The dictionary, explained

Where the words come from

The dictionary is built from SCOWL (Spell Checker Oriented Word Lists), an openly licensed word-list project widely used by spell checkers and word-game tools. Rather than compiling a private word list from scratch, this site reuses SCOWL's existing, already-vetted data through the wordlist-english package (MIT licensed), which organizes SCOWL's words by dialect and by how common each word is. Two of those sets are combined here: the spelling-neutral English core (words spelled the same in every major English dialect) and the British-specific set (words spelled differently in British/Commonwealth English than in American English, such as "colour" and "centre").

How common a word has to be to make the list

SCOWL organizes words into frequency bands, from the most commonly used English words down to the most obscure. This dictionary includes every band up to, but not including, SCOWL's most obscure tier, which is reserved for highly specialized, archaic, or extremely rare words unlikely to ever come up in a real Scrabble game. That cutoff is a deliberate trade-off: a stricter list would miss some genuinely valid words that Scrabble players actually use, while a list that includes the most obscure tier would pad results with words so rare they'd rarely help and would occasionally crowd out a much more plausible answer. The 76,266-word cutoff aims for the middle: comprehensive enough for real games, without the very longest tail of specialist vocabulary.

What's in the file, and what's deliberately left out

Every entry passed two checks before being included: lowercase letters A through Z only (no numbers, apostrophes, hyphens, or spaces), and a length between 3 and 15 letters. That means possessives, hyphenated compounds, and multi-word phrases are excluded, along with 1 and 2-letter words. The 2-letter exclusion is a deliberate simplification, not an oversight: official Scrabble dictionaries disagree with each other about which 2-letter words are valid (North America's NWL/OSPD and the international Collins Scrabble Words don't fully overlap on that narrow category), so a general-purpose dictionary that isn't tied to one specific official rule set leaves it out entirely rather than silently pick a side. Duplicate entries across the combined source lists were removed, and the final file is sorted alphabetically, one word per line.

A full, combined copy of all 76,266 words is served as a plain, publicly downloadable file at /data/word-list.txt. The word finder tool fetches that same file directly (once per visit, cached afterward) rather than maintaining a separate copy, so the list you can download is always exactly the list the tool searches: there's no second, hidden dictionary anywhere in the site.

Licensing and reuse

SCOWL's word lists and the wordlist-english package that organizes them are both openly licensed for reuse (SCOWL under its own permissive terms, wordlist-english under the MIT license). This site reuses that existing, already-built data rather than compiling a private word list from scratch, in line with the general principle that a proven, openly licensed data source beats reinventing one. No proprietary or scraped word list is used anywhere on this site. The same straightforward approach applies to your own data: see the privacy page for exactly what this site does and doesn't collect when you use the word finder.

How this compares to an official Scrabble dictionary

It's worth being upfront about the biggest limitation: this is a general English dictionary, not an official tournament word list like the OSPD/NWL (used in North America) or Collins Scrabble Words/CSW (used internationally). Official Scrabble dictionaries are curated by specific rules committees and include some short, unusual words (and exclude some perfectly ordinary ones) for reasons specific to competitive play rather than general usage. The word finder's results are a strong starting point for checking whether a word is plausible and for comparing scoring options, but if a specific play needs to be tournament-legal, cross-check it against whichever official list the game you're playing actually uses. See the examples page for how the score sorting is meant to be used in practice: as a way to compare and rank options, not as a claim that every result is guaranteed to be an accepted tournament word.

Why one combined file instead of splitting by length

A rack search has to check every word length up to the size of the rack at once (a 7-letter rack can produce valid 3, 4, 5, 6, and 7-letter words all in the same result set), so splitting the dictionary into one file per length the way a fixed-length tool would wouldn't actually save anything here, since nearly every search still has to touch most of the length buckets anyway. Serving one combined 722KB file, fetched once in the background after the page has already loaded and cached in memory for the rest of the visit, keeps the implementation simple without costing anything on first paint: the fetch is deferred until after load, so it's never part of what the browser has to finish before the page looks and feels ready.