The bar every word passes
Nothing lands in this dictionary just because a machine coined it. Every candidate — human-submitted or AI-coined — faces the same three gates, in order, before publication. This page is the public statement of that bar.
Gate 1 — the craft checklist
All ten must hold before any search even runs:
- Names a real phenomenon — an observable, describable social pattern. Not a joke, not a private in-joke.
- Pronounceable — a reader who sees it can say it.
- Short — preferably four syllables or fewer.
- Real word-formation — built from productive English patterns: compounds, clean blends, live affixes (-washing, -flation, -creep, micro-). The morpheme boundaries must be real: the word parses.
- The fusion must do work — see below.
- Never a brand or trademark.
- Etymology stated — the roots and why they fit, on every entry.
- Machine-coined entries carry more, not less — a proposed IPA pronunciation, related-entry placement (empty by judgment, never by omission), and pre-selected categories.
Rule 5 is the one that kills the most candidates. If two words already form a natural, everyday phrase — and the coinage means exactly what that phrase means — we don't publish it. Fusing mile marker into milemarker adds nothing: that is respelling, not naming. A fusion earns its place only when the phrase doesn't already exist, the meaning is more than the sum of its parts, or the form itself is transformed — clipped, blended across boundaries, shifted in grammar.
Gate 2 — the novelty search
Three layers, because exact-string search alone misses most collisions:
- The exact form — quoted searches for the word alone, plus "meaning" and "definition" phrasings, plus dictionary checks.
- The source language — a word built on foreign roots is searched in its own language and its standard English translation, not just the fused English spelling.
- The split phrase — every compound is searched unfused, because "run club" doesn't stop existing because someone spelled it runclub.
No search engine proves a negative, so our claim is always dated and scoped: no usage with this meaning was found in the queried indexes, on the checked date — and the evidence is published with the entry. A result only counts against a candidate when it means what the candidate means; finding the form in an unrelated sense does not fail it.
Gate 3 — distinctness and the editorial line
- An established word, a named trend, or a living brand already owning the concept or its neighborhood blocks the coinage — even when the exact string is free.
- Words here target systems, institutions and practices — never an ethnic, religious, or other group of people. No conspiracy narratives. An entry that crosses this line is rejected regardless of novelty.
What happens to a verdict
- Approved — published with its verification badge (the badge meanings are on the about page).
- Revise — the phenomenon is real and unnamed but the form collides or does no work: the form is reworked, not the idea.
- Rejected — never inserted, with the evidence that drove the decision kept on record.
Every non-approval carries the evidence that drove it, with the search engine and date named. Raw result counts are never the verdict — the question is always whether anyone already uses this meaning, under any spelling.
See it in practice
The pipeline, the badges and the participation ladder are on the about page. Think you've spotted a phenomenon with no name? Coin a word →