Abstract
Generative artificial intelligence (GAI) has significantly transformed language education. With its rich affordances, it has made second language (L2) teaching, learning, and assessment more efficient, effective, and accessible (Chapelle, 2025; Hyland, 2026; Zou et al., 2025). As reported in several systematic reviews, the technology has enhanced the teaching, learning, and assessment of L2 skills, including listening, speaking, reading, and writing, as well as vocabulary and grammar (e.g., Crompton et al., 2023). Moreover, according to various meta-analyses, the tool has had a positive impact on L2 proficiency and affective-cognitive outcomes, particularly in the development of productive skills (e.g., Dong, 2026; Saarela et al., 2026). Lastly, based on many empirical studies, the tool has facilitated improvements in L2 learners’ motivation and active engagement (e.g., Kohnke and Moorhouse, 2025; Zapata, 2026).
In some contexts, GAI has efficiently facilitated the implementation of English language teaching (ELT) approaches, such as the process approach to teaching writing, content and language integrated learning (CLIL), multimodality, multiliteracies, translanguaging, English as a medium of instruction (EMI), task-based language teaching (TBLT), textual enhancement, and digital multimodal composing (DMC). In this issue of RELC Journal, you will see how these approaches were implemented in specific contexts. Though these approaches seemed to work well without the use of GAI (see the articles in this issue), several studies claim that GAI tools can effectively enhance their implementations in language classrooms, thus yielding more impactful outcomes (e.g., Ci and Jiang, 2025; Li et al., 2025; Ramirez-Aroca and Javadinejad, 2026; Tang and Honeycutt, 2026; Tzirides, 2026).
In addition to effectively supporting ELT pedagogies, GAI has also facilitated language assessment practices and research processes. As reported in recent systematic reviews, GAI can create assessment tasks and automatically evaluate extended responses (Chapelle, 2025). It can also support the implementation of assessment approaches such as dynamic assessment and adaptive assessment (Zapata, 2026). Lastly, it can provide instant, unlimited, and actionable feedback that improves writing quality (Crosthwaite and Sun, 2026; Hyland, 2026). In terms of research and knowledge production, GAI can expedite research and paper writing while adhering to frameworks that underscore transparency, methodological rigor, ethics, social value, and human accountability (Floris and Renandya, 2025; Moorhouse et al., 2026).
As shown above, GAI presents positive impacts on L2 learning, teaching, assessment, and research. However, the findings reported in these studies are inconclusive, as outcomes vary with evolving conditions and contexts. Similarly, while GAI's impacts on assessment are promising, challenges remain regarding the validity, reliability, fairness, and practicality of GAI-generated assessments, as well as the validation of the accuracy, fairness, and transparency of AI-generated scores and feedback (Chuang and Yan, 2025; Zou et al., 2025). In research and knowledge production, due to many reported cases of irresponsible and non-transparent use of GAI at various stages of research, public trust in the accuracy and reliability of research findings has become fragile (Bernard, 2026; Zapata, 2026). According to reports, scientific literature is being flooded with synthetic, AI-generated papers (Bernard, 2026). Similarly, reports also claim that more than half of new articles on the internet are written by AI and that the whole digital ecosystem is inundated with AI slop—a low-quality, repetitive, misleading, and mass-produced digital content generated by AI, including junk images, deepfakes, and fake news, among others (Agnellini, 2025). This type of content constitutes so-called digital plastics, causing long-term pollution in our digital ecosystems (Roe et al., 2025). Since many people, particularly young learners, are not particularly adept at identifying and addressing digital plastics, there is an urgent need to develop emergent multiliteracies to mitigate potential risks to (language) education and societal development.
AI pollution and its impact on language education
AI pollution refers to the contamination of digital environments (e.g., internet, social media platforms, and AI-powered apps) with massive volumes of low-quality, generic, and misleading AI-generated content, also known as AI slop (Agnellini, 2025; Roe et al., 2025). This type of content, presented in various forms such as texts, images, and videos, appears authentic but is actually synthetic, offering no genuine value or unique insight. As a digital pollutant, it can cause massive and long-term devastation in the digital environment. It can also cause “model collapse” when it is repeatedly used as training data for newer generations of AI. Model collapse occurs when an AI degenerates into producing low-quality outputs over time, relying on its previous sloppy outputs. When AI pollution persists, public trust in the digital environment will erode, and the environment itself will no longer serve its purpose. According to Roe et al. (2025), this problem is caused by several factors, including cognitive, socioeconomic, and sociopolitical factors. It is also influenced by the evolving culture of instant gratification and the growing demand for “cognitive junk food” amid an increasingly fast-paced, stressful lifestyle.
