When academic language becomes suspicious: GenAI, authorship and linguistic legitimacy

29 October 2026

As Generative Artificial Intelligence (GenAI), particularly Large Language Models (LLMs), becomes increasingly important in higher education (HE), much attention has been given to distinguishing cognitive offloading from authentic intellectual work. A deeper examination of this narrative, however, reveals concerns surrounding authenticity, authorship and how we judge academic writing. A particular interest of this blog is to examine how perceptions of AI-generated language may influence the recognition of authentic intellectual work, especially for learners from different linguistic backgrounds. 

In their work, Leppänen et al. (2025) identify changes to student writing following the emergence of LLMs, including an increased prevalence of words such as 'delve', 'crucial' and 'foster'. While these findings provide insights into changing linguistic patterns, they raise questions about how such characteristics might be interpreted beyond the research context. In this regard, concerns about the heightened scrutiny faced by researchers writing in English as an additional language, alongside Albeihi and Rice’s (2025) examination of how GenAI may reinforce existing linguistic hierarchies, stand out. This raises an important question regarding whether linguistic characteristics are increasingly being treated as markers of intellectual authenticity rather than expressions of academic identity. 

Academic learning has traditionally involved learners developing knowledge and identity through engagement with peers, educators, scholarly materials and their surroundings. Within academia, learners gradually acquire the linguistic conventions of their disciplines, shaping their academic identities beyond their immediate sociocultural environments. In the era of GenAI, however, we may be witnessing a shift in how these linguistic characteristics are perceived. Just as academic disciplines shape the language researchers use, emerging evidence suggests that LLMs may also be influencing scholarly vocabulary and expression (Matsui, 2025; Abramski et al., 2025). This raises an interesting question: could linguistic patterns associated with LLMs eventually become part of accepted academic conventions? 

If I return to my own PhD thesis, written before the widespread availability of LLMs, I find words such as 'delve' and 'crucial' appearing in abundance. These formed part of the academic vocabulary I acquired through years of engagement with scholarship. Does their presence now delegitimise my intellectual work or authorship? If not, why might 

linguistic characteristics previously accepted as part of academic writing now become associated with possible malpractice? 

Even as I write this blog, I find myself increasingly reflexive of my own position as the author. I consciously revisit and change words I would otherwise use naturally, not because they fail to communicate my argument, but to avoid having my authorship questioned. If I am already policing my own academic voice, what does this mean for learners still developing theirs? 

Linguistic identity and the insider–outsider dynamic 

Academic writing has historically been shaped by dominant language ideologies, particularly the privileging of certain forms of English. For learners writing in English as an additional language, academic vocabulary is often consciously acquired through engagement with scholarly texts and disciplinary expectations. Yet, if these expressions become associated with LLM-generated content, we risk introducing another layer of scrutiny into an already unequal linguistic environment. This creates a potential insider–outsider dynamic where intellectual legitimacy may depend not only on what learners communicate, but also on whether their language aligns with educators' expectations of authentic writing. A concern is our own assumptions about learner capability. When written work exceeds what we believe a learner is capable of producing, how do we distinguish legitimate academic concern from our own preconceived perceptions? 

Rethinking authenticity in the age of GenAI 

Educators have a responsibility to uphold academic integrity. However, identifying linguistic patterns is not equivalent to establishing intellectual misconduct. The former offers insights into changing language use. The latter requires careful consideration of evidence, context and intellectual engagement. In an era where AI detection remains contested, we must be cautious about replacing technological uncertainty with subjective judgements of authenticity. In doing so, we risk reinforcing existing linguistic prejudices under the guise of protecting academic integrity.  

If the language we once encouraged learners to acquire is now becoming a potential source of suspicion, what does this mean for academic identity, belonging and the recognition of authentic intellectual work? 

References 

Abramski, K., Improta, R., Rossetti, G., & Stella, M. (2025). The “LLM World of Words” English free association norms generated by large language models. Scientific Data, 12(1), 803. https://doi.org/10.1038/s41597-025-05156-9 

Albeihi, H. H. M. & Mary F. Rice. (2025). Generative AI and Language Diversity: Implications for Teachers and Learners. 16(1). 

Leppänen, L., Aunimo, L., Hellas, A., Nurminen, J. K., & Mannila, L. (2025). How Large Language Models Are Changing MOOC Essay Answers: A Comparison of Pre- and Post-LLM Responses (Version 1). arXiv. https://doi.org/10.48550/ARXIV.2504.13038 

Matsui, K. (2025). Delving Into PubMed Records: How AI-Influenced Vocabulary has Transformed Medical Writing since ChatGPT. Perspectives on Medical Education, 14(1), 882–890. https://doi.org/10.5334/pme.1929

Dr Manesha Peiris: School of Physical and Chemical Sciences

https://www.qmul.ac.uk/spcs/staff/academics/profiles/mpeiris.html