About this Would You Rather generator

Why it exists

"Would you rather" is one of the oldest zero-prep party games there is, but most online generators either need an account, load in ads before showing a single question, or wander into content nobody wants read aloud in a classroom or a work meeting. This page skips all of that: a bundled, hand-written question bank, a category filter, and a picker that avoids immediate repeats — nothing more.

How the question bank was built

Every question was written by hand, in pairs, and assigned to exactly one of eight categories: Classic, Funny, Food, Tech/Internet, Travel, Deep/Thought-provoking, Kids/Family, and Work-safe. The bank currently holds 128 questions, 16 per category.

Content policy. The entire bank is written to be strictly SFW / PG — no sexual, violent, drug-related, or otherwise adult content, in any category. That is a hard rule, not a soft guideline: the source data was run through an automated word screen for a list of clearly unsafe terms in addition to being written and read by hand, and every entry has to pass both before it ships. Kids/Family and Work-safe are the two categories written with the largest safety margin, aimed at classrooms and office ice-breakers respectively — but nothing in Classic, Funny, Food, Tech, Travel, or Deep is meant to need that same margin either.

How the picker works

The bank and the picker are a single, dependency-free JavaScript module (bank.js) with no DOM code in it at all, so it can be — and is — tested outside the browser under plain Node. Selecting Next draws a random question from whichever category is currently filtered (or the whole bank, for "All categories"), then checks it against the question that was just shown. If it's an exact repeat and the pool has more than one question, it re-rolls, up to 50 attempts, before falling back — a limit that exists only as a safety valve and is never reached in practice. A category with exactly one question is allowed to show that same question again, since there's nothing else to show.

Filtering by category rebuilds the pool immediately and draws a fresh question from it — it doesn't wait for the next click of Next.

How it was verified

Before shipping, the bank and picker were run against a set of automated checks under Node:

• Every one of the 128 entries has two non-empty options and a category that exists in the category list.
• No two questions share an id.
• An automated word screen flags zero entries against a list of clearly unsafe terms (sexual, violent, drug-related language), on top of the manual authoring pass.
• Filtering by any of the eight categories returns only questions from that category, and each category has at least ten questions.
• Filtering by "all", an empty value, or an unrecognised category id behaves correctly — "all" and empty return the whole bank, an unrecognised id returns nothing rather than silently falling back.
• Across 500 consecutive draws from a category with more than one question, the picker never returns the same question twice in a row.
• A category with exactly one question correctly returns that question even though it matches the "last shown" id — the no-repeat rule only applies when there's an alternative.
• Over 4,000 draws from a small category, every question in it gets drawn at least once — the randomness isn't silently biased toward a subset.

How the site is built

• Static HTML, one small stylesheet, two small scripts. No framework, no build step, no bundler.
• Nothing is fetched after the page loads. The full question bank ships inline with the page; picking a question, filtering by category, and copying text are all local operations.
• System font stack. Your operating system's UI font, so nothing has to download and nothing shifts as it loads.
• Your picks stay on your device. The optional tally is written to your browser's own localStorage, if it's available, and is never uploaded anywhere.
• Dark mode follows your system setting via prefers-color-scheme.

Scope and honesty about limits

This is a fixed, hand-curated question bank, not a language model generating new questions on demand — what you see is what was written ahead of time, nothing more and nothing less. The tally is a private, single-device counter for fun, not a survey of how other people answer. If a question ever reads as inappropriate for a setting it's supposed to be safe for, that's a bug in the content, not an edge case the tool is meant to handle at runtime — see contact to report it.