A school leader introduces an AI-supported feedback process. One teacher asks how the generated comments will be checked. Another questions whether the process adds more work than it removes. A third does not want to try it.
Calling all three teachers “resistant to AI” supplies a label. It does not identify the next useful action.
Teacher resistance to AI can involve the technology, the proposed task, the conditions of introduction or a combination of these. School leaders need to establish what a teacher is questioning before choosing clarification, practice, a change to the proposal or another response.
The title's two possibilities are therefore not mutually exclusive. A teacher can be interested in AI and opposed to a particular use. A poorly explained introduction can coexist with a genuine learning need. An objection can also be mistaken.
The work begins with understanding the claim.
What research can contribute
A 2026 study by Daniela Marzano and Sara Pautasso examined responses from 898 Italian teachers in preservice training. The voluntary online survey was conducted between January and June 2025 among participants in qualification programmes. Their accounts combined interest with concerns about learning, assessment and professional practice.
The study's composition, self-report measures and substantial non-response to a fear-related question limit interpretation. It does not establish the prevalence of resistance among all Italian teachers, and its terminology should not become a diagnosis applied to colleagues.
A separate study by Habimana and colleagues, published in May 2026, surveyed 120 secondary mathematics and science teachers in Rwanda and interviewed 18 teachers. Participants' accounts included perceived usefulness alongside concerns about infrastructure, workload, training and professional judgment.
The survey population had prior exposure to educational technologies, and interviewees were selected for relevant AI experience. These are contextual perceptions and associations, not independently measured gains in student learning or a causal test of an implementation method.
Together, the studies support asking more specific questions. They do not tell a principal what an individual teacher believes.
The approach below is an editorial proposal for that conversation.
Establish what is actually being rejected
Consider a fictional science department discussing an approved AI-assisted process for drafting feedback.
A teacher objects that a generated comment attributes understanding that the pupil's response does not demonstrate. The department leader hears hesitation and offers another demonstration of the interface.
The demonstration may address a skill the teacher does not lack. The unresolved question concerns the relationship between the pupil's work and the feedback.
Before deciding on a response, ask the teacher to identify the proposed action they question.
Is it using AI for any professional purpose? Using it for this task? Sharing particular information? Allowing a draft to reach a pupil without sufficient review? Taking on a new checking responsibility under the current workload?
These positions differ even when each produces the same initial answer: “I do not think we should do this.”
The scenario is fictional, not a reported school deployment, a product evaluation or a Learning Coach client case.
Ask for a concrete example
An abstract argument about whether AI is good for education can continue without resolving the proposal in front of the team.
Ask for an example that makes the concern assessable.
In the fictional department, a prepared pupil response might state a correct scientific term without explaining the mechanism. The generated feedback might praise an explanation that is absent. The teacher can point to the mismatch.
That observation does not prove that every generated comment will fail. It does give the team something to examine.
Use synthetic or otherwise appropriately authorised material. A discussion about a concern should not require staff to circulate identifiable pupil work through additional systems.
The leader can then check whether the example accurately represents the proposed process. Perhaps the teacher believes comments will be sent automatically, while the proposal requires review. Perhaps review is intended but nobody has allocated time for it.
Clarifying the proposal may resolve one issue and reveal another.
Distinguish a fact, a prediction and a value judgment
Concerns can contain different kinds of claim.
“The process sends comments without review” is a factual claim about how it operates. Someone should establish whether it is accurate.
“Reviewing every comment will take more time than writing feedback directly” is a prediction about workload. It needs examination with suitably comparable work.
“Teachers should retain responsibility for the feedback pupils receive” expresses a professional position. The discussion needs to establish how that responsibility would operate in practice.
Treating all three as fear of change misses their different evidential needs.
Likewise, a confident claim in favour of adoption deserves scrutiny. “This will save time” is also a prediction until the relevant process has been examined. A successful demonstration does not establish the overall burden or educational value.
Ask both the proposer and the objector what supports their view, what remains uncertain and what information would help.
Do not prescribe training for every disagreement
Some difficulties are learning needs. A teacher may not understand what the proposed assistance does, how to inspect its output or where to obtain help.
Relevant explanation and supported practice can address those gaps. Practice should use suitable material and make its purpose clear.
Other difficulties require a decision.
If the teacher understands the tool but the school has not established who checks the comments, another training session will not assign responsibility. If the information conditions remain unresolved, confidence with the interface will not settle them.
In the fictional department, the teacher might demonstrate that they can produce and inspect a draft while still questioning whether the process is worth using. That is a substantive judgment to examine, not necessarily a failure to learn.
Conversely, a claim of professional concern should not prevent reasonable questions about understanding. Ask the teacher to explain the difficulty precisely enough that the team can respond.
The aim is to match support to the problem.
Make a limited inquiry possible
Where the proposal is appropriate to investigate, a small, supported examination can be more useful than a contest of opinions.
For the fictional feedback process, colleagues could compare prepared examples. What does the draft get right? Which statements need correction? How much checking is required? Does the final comment identify a useful next step for the pupil?
