1　Introduction

This chapter introduces a dedication to John Swales, a seminal figure in ESP/EAP renowned for his genre analysis work and pedagogical contributions such as the CaRS model, which have deeply influenced academic writing research and teaching. It highlights Swales’s vision urging EAP to move beyond narrowly circumscribed structural studies toward research with practical pedagogical relevance. The chapter notes the traditional EAP pedagogical approach of “examine and report back,” fostering rhetorical consciousness through analysis of disciplinary texts. It discusses recent disruptions caused by large language models (LLMs) that can perform analytical tasks, correct English usage, and rapidly generate academic texts, alongside the global expansion of English Medium Instruction (EMI), which demands new teacher training and interactional skills. The author outlines their perspective on the evolving focus of EAP research from structural towards dialogic discourse features, situating English use within EMI contexts as a standard language, lingua franca, and element of translingual academic practices. The chapter presents a framework positioning academic English on a continuum from standard to non-standard across normative, socio-ideological, and interactional dimensions, analyzing LLM affordances and constraints in relation to these. By examining the interactional dimension, the chapter illustrates LLM limitations concerning authorial stance in academic writing. It concludes by advocating for EAP’s enhanced emphasis on nurturing authorial voice and fostering spoken interaction among multilingual English users in EMI and academic settings amid advancements in AI and global English use.

2　EAP: From structure to dialogue

This chapter introduced Swales’s approach to genre analysis, which was inspired by Propp’s structuralist narrative analysis identifying key narrative units. Swales applied a similar framework to academic genres, notably through his CaRS model outlining rhetorical moves in research article introductions across disciplines. This structuralist approach evolved within the ESP genre school, establishing textual patterns in academic and professional discourse and significantly aiding novice researchers in engaging with international academic communities. The ESP genre pedagogy initially targeted L2 English users in anglophone STEM fields but has since expanded to various disciplines worldwide, requiring adaptation for local academic and rhetorical traditions, especially in humanities and social sciences where multilingual research practices are common. Swales critiqued English dominance in academia, advocating for the recognition of other languages.

The chapter further discussed the impact of corpus-analytical methods on register variation studies and the creation of the Michigan Corpus of Spoken Academic English (MICASE), which facilitated analysis of spoken academic discourse, including interactive speech events, and enabled comparisons with everyday dialogue. Corpus-based research also emphasized dialogic features of academic writing, focusing on metadiscourse, stance, and engagement to anticipate reader response. Recent ESP research has investigated digital academic genres emerging from Web 2.0 platforms, characterized by hybrid, multimodal texts with explicit writer-reader interactions, increased dialogicity, and tolerance for non-standard English. Multilingualism has grown alongside Web 2.0, although English remains dominant.

Amid these developments, the chapter highlighted the influence of Generative AI, particularly large language models (LLMs), on academic genres and English usage in academic settings, noting emergent EAP research that compares AI-generated academic discourse with human writing. This evolving landscape presents challenges and opportunities for EAP professionals in both research and pedagogy. The chapter emphasized the shift in EAP research focus from the structural features of traditional academic genres toward dialogic discourse characteristics and digital genres driven by technological advances, while also acknowledging rapid changes in the English language due to its expanding L2 user base.

3　The “E” in EAP: Standard English, lingua franca, or translingual practice?

This chapter discussed the evolving roles of English in global higher education, highlighting the shift from English for Academic Purposes (EAP) as language support for international students in anglophone universities toward English Medium Instruction (EMI) in non-anglophone contexts. It presented findings from a systematic review indicating that learning disciplinary content in English enhances students’ academic English skills, particularly receptive abilities and vocabulary, while emphasizing that academic writing remains a persistent challenge requiring ongoing EAP support. The chapter explored the traditional focus of EAP on standard written English (SWE) as a norm derived from native English varieties, often seen as necessary for all learners including L2 users, and noted critiques challenging native norms and the adequacy of standard-based language assessment.

The status of English was reframed through research on World Englishes (WE) and English as a Lingua Franca (ELF), revealing a polycentric view of English with multiple global standards and varieties. World Englishes research examined structural features and societal functions of diverse Englishes beyond postcolonial settings, whereas ELF studies focused on common linguistic features in academic communication among L2 users. Despite shared concerns with variation, synergy between WE and EAP remains limited, with some scholars, such as Shaw, critiquing the elite academic focus of EAP/ELF/EMI studies and urging broader sociolinguistic equity considerations, aligning with calls for wider EMI research agendas.

