The published standard
How we vet every question
A question cannot reach the site unless the build can prove that it passes the safety, age-band, grammar, balance and duplicate checks below. The build refuses to publish the corpus when any rule fails.
- Current build
- 753 questions
- Age ranges
- 3 separate bands
- Coverage floor
- 25 per available band and category
The release gate
What the machine checks
Questions are authored as two option fragments that complete the words “Would you rather.” One dependency-free checker judges authoring, builds and tests, so the standard does not change between those stages.
-
Shape and read-aloud grammar
Each option must be a clean lowercase verb phrase with at least two words. The checker rejects digits, sentence punctuation, doubled spaces, an extra “or,” unknown starting verbs and characters outside the allowed form.
-
Age-band length and word difficulty
Every band has its own caps for words, characters, combined pair length and a deterministic per-word syllable estimate. The combined cap also keeps long questions inside the space reserved by the interface.
-
Safety and category boundaries
Exact words, prefixes and phrases are checked against global, brand, youngest-band and gross-category exclusion lists. A term fenced into Gross cannot appear in Food, Silly or any other category.
-
A choice that is actually balanced
The two options cannot be identical, heavily mismatched in length or framed so that one is plainly better. These checks reduce wording that steers the answer before a child has made a choice.
-
Corpus-wide coverage and duplication
The checker rejects exact, reversed and high-overlap near duplicates. It also requires at least 25 questions in every age-and-category cell the interface can offer.
-
A checker that proves it can fail
A separate fault-injection gate exercises all 36 declared question-level and corpus-level rules and requires the expected rule to fire. A clean control must remain clean. Adding a rule without a matching fault case fails that gate.
Three different jobs
Why the age bands are separate
The bands are editorial product boundaries, not a claim that every child develops at the same pace. They control vocabulary, abstraction, topic access and how much text the room has to process at once.
| Ages | Writing target | Mechanical cap per option | Category boundary |
|---|---|---|---|
| 4–6 | Concrete, present-tense, sensory and cheerful | 8 words, 46 characters, 3 estimated syllables per word | No Gross or Big Questions; extra fear, permanence and loss exclusions |
| 7–9 | More imaginative situations with direct choices | 11 words, 64 characters, 4 estimated syllables per word | Gross is available as a separate, removable category |
| 10–12 | Longer scenarios and genuine trade-offs | 14 words, 82 characters, 5 estimated syllables per word | Big Questions and the strongest Gross vocabulary begin here |
The exclusion standard
What does not belong in the question bank
The matcher uses reviewable words, prefixes and phrases rather than a vague “clean” label. The source lists are the enforcement authority; these groups explain their scope in plain language.
Excluded in every band
- Death, self-harm, weapons, violence, injury, abuse and coercion
- Alcohol, smoking, vaping, tobacco and drugs
- Dating, romance, sexual content and nudity
- Appearance or body shaming
- Religion, politics, protected-trait and identity prompts
- Family trauma, poverty and housing insecurity
- Disability insults, slurs and demeaning labels
- Severe fear, harmful occult themes, illness and mental-health crises
- Crime, prisons and kidnapping
- Brands, franchises, named characters and real public figures
Additional rules for ages 4–6
- No frightening creatures, haunted settings or nightmare language
- No disasters, dangerous isolation, trapping or getting lost
- No “forever,” immortality, infinity or permanent-loss framing
- No time travel, alternate dimensions or body swapping
- No Gross category
Gross is fenced, not mixed in
Bodily-humour vocabulary begins only in the Gross category for ages 7–9 and 10–12. A second, stronger list is allowed only in Gross for ages 10–12. Turning Gross off removes those matched terms from the active pool by construction.
Release identity
How a corpus change is recorded
The current question count above is inserted from generated data during the build. The same generator calculates a stable SHA-256 release-data fingerprint over the questions, prompts, age/category rules and lint policy, then writes it to the generated module and release statistics. This page does not hand-copy a fingerprint that could drift from the shipped data.
A small, attributable subset is also available as a machine-readable question sample.
Limits and corrections
What this process cannot promise
- Word and phrase matching cannot understand every possible context. A list can miss a new phrasing or reject a benign use, which is why its exceptions are explicit and the fault gate tests the matcher itself.
- The syllable count is a deterministic heuristic, not a clinical reading assessment. The age bands are not medical, psychological or curriculum advice.
- This standard does not claim that a named human reviewer, teacher or child-development professional individually approved every question.
- A mechanically valid question may still be a poor fit for a particular child, family or classroom. Adults should choose the age and categories that fit the room in front of them.
To report a question that should be corrected or removed, use the operator contact route published on the Privacy page. There is no contact form and this page does not collect a message.