How does AI pollution adversely affect language learning, teaching, assessment, and research? In L2 learning and teaching, the use of low-quality AI-generated texts may lead to misinformation, mediocrity, and the erosion of learners’ trust and engagement with online content and learning activities. In L2 assessment, low-quality AI outputs can also introduce inaccuracies, undermine the authenticity of tests, raise ethical concerns about fairness and validity, and lead to an inaccurate evaluation of learners’ true language proficiency (Chuang and Yan, 2025). GAI's inaccurate and incomplete feedback can also misinform and mislead learners, leading to disengagement with the tool (Henderson et al., 2025). In research, the use of synthetic, AI-generated papers may obscure the literature with unreliable findings. Unconscious or deliberate use of AI-generated papers as references can mislead ongoing research, perpetuate inaccurate knowledge, paradigms, and arguments, weaken the interpretation of findings, cause harm and frustration, and destroy the integrity of scientific inquiry (Bernard, 2026).
Combatting AI pollution through enhanced multiliteracies education
With the massive proliferation of AI slop in digital environments, people must develop multiliteracies, a set of skills involving diverse and complex processes of meaning-making and representation that take various forms or dimensions (Kalantzis and Cope, 2023; Meniado, 2026b). In the AI era, highly important forms of multiliteracies include multimodal literacy, digital literacy, media literacy, information literacy, critical literacy, and AI literacy. Multimodal literacy refers to the ability to understand and represent meaning across multimodal texts by integrating a range of modes, each using semiotic resources that carry socially or culturally constituted meanings (Lim and Tan-Chia, 2023). A closely related dimension is digital literacy, which refers to the ability to access, manage, understand, integrate, communicate, evaluate, and create information safely and appropriately through digital technologies for various purposes (Meniado, 2026b).
Other forms of multiliteracies include media literacy, which is the ability to produce meaningful content and critically analyze, interpret, and evaluate the credibility of messages available from traditional and new media (e.g., social media, news sites), and information literacy, which is the ability to search, evaluate, use, and create information effectively and responsibly to achieve specific goals (Meniado, 2026b). Another equally important form is critical literacy, which refers to the ability to question and spot inaccuracies, biases, or risks in a particular text. Last, and perhaps the most important, considering the worsening problem of AI pollution, is AI literacy, which refers to a set of competencies that enable individuals to interact with AI technologies, encompassing both technical proficiency and socio-emotional competencies (Meniado, 2026a). This definition has expanded to include criticality in AI literacy, giving rise to the concept of critical AI literacy (CAIL) in response to the growing adverse impacts of AI pollution on various aspects of human life. According to Roe et al. (2025), CAIL refers to “the ability to critically analyze and engage with AI systems by understanding their technical foundations, societal implications, and embedded power structures, while recognizing their limitations, biases, and broader social, environmental, and economic impacts” (2). It also includes the ability to identify and evaluate AI content in the digital wilds, as AI outputs leave enduring footprints online.
Integrating or explicitly teaching multiliteracies, particularly (critical) AI literacy, is important to ensure that learners become critical consumers and responsible producers and distributors of AI outputs in digital environments. As AI technologies have advanced to the point where humans already struggle to distinguish what is real from AI-generated imitations, responsive policies and programs across different layers of educational systems, industries, and societies must be implemented. We cannot rely solely on technology (AI detection tools) to save us from this digital mess, as research shows that using technology alone to detect AI misuse can be unreliable and risky (Roe et al., 2025). We need a multi-layered, multisectoral approach to fostering responsible consumption, production, and distribution of AI outputs within the digital ecosystem (Chapelle, 2025).
At our level as language classroom practitioners, what can we do to help address the problem? Perhaps we can start by developing/enhancing the above-mentioned literacies within ourselves. We cannot give what we do not have, and we cannot teach what we do not know. To integrate multiliteracies in our language classrooms, we can adopt an established framework suitable to our context—for example, the multiliteracies pedagogy developed by the New London Group (Kalantzis and Cope, 2023; Meniado, 2026b). The framework is designed in four pedagogical steps, with each step connected with a corresponding knowledge process: (a) situated practice (experiencing); (b) explicit instruction (conceptualizing); (c) critical framing (analyzing); and (d) transformed practice (applying) (see Kalantzis and Cope, 2023: 9). We can integrate these pedagogical steps into the stages of our teaching approaches (e.g., CLT, TBLT, the process approach, CLIL). There are many things that we can do, but, at each stage, we must explicitly develop relevant forms of multiliteracies (e.g., digital, AI, multimodal, critical). We can guide our students in analyzing the elements, purpose, and intended meaning of a text, and in critically evaluating its authenticity, effectiveness, limitations, biases, and ethical issues. We can also guide them in using AI tools responsibly to produce and distribute their original (multimodal) texts, as they apply what they learned from the input texts. There are many possible activities. With collaboration and exploration with our fellow teachers, I am certain that we can discover more possibilities.
