How Atrivial generates each quiz
Drafted questions, one right answer, three believable wrong ones and your final check.
When you click Generate questions, Atrivial runs your content through a short series of steps before anything shows up in Questions. This is what happens at each one, including where the checking happens and who does it.
You'll put these questions in front of your readers under your own name, so you should know exactly how they were made.
Atrivial cleans the source first
Atrivial reads your article, feed post, pasted text or file and removes what isn't content, like navigation, ads, promotions, cookie notices and related-article lists. Anything it isn't confident is clutter stays in. It doesn't summarize or rewrite your piece at this stage.
Long sources get trimmed to their opening portion at about 60,000 characters. If the part you care about comes late in a long document, paste that section on its own.
Questions are drafted to stand on their own
Each question has to make sense to someone who hasn't read your piece. So there's no "according to the article," and nothing tests whether a reader remembers your exact wording.
The drafting step looks for what your content develops, such as causes and effects, how something works, common misconceptions, contrasts and concrete scenarios. The aim is a question that sparks curiosity and makes your piece the natural next read.
When an article is thin, Atrivial may add well-established general knowledge closely tied to its main idea. Its instructions rule out adding anything time-sensitive, niche, disputed, number-heavy, medical, legal or financial that your content doesn't support. They also say not to present advice or opinion from your piece as universal fact.
Questions are drafted to fit under 120 characters, with answers under 30, so they're easy to scan.
One right answer and three believable wrong ones
Every trivia question gets exactly one right answer and three wrong ones, and the drafting rules are strict. The right answer has to be fully correct under the exact wording of the question and the only defensible choice. Each wrong answer has to be wrong under that same wording, with no partial credit.
The wrong answers also have to be believable. They should match the right answer's grammar and format, and none should give the game away by being absurd or oddly worded. If several answers could be true in everyday life, the instructions tell the model to narrow the question, try a different angle or generate a different question altogether.
Atrivial discards any draft that's missing its question, right answer or any of its three wrong answers, then runs extra rounds to make up the count. Each question also gets a short fun fact, and a trivia question isn't complete without one.
Where the fact check happens
For quizzes generated from your content, there's no separate fact-checking pass after drafting. That holds on every plan and at every AI model quality setting. The answer rules above live inside the drafting step. They tell the model what a fair question looks like, and they can't guarantee it got there.
So the last check is yours, and the Portal gives you room for it. Your first quiz opens in Questions as soon as it's generated, and you can edit any question, answer or fun fact there. Readers only reach the quiz through the email code, link or embed you put out, so you can read it through before your send.
For recurring quizzes, Review before publishing is on by default. Each new quiz waits in Upcoming Draft until you click Approve, and if you don't approve it in time, the current quiz stays live.
A quick way to check is to read each right answer against your piece, then ask whether a sharp reader could argue for one of the wrong ones.
Settings you control
Advanced settings, the last screen of Input, holds three options. AI model quality offers Fast, Balanced and High Quality, and Balanced is the default. Fast uses less of your plan's generation allowance and High Quality uses more. In Writing style, you might add "Use a warm, conversational voice." Question instructions is for rules such as "Avoid partisan political questions."
High Quality is a Business plan option. It runs on a slower, higher-quality model and generates fun facts in a separate step. Individual includes Fast and Balanced. Both plans get the same answer rules and the same editing tools.
If a set misses the mark, Regenerate Questions in Input generates a new set from the same source.
See which plan gives you the AI model quality and generation allowance you need.
Compare plans
