Most teachers in 2026 are dealing with the same pattern. Student work is more polished than it has ever been. The thinking behind the work is harder to read. AI tools are everywhere — at home, on phones, in the browser tab next to the assignment — and the cognitive habits the curriculum was built to develop are getting fewer reps than they used to.
This post is for teachers and lecturers who want to keep deliberate practice in the room without pretending AI doesn't exist. The published evidence is on the cautious-pragmatist side, not on either of the extremes that dominate education conferences.
Why this matters — the published evidence
The 2025 EDUCAUSE Review piece on the "productivity paradox of AI on students' thinking" names the pattern directly. Students who lean on AI produce better artefacts and report weaker engagement with the originating cognitive act. The artefact-and-act gap is the load-bearing observation.
The Conversation's 2023 piece on ChatGPT and student writing motivation reported the same direction qualitatively early in the AI-in-classroom timeline. AI-assisted students often produced more polished work and developed weaker motivation to write or think for themselves. The qualitative read foreshadowed what later surveys found at scale.
The 2024 Springer overview of AI overreliance in education (Springer Nature) covers the workplace-and-classroom transfer. The mechanism is the same as in the general cognitive-offloading literature: cognitive acts that get fewer reps tend to weaken, and AI in the classroom shifts which acts get reps.
The constructive frame is the deliberate-practice literature (Ericsson, Krampe & Tesch-Römer, 1993). Specific cognitive skills respond to daily, effortful, on-purpose engagement with the skill. Teaching is the structured form of that engagement. The classroom is one of the few places where deliberate practice on cognitive skills happens at scale.
What deliberate practice looks like in front of students
Three classroom moves carry most of the load. They are not new. They are harder to hold to in an AI classroom because the alternative is so frictionless.
Modelling the originating act. When you write an example paragraph on the board, when you solve a maths problem in front of the class, when you reason aloud through a primary source — the students see the cognitive act they are being asked to develop. AI doesn't do this. The artefact appears with no visible originating process. Modelling is the part of teaching AI cannot substitute for.
In-class unmediated practice. Phones in a tray. Laptops closed. Twenty minutes of writing, calculation, reading, or reasoning that has to happen in the room, with the class as witnesses. The volume does not need to be large; the unmediated existence of the practice does. The deliberate-practice literature is consistent on this — the act has to happen, on-purpose, with feedback.
Feedback on the act, not just the artefact. Marking AI-assisted work tells you about the artefact and very little about the student. In-class practice plus immediate, specific feedback on the cognitive act keeps the loop on the part of the work the AI cannot fake.
The assessment redesign question
The harder structural change is assessment. The traditional take-home essay or problem set was a defensible proxy for the cognitive act when the only way to produce the artefact was to do the act. That assumption is no longer safe.
The honest options for assessment redesign cluster into four families:
In-class assessment of the cognitive act. Timed, unmediated, in the room. The simplest defensible measure of what the student can actually do without AI.
Process artefacts alongside the final artefact. Drafts, voice notes, in-progress screenshots, oral defences. The process becomes the evidence. AI-assisted final artefacts are still allowed; the process artefacts are what get assessed.
Oral exams and viva-style defences. The student has to talk through what they wrote or solved. The artefact may have AI in it; the defence has to come from the student.
AI-allowed but transparent. The student declares what AI did, the assessment criteria treat the AI contribution explicitly, and the marking weight on the originating act is separated from the marking weight on the final polish.
None of these are perfect. All of them are more defensible in 2026 than an undeclared take-home essay graded on artefact quality alone.
What this is not
A few honest disclaimers, because the AI-in-classroom category is full of overclaim and panic in roughly equal measure.
This is not a case for banning AI in schools. The published evidence does not support an abstinence position, the tools are not going away, and students will use them outside school whether they are allowed inside it. The recommendation is calibration — keep deliberate practice on the timetable, redesign assessment, and treat AI as a tool that gets used transparently.
It is not a claim that AI use causes lasting cognitive harm to students. None of the cited evidence supports that claim. The narrower claim — that the originating cognitive acts of writing, calculation, reading, and reasoning get fewer reps when AI is in the loop, and acts that get fewer reps tend to weaken — is what the evidence supports.
It is not a substitute for school-level policy. Phone policies, AI-use policies, and assessment standards are decisions for departments and schools, not individual teachers acting alone. The moves above are what an individual teacher can do regardless of how the policy questions get answered.
How Senwitt fits
Senwitt is not a classroom curriculum and we do not pretend it is. What it offers is a daily seven-minute Set across six Senwitt Skills — Writing, Math, Code, Memory, Reading, and Reasoning — that students or teachers can do independently to keep the underlying skills warm. The deliberate-practice frame comes directly from the Ericsson 1993 paper.
For teachers, the for teachers page and the for students page lay out the case in more detail. The research/your-brain-on-chatgpt page covers what the cognitive-debt research actually supports.
A note on the transition cost that often goes underspoken in education conferences. Redesigning assessment is not free. In-class assessment requires room and time. Process artefacts require students to keep work-in-progress they were not keeping before. Oral defences require staffing the conversations. Teachers in 2026 are being asked to absorb this transition cost on top of the workload they already carry, and any honest piece on the AI-classroom problem has to acknowledge that the right pedagogical move is sometimes blocked by structural realities — class sizes, contact hours, marking time, departmental policy. The recommendation in this post is therefore deliberately limited to what an individual teacher can do without departmental sign-off. Modelling the originating act, protecting one in-class unmediated practice block per topic, and giving feedback on the act as well as the artefact are within reach. The full assessment-redesign question is real and worth pushing on, but the daily classroom-survival version of deliberate practice does not depend on winning that argument first.
