The AI Trap in L2 Research: Theory First, Technology Second
Willy A Renandya, 17 Nov 2026
Artificial intelligence is rapidly reshaping the ELT field, creating fresh opportunities for vocabulary development, speaking practice, and growth in reading and writing skills. Although it is exciting to jump into this fast-paced, AI-driven research, many studies share a common weakness: they begin with the technology rather than with an understanding of how languages are genuinely learned and taught.
To produce impactful research, you must move beyond surface-level questions like “Do students like ChatGPT?” or “Does AI raise test scores?” Instead, established second language acquisition (SLA) theories and pedagogical frameworks should guide you to examine how, why, for whom, and under what conditions AI actually supports language learning.
Eight Ways Theory Can Strengthen Your Research
Using established second language acquisition (SLA) theories and teaching frameworks helps you move past simply testing a new gadget and focus on how people actually learn languages. Here are eight simple ways these frameworks can make your research questions stronger and your studies more effective
- Clarify the Role of Language Input: AI generates endless reading and listening texts, but more input does not automatically produce learning. Input-based theories remind us that materials must be meaningful, comprehensible, and matched to a student’s proficiency level. Instead of asking if ChatGPT improves reading, ask whether proficiency-adjusted AI texts promote vocabulary acquisition through repeated exposure.
- Understand How Learners Process Language: Instant AI writing feedback is helpful, yet receiving corrections does not guarantee acquisition. Drawing on the Noticing Hypothesis, investigate whether AI feedback helps learners actively notice grammatical errors and apply those corrected forms accurately in later drafts.
- Promote Learner Independence: Instant answers can easily create dependency rather than autonomy. Using Self-Regulated Learning theory, examine whether AI tools help students set goals, monitor progress, and evaluate feedback, rather than merely asking if they enjoy using the tool.
- Focus on Meaningful Interaction: Chatting with a bot does not automatically build speaking proficiency. Guided by the Interaction Hypothesis, study how AI conversations encourage learners to negotiate meaning, request clarification, and modify their output during communication.
- Separate Performance from Actual Learning: A polished, AI-assisted essay proves what a student can do with scaffolding, not what they have internalized. Use delayed post-tests and unassisted writing or speaking tasks to measure lasting gains in accuracy, fluency, and complexity.
- Address Diverse Learner Needs: Classrooms contain learners with varying goals and proficiency levels. Using frameworks like Differentiated Instruction and Vygotsky’s Zone of Proximal Development, evaluate whether AI-adapted tasks provide the right balance of challenge and support to deepen comprehension.
- Support Teacher Development: Research must look beyond students to examine how educators evaluate AI outputs and make sound instructional choices. Frameworks like TPACK (Technological Pedagogical Content Knowledge) help you investigate how professional development builds better teachers, not just tech-savvy ones.
- Evaluate Teaching Materials Critically: Speed of generation does not equal pedagogical quality. Apply materials-development principles—such as Nation’s Four Strands (balancing meaning-focused input, meaning-focused output, language-focused learning, and fluency development)—to assess whether AI worksheets are culturally relevant, engaging, and aligned with lesson goals.
Avoiding the “Perception-Only” Trap
Many early studies rely solely on surveys measuring whether teachers and students find AI useful. While gauging user acceptance is a valid starting point for emerging tools, perceptions alone cannot prove that learners acquired new vocabulary, improved grammar, or gained communicative competence. Strong research bridges the gap between technology acceptance and measurable language development.
In addition, people often rate a new tool highly simply because it feels fresh and exciting, not because it actually improves their English over time. To avoid this common trap, researchers must look past simple user satisfaction and include real language tasks, writing samples, or formal tests to assess true learning progress.
A Checklist for Your Research Design
Before finalizing your next research proposal or manuscript, ask yourself:
- Which established SLA theory or pedagogical framework anchors my study?
- What specific learning or teaching process is AI expected to facilitate?
- How is AI embedded into meaningful classroom activities?
- Does my design account for differences in learner proficiency, goals, and needs?
- Does the study examine how teachers make informed pedagogical decisions?
- Are the AI-generated materials appropriate and pedagogically sound?
- What concrete evidence shows actual learning or teaching improvement?
- Can students demonstrate these gains independently without AI assistance?
Next Steps
AI is a remarkable tool, but technology alone cannot explain how people acquire a second language. Much less how technology can maintain and extend language acquisition. Grounding your work in SLA theory ensures that you ask sharper questions, design rigorous studies, and interpret your data responsibly. We should keep pedagogy and learning theory in the driver’s seat, and let AI serve as the vehicle.
