WAYGROUND · RUNG IV · INNOVATE

Students audit what the AI wrote

Wayground will write a quiz from a text in about thirty seconds, free, and its own documentation warns that it may generate inaccuracies. Most people treat that warning as a caveat. It is the assignment.

Idea starters, not finished practice. None of this is an expectation. On whether a given AI question is actually wrong in your subject, you are right and I am not — and so is the student who argues with you about it.

The feature, and the sentence underneath it

Wayground AI generates an assessment from a document, a prompt, a website, a YouTube video, a curriculum standard, or a photo of a worksheet. It is on the free plan. Feed it a chapter and you have twelve questions before the kettle boils.

Underneath it, in their own documentation, is this: Wayground AI is continuously improving, but may occasionally generate inaccuracies. Please review the content before publishing.

That sentence is doing a lot of work. It is a disclaimer, and it is also an accurate description of a machine that produces confident, fluent, plausible assessment items about a text it has processed but does not understand. Which makes it the most useful object in the room, if you point students at it instead of hiding it from them.

Wayground AI — what it generates from →

The move, in three parts

1 · Generate, and do not fix it

Point the AI at the text the class just read. Take the twelve questions it writes. Do not clean them up.

Read them yourself first, once — you need to know what is wrong before they tell you, and you need to be sure nothing in there is harmful rather than merely inaccurate.

2 · Give students the audit brief

Three of these questions are bad. Find them. For each one, say what is wrong with it and write a better version.

Do not say which three, and be honest with yourself that there might be two or five. The number is a nudge, not a fact, and if a student argues that a fourth one is broken and they are right, that is the best outcome of the day.

3 · Make them defend it out loud

Ninety seconds per pair, said not written. The reasoning is the evidence, and spoken reasoning is both faster to grade and much harder to outsource.

Every student speaking — oral evidence you can grade →

Why this clears the bar

The Innovate standard in this course is strict: not faster, not easier — previously impossible. Most AI classroom moves fail it. This one does not, and the reason is worth stating precisely.

A student critiquing a bad question is not new. A student critiquing a bad question written about the text they read this morning, produced in thirty seconds, at no cost, in a quantity that lets everyone have a different set — that is new. Before, a flawed test item was either a real one you were not going to hand out for dissection, or one you wrote badly on purpose, which students can smell.

And there is a second thing, which is the part I think matters more. The class is being asked to evaluate a machine that sounds authoritative and is sometimes wrong. That is not a hypothetical skill for them — it is the daily condition of their lives, and here they get to practice it on material where they are the ones with the expertise and the machine is the one guessing.

That inversion is rare and it is worth a block. Change the evidence, not the topic →

The strongest version of the task is comparative. Generate two sets from the same text with different prompts and ask which set is better and why. The answer is never obvious, the argument is about what a good question is, and that is a conversation most students have never been invited into.

What it looks like in a room

An 80. Students read the chapter for homework. You arrive with twelve AI-written questions on the screen, unedited.

Within four minutes somebody has found the easy one — a question whose answer is stated verbatim in the first paragraph, testing nothing. Fine. That is the warm-up.

Around minute twelve a student finds the interesting one: a question that is technically answerable from the text but rests on a detail the chapter mentions once, in passing, in a subordinate clause. It is not wrong. It is trivial dressed as difficult.

And now the class is arguing about what a question is for, which is a conversation about the material disguised as a conversation about the quiz. They cannot have that argument without knowing the chapter well, and they are having it voluntarily.

The AI wrote a bad question. The bad question was the lesson. You did not write anything.

Where it breaks

You do not read the output first. Non-negotiable. Generated content about a historical atrocity, a text with contested framing, or anything sensitive can be wrong in ways that are not merely academic. Read all twelve before the class does. This costs three minutes and it is the whole risk control.

Students use AI to audit the AI. They will try. The oral defense is the answer — not as a trap, but because the reasoning is what you were assessing and speaking it is the fastest honest route to it. Say that out loud rather than policing it.

The questions are all fine. Sometimes they are, especially on straightforward factual material. Have a plan B, and if it happens, that is a finding worth saying: the machine did well on this text. Why do you think this text was easy for it? That question is at least as good as the original assignment.

It becomes an AI lesson. The subject is the chapter. If the block ends with students knowing more about how language models work than about the material, the task has drifted.

The ceiling and the gradebook. Twenty stored activities on the free plan, and no Canvas integration below a school plan. Grade the written revisions or the oral defense as a Canvas assignment — the artifact is theirs, not Wayground’s.

Next

Rung III · Deepen

Run the corrected version with redemption questions on, and measure whether the fixes helped.

The second attempt is the assessment →

Rung II · Enhance

Read aloud and extra time for the class, so the audit is a reading task everyone can actually do.

The accommodations are on for everyone →

Back to the hub

The name change, the free feature table, the ceiling, Canvas, and who to ask.

Wayground — twenty and twenty →

They spend their lives being told things confidently by machines. Here is thirty minutes where they are the expert and the machine is the one guessing.

Jason Beyer · Educational Technology · Regis Jesuit High School · A.M.D.G.

HelpDesk@regisjesuit.com · 303.269.8080 · LAST VERIFIED — AUGUST 2026