The chapter introduced translanguaging as an emerging framework that embraces multilingual speakers’ full linguistic repertoires without rigid language boundaries. It addressed how translanguaging facilitates language learning and creativity, citing examples of re-appropriated English forms infused with other language influences, and noted the impact of such practices on academic writing as variably aligning with or diverging from standard language norms. Despite extensive translanguaging research in academic settings, integration with EAP remains minimal.

Further conceptual advancement came from Canagarajah’s spatial repertoire approach, which values the use of multiple semiotic resources—linguistic, visual, paraverbal, and bodily—in communication, emphasizing collaborative meaning-making in context. Canagarajah’s recent work on Communication for Specific Purposes (CSP) applied this lens to research group meetings, showing effective international scholarly communication beyond grammatical proficiency. Mauranen similarly advocated for inclusion of nonverbal elements in Language for Specific Purposes (LSP) research. The chapter concluded by questioning whether the CSP framework will be adopted in future EAP scholarship.

4　Academic English and its dimensions

This chapter discussed differing conceptualizations of academic English, highlighting that its use is not uniform but varies across contexts such as standard language, lingua franca, and translingual practices. Research in Swedish EMI settings revealed that advanced L2 users regulate British English standards while also engaging in informal translingual practices like “Swenglish,” adjusting language based on formal or informal contexts. The chapter proposed a continuum model where both native and non-native English users’ language practices fluctuate between standard and non-standard forms, influenced by contextual norms and shifting trends. The rise of AI-generated language introduces more standardized linguistic outputs coexisting with human use.

Language was framed along three dimensions: normative, socio-ideological, and interactional. The normative dimension treats language as institutionalized codes defined by grammar, vocabulary, and phonology, and is prominent in language policies and academic texts. The socio-ideological dimension relates to users’ perceptions of language reflecting specific worldviews shaped by cultures, disciplines, or professional groups, connecting with notions of discourse communities and disciplinary epistemologies. The interactional dimension concerns how utterances respond dialogically to others, expressing individual voices and stances, a focus of academic writing studies emphasizing authorial positioning.

Finally, the chapter noted that large language models (LLMs) like ChatGPT perform well on the normative dimension, managing standard language forms effectively. However, these models are less capable in handling the socio-ideological dimension’s nuanced disciplinary worldviews, while grappling most with the interactional dimension’s dialogic and interpersonal subtleties. Further elaboration on these challenges is reserved for the next section.

5　Large language models and what they mean for EAP

This chapter discussed the transformative impact of large language models (LLMs) like ChatGPT, Claude, Copilote, and DeepSeek across various tasks, from everyday language checks to complex academic assignments such as data analysis and literature reviews. Initial concerns within the EAP field regarding writing instruction have evolved into a nuanced understanding of LLM affordances and limitations. LLMs effectively replicate structural features of human communication, including genre conventions and lexicogrammatical patterns common in academic writing, and provide quick access to less visible genres like peer review, benefiting novice researchers. They also hold potential for addressing epistemic biases by amplifying underrepresented voices. However, limitations arise from their reliance on statistical probabilities and syntactic patterns rather than true semantic understanding, leading to outputs approximating standardized English varieties, especially US English, and diminishing sensitivity to register, authorial voice, and stance. Socio-ideological constraints stem from biased training corpora, as early data sources reflected primarily educated, white, anglophone, male perspectives, with subsequent models still skewed by data selection and hidden prompts that may produce flattened, non-controversial argumentation. Consequently, the use of LLMs risks enforcing standardized knowledge-making practices, potentially fostering "epistemic monoglossia" by privileging English language and its associated frameworks over epistemic diversity. Despite these challenges, human expertise remains crucial in crafting academic texts that involve critical analysis and original perspectives, highlighting areas where EAP’s extensive research can inform novel pedagogical approaches.

6　What LLMs struggle with: The case of authorial stance

This chapter described the challenges large language models (LLMs) face in accurately producing authorial stance, a key element of authorial voice in academic writing. It highlighted that early iterations of LLMs, like ChatGPT, often generated text with inappropriate tone or style, such as promotional language creeping into academic discourse. Evaluative adjectives and reporting verbs in LLM-generated student essays were noted as inadequate or repetitive, revealing a stylistic mismatch with academic norms. The chapter emphasized that authorial voice evolves socially within academic disciplines and that stance expressions—linguistic markers conveying personal attitudes and assessments—are central to knowledge construction through writer-reader interaction.

