Speakers
Description
With the increasing presence of generative artificial intelligence (GenAI) in learning and teaching practices, English as a Foreign Language (EFL) pedagogy is being fundamentally reshaped. In particular, GenAI has emerged as a core resource influencing students’ approaches to learning and knowledge construction (Johnston et al., 2024; Ravšelj et al., 2025). Recent research indicates that students increasingly rely on GenAI chatbots not only for general writing support but also to resolve lexical problems and vocabulary challenges (Franjić & Božinovski, 2025; Xiao, 2024). Several studies have shown that large language model (LLM)-based chatbots such as ChatGPT can outperform traditional learner dictionaries in receptive and productive lexical tasks, especially in terms of speed and perceived usefulness (Božić Lenard & Šokčević, 2024; De Schryver, 2023; Lew et al., 2024; Ptasznik et al., 2024).
This is not to say that GenAI chatbots offer the same quality of lexicographic data as traditional dictionaries; rather, quality may no longer be a decisive factor for users anymore (de Schryver, 2023). This raises a critical question for lexicography: if dictionaries are no longer viable tools for most users, what role—if any—do lexicographic functions and knowledge still play in GenAI-mediated language use?
While a growing body of research has compared GenAI tools and dictionaries, it has largely focused on English-major students or controlled task-based comparisons (Chen, 2025; de Schryver, 2023; Lew et al., 2024; Mai, 2024; Ptasznik et al., 2024). Much less is known about non-English-major tertiary EFL students—who often lack dictionary literacy and increasingly rely on GenAI as their primary lexical access point. While some recent studies have begun to address GenAI use among ESP/EFL learners outside language-specialist programmes (e.g. Božić Lenard & Šokčević, 2024; Zhong, 2025), research remains limited.
This small-scale, user-centred study investigates lexicographic functions under GenAI mediation rather than dictionary use per se. Conducted in classroom settings in April 2026, it draws on data from approximately 40 non-English-major tertiary EFL students enrolled in English-language courses in Slovenia and Japan, all of whom regularly use GenAI tools and have received no formal dictionary training. The study examines (1) how students self-report their use of GenAI for lexical purposes, (2) how these perceptions align with observed lexical judgement when evaluating short GenAI-like texts containing collocational, patterning, and register-related issues, and (3) whether brief GenAI-mediated vocabulary exploration is associated with immediate uptake of target lexical items in a subsequent paper-based task.
The study adopts a one-session classroom design combining a brief survey (based on Ptasznik & Lew, 2025) with two task-based components (cf. Chen, 2025). First, a lexical judgement task completed without GenAI probes students’ ability to identify and revise near-miss collocations, grammatical patterning errors, and register mismatches, allowing comparison between self-reported confidence and actual lexical evaluation behaviour. This is followed by a time-limited GenAI-for-learning phase, in which students use a GenAI tool of their choice to explore selected lexical expressions, and a subsequent paper-based uptake task assessing their ability to use these items appropriately without tool support. The design enables triangulation of beliefs about GenAI as a lexical resource with observable judgement and short-term uptake, offering insights into the extent to which GenAI functions as a dictionary-like interface in L2 vocabulary learning and use.
The findings are expected to show that, even when dictionaries disappear as explicit tools in everyday writing practices, lexicographic functions remain central but become opaque under GenAI mediation (‘invisible lexicography’). In contemporary EFL contexts, the pedagogical relevance of lexicography may lie in sensitising learners to the lexical dimensions of AI output—such as sense, collocation, patterning, and register—that require human interpretation and evaluation.