A twelve-week teacher-education course raised self-reported GenAI literacy across five dimensions

Knowing how to obtain a fluent AI response is only one part of being ready to teach with generative AI. Ting Huang, Chenze Wu and Meixin Wu examined a course that treated literacy as a combination of technical proficiency, critical evaluation, communication, creative application and ethical competence. Their intervention placed these dimensions inside ordinary English teacher preparation, rather than offering a one-off demonstration of a tool. The central question was whether repeated, guided use would change how future teachers understood their own capacity to use AI responsibly.

The study took place in a compulsory English curriculum and pedagogy course at a university in southeastern China. Four existing classes participated: two received the intervention and two followed the regular curriculum. Ninety complete responses were analyzed, with 45 in each condition. The twelve-week course ran from September to December 2024. Researchers used pre-course and post-course questionnaires, including a 25-item literacy scale and open-ended responses. Analysis compared outcomes after accounting for baseline scores, alongside qualitative coding of students' explanations.

The intervention integrated GenAI into multiple teaching topics and included explicit prompting practice, evaluation of outputs and discussion of ethical issues. KIMI was used in tasks involving lesson plans and presentation materials. Students first created a lesson plan and supporting slides independently, then later used AI to improve them and reflected on the process. They also recorded their interactions and explained how the tool had contributed. This sequence makes a useful distinction between possessing a finished artifact and examining the decisions behind it.

The intervention group reported higher post-course overall literacy than the comparison group, with corresponding advantages across all five dimensions. Open-ended responses added examples of more considered use, including adapting prompts and evaluating teaching applications. The reported improvement should nevertheless be described as a change in questionnaire-based literacy. Even a reliable self-report scale does not directly establish that someone can identify a fabricated source, protect student information or design an effective lesson when independently assessed. Confidence and awareness are relevant outcomes, but they are not interchangeable with demonstrated competence.

Several design features restrict broader claims. Classes were assigned as intact groups, and participation came from one institution with a predominantly female cohort. Researchers helped develop and implement the intervention, so local teaching expertise and expectations may have influenced outcomes. There was no delayed post-test, and open-ended survey responses could not be followed up through interviews. The study therefore supports further evaluation of sustained literacy training without establishing a universal causal effect, durable competence or benefits for the pupils these future teachers may eventually teach.

For Hong Kong teacher education, the practical contribution is an inspectable training sequence. Programs could combine the five dimensions with short performance tasks: check a questionable claim, improve a weak prompt, identify a privacy problem and justify an adaptation for a particular class. An independent lesson design before and after training could complement self-ratings, while a later placement review could examine transfer. These are proposed additions beyond the reported study. The course offers a useful starting structure, provided the evaluation distinguishes what participants believe they can do from what their teaching decisions actually demonstrate.