Reflections and future directions
The proliferation of AI slop, which causes AI pollution in our digital environment, is becoming entrenched in our cultural, socioeconomic, and sociopolitical realities and will persist for some time. This problem, which extends beyond the digital sphere, is affecting educational ecosystems worldwide, with particularly adverse impacts on language learning, teaching, assessment, and research. A problem of wide magnitude, it requires systemic, multi-layered solutions from various actors and sectors of society. Macro-level solutions should start by addressing the main reasons why this problem exists—epistemic, cultural, social, and economic inequities. Government policies and programs should aim to narrow these gaps. Enhanced multiliteracies education with a strong emphasis on developing critical AI literacy should also be integrated into the language curriculum to produce more critical and responsible AI users. Close collaborations between institutions and AI watchdogs can also be established to clean up AI slop to ensure a safe and sustainable digital environment.
At the meso-level, educational institutions should establish their contextualized AI frameworks to guide teachers in selecting, using, and evaluating AI tools and AI-generated materials for instructional, assessment, and research purposes. Responsive and mission-oriented AI governance in schools can promote sustainable educational outcomes. Training language teachers in various forms and modes (Meniado, 2023) to explicitly teach emergent digital multiliteracies, particularly critical AI literacy, in their language lessons can also ensure sustained efforts to produce AI-literate learners. The SEAMEO Regional Language Centre in Singapore is responding to this need by offering a new course on teaching digital multiliteracies for English teachers in the Southeast Asian region starting next year. At the classroom level, teachers can exemplify how to apply critical AI literacy in real-life situations while using human-centric pedagogies to teach language skills. They can also serve as innovators by exploring and experimenting with new pedagogies in teaching multiliteracies, particularly critical AI literacy.
In this issue of RELC Journal, you will find articles on innovative ELT pedagogies, perspectives, and resources that offer possibilities for integrating the development of emergent digital multiliteracies among L2 learners. You will see the research article by Ryu et al. (2026) illustrating how machine translators can facilitate the foreign language writing process, and the research article by Kruawong and Imsa-ard (2026) describing the different CLIL assessment practices of Thai teachers in English-as-a-foreign-language (EFL) contexts. Moreover, you will come across the research paper by Liu and Lim (2026), which discusses English language teachers’ beliefs about teaching multimodal literacy and the factors that affect these beliefs, and the paper by Romanowski (2026), which presents Chinese EFL teachers’ perceptions of translanguaging in an EMI context. Further, you will also find Zeng and Cheung's (2026) article explaining the effects of educational stage and teaching experience on foreign language teaching anxiety (FLTA), and Lee and Thomson's (2026) paper discussing the effects of textual enhancement on reading fluency and reading comprehension. To round up this section of research papers, there are articles by Wu et al. (2026) about learners’ perceptions and experiences of swearing in English, by Hsu and Chang (2026) describing the effectiveness of online courses on elementary students’ language abilities and anxiety, and, finally, Peng and Shi's (2026) article about the use of rhymes in textbooks for young learners.
As you read further, you will also see the innovations in practice report by Aubrey and Philpott (2026), showing how a reflection activity can enhance a teacher's self-confidence and implementation of the TBLT approach in a lesson cycle, and the report by Liu and Lam (2026) discussing the components, processes, and impacts of a gradeless DMC assessment practice. Along with these reports are the thematic review papers by Malmström and Zhou (2026), describing the current state of EMI teacher collaborations, including their various forms, issues, challenges, and strategies for addressing them, and Savski's (2026) paper discussing the five principles of reflective translanguaging in Southeast Asian ELT. In this issue, you will also find the technology review article by Hong and Lin (2026), describing and evaluating Cathoven, an AI assistant that can help you make language inputs comprehensible and provide targeted feedback on your students’ outputs, and the “Conversations with experts” article by Dovchin and Marlina (2026), discussing language and injustice. Lastly, you will see four book reviews: Chanh's (2026) review of Introduction to Instructed Second Language Acquisition (3rd edition); Liu's (2026) review of Artificial Intelligence and Language Teaching; Akbarian and Hosseini's (2026) review of Researching Incidental Vocabulary Learning in a Second Language; and Moodie's (2026) review of Professionalising English Language Teaching: Concepts and Reflections for Action in Teacher Education.
All the above-mentioned articles present practices and resources that can be further enhanced through the responsible use of GAI and the deliberate integration of multiliteracies in language education. I hope that you will find the articles inspiring and transformative. I also hope that, as you adopt or adapt the pedagogies, perspectives, and resources presented in the articles to your context, you will integrate and explicitly teach emergent digital multiliteracies, particularly critical AI literacy, among your learners. Lastly, I hope that you will explore and experiment with new approaches in addressing AI pollution in your contexts and share outcomes and insights in the next issues of RELC Journal. As classroom practitioners and researchers, let us all work together to develop innovative ways of combating AI pollution in all its forms, ensuring a safe and sustainable digital environment for our learners and future generations.
