Image Credit: Christina Morillo from Pexels
by the Educator Collaborative Associate Member, Melissa Wicker
It’s no surprise that AI is front and center in the field of education right now. It is hard to sit in a faculty meeting, a conference session, or a curriculum review without it coming up. Most of the conversation, though, seems to be about students: should they use it, how do we catch them misusing it, what does “AI literacy” mean for a ten-year-old. Those are important questions, but there is a quieter one that I think about more because it is the one sitting on my desk each week: what does it mean to prepare teachers to integrate AI into their own professional practice?
That distinction matters. In work in educator preparation, I have come to see AI integration less as a classroom tool and more as an issue of professional judgment. The pre- and in-service teachers I work with are not taking courses to become AI experts. They are becoming literacy experts who happen to have AI available to them, much the same way they have a curriculum guide, an instructional coach, or a colleague down the hall. The job is not to make them technologists. It is to make sure that when, not if, they reach for the tool, they know how to use it ethically and how to use it well.
Where most programs start, and where I think we need to start instead
In many teacher preparation programs, student engagement with AI may not extend past surface level. They get some general practice with a platform and maybe a demonstration of how to differentiate a rubric or differentiate a passage, but the instruction rarely extends past operating the tool. A recent scoping review of AI literacy in teacher education preparation found that teachers’ professional knowledge is still rarely addressed head-on in the literature. Most existing work leans toward adopting AI edtech or teaching about AI rather than building the practical and ethical judgment teachers actually need (Sperling et al., 2024). A survey of preservice teachers in Germany found something similar: what they wanted was not more tool demonstrations but foundational digital literacy training they could actually build on (Heine & König, 2025).
That gap, tool exposure standing in for real, intentional instruction, is useful as a starting point, but it still only teaches teachers how to operate a tool without teaching them to evaluate one. And evaluation is the harder, more durable skill. Technology changes rapidly. Judgement does not.
So in my courses, I try to flip the order. Before we talk about what AI can generate, we talk about what good literacy instruction requires: what a well-differentiated passage actually needs to do for a striving reader, what makes a discussion question generate authentic thinking. Only when candidates can articulate that themselves do we bring AI into the room, and the question becomes does this output meet the bar I just described? Sometimes it does. Often it needs revision, and often the revision itself is the lesson. One study that has stuck with me had teacher candidates use ChatGPT-generated responses, including its mistakes, as materials for practicing critical analysis of student-style problem-solving. The flawed answers turned out to be a genuinely useful training tool, not just something to correct (Drushlyak et al., 2025).
The same principle holds when we use AI to help analyze assessment data. AI can identify patterns quickly, but candidates have to bring the contextual and relational knowledge of the actual student–their history, their language background, what happened the week the assessment was given–instead of taking the AI’s analysis at face value. That gap between what the tool sees and what the teacher knows is the most instructive part of the lesson, and it is where candidates start applying what they are learning in ways that actually matter.
This mirrors a growing body of research about teacher preparation more broadly. Teacher AI competence is not just about knowledge of the tools. One recently developed scale measures professional judgment, ethics, and assessment as core dimensions alongside the technical knowledge (Chiu et al., 2025). An approach gaining traction in teacher education, critical co-discovery, has candidates and instructors exploring AI’s capabilities and limits together rather than candidates simply being handed a tool to practice with; this approach builds more reflective, agentive engagement than passive tool training (Dilek & Baran, 2025).
A framework I keep coming back to
When I am designing instruction or professional development around this, I try to keep candidates anchored in these three questions. First, what am I trying to accomplish instructionally? Not “what can AI do here,” but what is the actual literacy goal. Second, what would a strong version of this look like if I built it myself? This is an often-skipped step, but it makes candidates better evaluators. Third, Where does AI genuinely save time or add value, and where does it just add noise? Sometimes the honest answer is that it does not help! That is a legitimate answer, and candidates need permission to reach it.
I have watched my students go from uncritical enthusiasm (“just generate it for me”) or blanket suspicion (“I don’t trust it. I’ll do it myself”) to something more useful: a working relationship with the tool that treats it like any other resource. It may be helpful in specific situations and unreliable in others, but it is never a substitute for their own instructional knowledge.
Why this belongs in professional practice, not curriculum for kids
I have been intentional about framing AI integration as targeting educator professional practice, not instruction delivered directly to students. Teachers are the ones who need to build judgment about when AI helps and when it gets in the way because they are the ones making instructional decisions all day, every day, under real-time pressures. If we build that judgment within them first, they are far better positioned to make thoughtful decisions about their students later. And, once teachers have that judgment themselves, they are far better equipped to teach it forward. It is difficult to model discernment you have not developed but easier to help a student evaluate AI output once you have spent real time doing that evaluation yourself.
What I would ask other teacher-educators
If you are building AI into your own program, I would push you to resist starting with the tool itself. Start with the instructional standard. Let candidates explain what excellence looks like before you show them what a machine can produce. The tool will keep changing. The standard should not.
I would love to hear how other programs are approaching this. Are you framing AI integration around professional practice, student-facing instruction, or both? What is actually working in your candidates’ practice, not just your syllabus?
References
Chiu, T. K., Ahmad, Z., & Çoban, M. (2025). Development and validation of teacher artificial intelligence (AI) competence self-efficacy (TAICS) scale. Education and Information Technologies, 30(5), 6667-6685. https://doi.org/10.1007/s10639-024-13094-z
Dilek, M., Baran, E., & Aleman, E. (2025). AI literacy in teacher education: Empowering educators through critical co-discovery. Journal of Teacher Education, 76(3), 294-311. http://dx.doi.org/10.1177/00224871251325083
Drushlyak, M., Lukashova, T., Shamonia, V., & Semenikhina, O. (2025). Artificial intelligence in education: ChatGPT-based simulations in teachers’ preparation. Informatyka, Automatyka, Pomiary w Gospodarce i Ochronie Środowiska, 15(1), 144-152. http://doi.org/10.35784/iapgos.6180
Heine, S., & König, J. (2025). Applying artificial intelligence in teacher education: Preservice teachers’ attitudes and reflections in using ChatGPT for teaching and learning. European Journal of Teacher Education, 48(5), 934-963. https://doi.org/10.1080/02619768.2025.2540791
Sperling, K., Stenberg, C., McGrath, C., Åkerfeldt, A., Heintz, F, & Stenliden, L. (2024). In search of artificial intelligence (AI) literacy in teacher education: A scoping review. Computers and Education Open, 6, 100169. https://doi.org/10.1016/j.caeo.2024.100169
