What live classroom assessment can show when AI can write homework

A [Reuters report carried by GMA on October 29](https://www.gmanetwork.com/news/topstories/world/1007548/how‑teachers‑are‑remaking‑classrooms‑for‑the‑ai‑age/story/) describes individual US teachers changing how they assess students as AI can produce take‑home answers. San Francisco teacher Tristen Chang began replacing take‑home essays with oral exams last year. New York math teacher Bobson Wong asks students which problem‑solving method they prefer and why. The reporting describes classroom responses, rather than a new national policy. The useful assessment question is what the teacher needs to observe. A polished answer is an artifact: it can be read for accuracy, structure or interpretation. A live explanation adds an opportunity to ask how a student reached that answer. These are different evidence‑gathering tasks, not competing guarantees of authenticity. A follow‑up question can make the distinction concrete. Asking a student to explain a choice or respond to a changed assumption gives the teacher something to assess beyond the initial result. Whether that actually reveals the intended understanding depends on the question and the conditions of the assessment; the format alone does not establish it.

A teacher considering a change can start by naming the evidence the task should produce. If the goal is explaining an interpretation, what part of the explanation needs to be heard? If the goal is choosing a mathematical method, what makes the choice appropriate, and what would change it? If the goal is revising a piece of writing, which decision should the student be able to discuss? That approach keeps an assessment focused on a defined learning goal. It also makes the limits clearer. In‑class work offers an observation window; it is not a complete record of everything a student knows. An oral response may expose a line of reasoning, but it is not a universal test for AI use. A method‑choice question can ask for a justification without proving that the original answer was produced unaided. For each format, the practical design questions include what assistance is allowed, how the evidence will be judged and how students can demonstrate the same goal through an appropriate accommodation. A change of format should not quietly change the skill being assessed.

The [Reuters reporting](https://www.gmanetwork.com/news/topstories/world/1007548/how‑teachers‑are‑remaking‑classrooms‑for‑the‑ai‑age/story/) also preserves important qualifications. In teacher Meghan Hyland’s classes, handwriting has an exception for students requiring accommodations. Some teachers interviewed also use AI for preparation, materials or adaptation. Their concerns about student misuse do not amount to a uniform rejection of classroom AI. The comparison therefore should not collapse into paper versus screens or people versus software. A useful assessment brief can state the learning goal, the permitted help, the evidence a student will show and the way that evidence will be evaluated. That makes the expectations discussable before a finished answer becomes the only record. For a separate look at AI supporting instruction, [our report on Stanford’s classroom projects](https://opentools.ai/news/stanford‑classroom‑ai‑teacher‑roles‑questions‑prerequisites‑tutoring) follows question grouping, prerequisite mapping and tutors that prompt learners instead of supplying an answer. Those higher‑education examples help distinguish a teaching‑support task from an assessment task; they do not establish outcomes for the US school classrooms described by Reuters. OpenTools conducted no interviews or classroom tests for this analysis. The reported teacher examples are not a representative survey or evidence of learning gains. They motivate questions about assessment design; they do not establish an AI‑proof format or a policy every classroom should adopt. AI‑generated conceptual illustration of a teacher inviting a learner to explain a choice. The people, shapes and conversation are symbolic—not an actual school, student assessment or evidence that a task prevents AI use. Original artwork by OpenTools Team, informed by [Reuters reporting carried by GMA](https://www.gmanetwork.com/news/topstories/world/1007548/how‑teachers‑are‑remaking‑classrooms‑for‑the‑ai‑age/story/).