The Best AI for Teachers in 2026
Our teachers guide is live, and this thread is for the most contested call we made in it. The guide takes a side in the biggest argument in education right now: when it comes to student AI use, redesign your assessments instead of trying to detect your way out. We know plenty of educators disagree, so here is our reasoning, and the floor is open.
Full guide with the complete stack (most of it free), the planning, differentiation, and assessment workflows, and the classroom ethics section is here: <https://whataidoineed.com/best/ai/for/teachers>
**Why we came down on the redesign side:**
The detection numbers do not support the weight being put on them. Turnitin's AI detection (the most established option) produces both false positives and false negatives, and a false positive is not a minor error: it is a formal accusation against a student who did the work, sometimes a student whose writing style (ESL students are over-represented here) simply pattern-matches to AI. Several universities have pulled detection from their disciplinary processes for exactly this reason. A tool that cannot be relied on for the accusation cannot carry the policy.
Meanwhile the detection arms race only runs one direction. Students paraphrase, humanise, and iterate faster than detectors update, and every detector improvement teaches the workaround. Betting your academic integrity policy on winning that race is betting against the trend line.
**What redesign looks like in practice (not theory):**
In-class writing for the work that must be verifiably the student's own. Oral defenses: five minutes of "walk me through your argument" reveals more about authorship than any percentage score. Process-based assessment, where the outline, draft, and revision are each visible (Brisk's Inspect Writing makes the document history readable, which is observation rather than accusation). And for at-home work, the honest adjustment: assume AI is in the room and design tasks where using it well is the skill being assessed, because that IS the skill their adult lives will assess.
**The position we want pushback on:**
We are not claiming detection is useless. As one signal among many, opening a conversation rather than closing a case, it has a place. We are claiming it cannot be the foundation, and that the hours spent on the detection arms race buy more learning when spent on assessment redesign instead.
**For the thread:**
Teachers who redesigned: what actually worked, what flopped, and what did it cost you in prep time? The in-class-writing pivot has real trade-offs (class time, slower writers, accessibility) and honest accounts of those trade-offs are worth more than the success stories.
Teachers who still run detection: make the case. Especially if your context (large lectures, fully remote, institutional mandate) makes redesign genuinely impractical, because the guide's advice assumes options not everyone has.
And the question underneath all of it: what does academic integrity even mean for students who will spend their careers working with AI? Genuinely curious where this community lands.