

What Counts as “Generative AI Use” in Research Writing? The Definition Problem No One Has Solved!

Anagha Nair

“Looks Cited” Doesn’t Mean “is Cited.” Why 1 in 5 AI References Fail.

Anagha Nair

Only 5.7% Disclosed AI Use! Why Researchers Need Unified Guidelines Urgently

Anagha Nair

Researcher Alert! AI Writing Surges in Research: 1 in 5 Computer Science Papers and Over 1,000 Journals at Risk

Anagha Nair

Is AI Becoming an Unintentional Co-Author? Understanding the Risks and Ethical Concerns