The widespread use of stance expressions in academic writing, including hedges, boosters, and attitude markers, facilitates dialogic engagement by allowing writers to position themselves relative to existing knowledge claims, opening or closing discursive space. Recent research demonstrated increased use of stance over time, particularly important as researchers promote work within the modern "attention economy." Comparative studies between student essays and LLM-generated argumentative writing revealed that student texts tend to be more interactive and persuasive, while LLM outputs struggle with the complexity of authorial stance, which involves nuanced attitudes toward both personal and other researchers' claims.

A study by Kuteeva et al. investigated how ChatGPT-4 handles stance through different prompting techniques: zero-shot (no examples), few-shot (examples given), and role enactment (acting as an expert researcher with critical evaluation steps). Findings showed that ChatGPT often treats stance as a surface linguistic feature, leading to over-hedging, unsolicited content, or overstated claims, with role enactment prompts producing slightly better, though imperfect, results. The model’s reliance on statistical probability mismatches the socio-ideological and interactional dimensions intrinsic to human academic writing.

The chapter pointed out ChatGPT’s rigid and limited understanding of stance, demonstrated by its persistent overuse of hedging and inconsistent boosting of claims, which contrasts with the observed trend among human researchers to boost rather than hedge claims. Notably, LLM outputs for attitudinal stance sometimes included overly confident, non-academic expressions, highlighting register mixing. This finding aligns with research identifying rhetorical overstatement as a characteristic of problematic retracted STEM articles, where exaggerated knowledge claims misrepresent epistemic stance. These tendencies suggest that LLM-mediated revisions may hinder writers seeking to strategically enhance their research impact rhetorically.

Stance was reaffirmed as a complex discursive phenomenon deeply tied to authorial voice, positionality, and disciplinary conventions, requiring explicit teaching for both native and non-native academic writers. Mastery of stance demands subtle linguistic knowledge, contextual awareness, and the ability to imply nuanced meaning. Pedagogical interventions remain crucial in this area.

The chapter also discussed the growing need for critical GenAI literacy development alongside traditional academic writing instruction. It referenced a study by Ou et al. exploring how PhD students used LLMs to enhance text readability and publishability, mostly employing zero-shot-like prompts. Students often followed up with iterative cues to refine tone and evaluation, illustrating the necessity for teaching prompt design and critical engagement with AI-generated text. Ou et al.’s micro-curriculum was proposed as a valuable foundation for developing such pedagogical approaches aimed at increasing students’ understanding of LLM affordances and limitations.

7　Implications for EAP in the age of EMI and AI

This chapter discussed how the evolving educational and technological landscape impacts English for Academic Purposes (EAP). Despite English serving as a global academic lingua franca and medium of instruction in many regions for over a decade, EAP research has remained heavily focused on written academic genres within anglophone contexts and largely reliant on corpus-based methods. There has been limited exploration of English Medium Instruction (EMI) classroom interactions, although recent studies have started to address this gap. Traditionally, EAP’s research on structural and lexicogrammatical features has aimed to help students conform to English standards typical in anglophone countries. With the rise of advanced large language models (LLMs) capable of automated corrections and translations, the chapter emphasizes the potential for EAP to shift toward examining spoken interactions and dialogic features in EMI settings, such as utterance-level and interactional dynamics.

Increasing interest in dialogic aspects of academic discourse, which appear even in conventional written genres, suggests a logical progression to focus more on dialogic interaction. Questions arise about how disciplinary discourse, academic registers, and everyday speech merge in EMI lectures and discussions, opening new research avenues. While LLMs significantly enhance writing support, EAP practitioners still hold a unique role in leveraging students’ socio-material contexts, identities, and multilingual repertoires. The chapter highlights opportunities for research and pedagogical innovation aimed at nurturing individual authorial voice, genre experimentation, and critical evaluation of knowledge. Collaborative writing and peer review are proposed as effective means for language support and data generation. Finally, the development of critical generative AI literacy among students is identified as crucial for increasing awareness of these technologies’ capabilities and limitations, ultimately reinforcing EAP’s relevance in higher education.