Agree beforehand what the inquiry is intended to clarify. It might concern review effort or the kinds of unsupported inference that appear. It should not quietly become a trial of the teacher's enthusiasm.
Explain who will see the material and how the findings will be used. Do not promise confidentiality or freedom from consequences beyond what the school can actually provide.
Some proposals require additional assessment before even a limited inquiry can proceed. Existing school procedures and relevant expertise should determine those conditions.
This article offers no universal authorisation to test an AI system with pupils or their information.
Allow the proposal to change
Listening is difficult to take seriously if adoption in the original form is the only acceptable conclusion.
Suppose the fictional team finds that the assistant creates useful starting points for some comments but requires extensive correction for others. Several responses remain open.
The department could narrow the proposed use. It could revise the inputs or review arrangements. It could decide that the additional work does not justify the benefit. It could also identify a misunderstanding and continue with a clearer explanation.
An objection can improve a proposal without defeating its wider purpose. Equally, a decision not to proceed can be a reasonable result.
Our guide to turning school AI policy into classroom practice examines the broader organisational work. The immediate conversation needs to reveal which part of that work is missing.
Do not ask the teacher to carry every unresolved responsibility simply because they noticed the problem.
Keep expectations and professional disagreement clear
Schools can have legitimate requirements for staff preparation and agreed practice. Professional disagreement does not make every expectation optional.
Leaders should explain what has been decided, which matters remain open and what participation is reasonably required. A request to understand a proposed process is different from a requirement to use it in every task.
If someone declines an agreed learning activity, establish the reason. Access, scheduling, uncertainty about the activity and unwillingness to engage call for different discussions.
Employment obligations and formal concerns belong within the school's appropriate processes. A general article cannot determine how an individual situation should be handled.
The useful principle is precision. Staff should not have to infer whether a meeting is consultation, training, a decision announcement or an assessment of performance.
Clarity supports an honest conversation even when the eventual answer is not the one a participant wanted.
Notice whose concerns receive attention
A confident advocate may describe benefits fluently. A less confident colleague may struggle to formulate an equally important difficulty.
Do not confuse ease of presentation with the strength of the evidence.
Offer ways to raise questions before a group meeting, and use a suitable private route where necessary. Ask whether staff in different subjects, roles or working arrangements encounter different conditions.
Avoid inferring a position from age, career stage or a label such as “digital native.” An experienced teacher may use AI extensively while objecting to assessment automation. A newer colleague may need support but be reluctant to reveal it publicly.
The research discussed earlier concerns particular populations. It cannot justify assigning a motive to someone across the staffroom.
Look at the work, the explanation and the circumstances.
Close the conversation with an answer people can use
A concern can disappear from the agenda without being resolved.
Record the question, who will respond and what staff should do while the answer is pending. Where the school changes the proposal, explain which concern the revision addresses.
In the fictional department, a useful response might clarify that generated comments remain drafts, identify the review responsibility and arrange an examination of checking effort before wider use. If that examination does not support the proposal, say so.
Where a concern remains unresolved, make the uncertainty visible. “We have discussed it” is not an operational answer.
Later, revisit whether the response addressed the original difficulty. Continued questioning may reveal an incomplete answer rather than an unchanging attitude. It may also show that the disagreement persists after the facts have been clarified.
Either way, the next discussion has a more specific starting point.
Use a school event to practise interpretation
A teacher-development event can give leaders and teachers a shared fictional case.
Present the same objection in several forms. One colleague misunderstands the process. Another identifies a genuine checking burden. A third questions its educational purpose. Ask participants what they would investigate and how their response would differ.
Keep the activity separate from public discussion of identifiable staff. Its purpose is to improve the conversation, not sort colleagues into supporters and opponents.
A keynote can introduce the distinction between a label and an explanation. A facilitated session can examine examples. Neither establishes that the school has resolved its implementation difficulties.
For organisers, our guide to choosing an AI in education keynote speaker connects a session's contribution with the educational question the audience needs to pursue.
Understand the objection before choosing the response
Sources were checked on 28 October 2026. This article draws on two contextual studies, not a systematic review or evidence that one conversation method changes adoption. The fictional feedback example and suggested questions are original planning proposals.
The research does not establish that all hesitation is justified, that every objection concerns implementation or that more AI use is necessarily the desired result.
Learning Coach's position is that professional learning begins with a question specific enough to investigate.
Before asking how to overcome a teacher's resistance, can you explain what the teacher is objecting to—and what an adequate answer would require?
Bring that question to your next school leadership or teacher-development event. Explore Learning Coach's international keynotes and request a tailored proposal.
References
Marzano, D., and Pautasso, S. Teacher technophobia toward Generative AI: evidence from an Italian survey. Frontiers in Education, 14 May 2026. Preservice teacher self-reports; sampling and non-response limits discussed above.
Habimana, O., Iyamuremye, A., Fadillah, M. A., and Isabwe, G. M. N. Teachers' acceptance of AI-powered applications in teaching and learning mathematics and sciences. Discover Education, 2 May 2026. Rwanda secondary STEM survey and interviews; perceptions rather than causal learning evidence.
