our free wordlist
download for crossfire spreadthewordlist.dict
download for crossword compiler spreadthewordlist.txt
last updated October 1, 2024 (join our mailing list for quarterly update emails)
size 302,179 answers, with 110,913 scored 50 or higher
license Creative Commons BY-NC-SA (the “NC” means you cannot adapt our list for commercial use)
download answers 50+ only (for experts) spreadthewordlist_unscored_high.txt
DOWNLOAD THESE FILES WITH UPPERCASE LETTERS HERE
download a scrabblized version for pangrams here!
scoring levels
50: likely good
40: questionable
30: stale or variants (use with caution)
20: thematic (use with caution)
10: likely a misspelling/parsing error (use with extreme caution)
0: offensive
we recommend setting a minimum score of 50. because of the data-driven nature of this project, it is not the case that everything scored at 50 or higher will meet an editor’s standards — please use your judgment!
tldr readme
our wordlist is compiled exclusively from freely available datasets. please feel free to share it.
the scoring algorithm involves three types of sources: freely available crossword-specific datasets, other freely available datasets, and a manually curated blocklist (see below). our wordlist is not attempting to mirror or incorporate crucial efforts towards the inclusivity of crossword puzzles by such initiatives as the expanded crossword name database, and we recommend supplementing this wordlist with those external resources. check out our resources and faq pages for more info.
crossword-specific datasets: answers that have appeared since 2020 in the USA Today, Universal, and Modern crosswords are our most-trusted sources, which means an answer has to appear relatively few times for us to score it as a 50. higher scores are also more likely to be given to answers that have appeared in other publications since 2020 with varying policies for long (7+ letters) and short (6- letters) answers.
other datasets: to ensure our wordlist is not solely composed of past crossword answers, freely available datasets sourced from Wikipedia, news search terms, Jeopardy terms, and Scrabble words have been incorporated. various resources have been used to prioritize US English spellings.
blocklist: we have manually developed a list of offensive content; this is supplemented with similar externally developed lists. the last step of our wordlist creation algorithm is to scan it for any answer that is in our blocklist; those answers are set to zero if they appeared in a crossword-specific dataset and are removed otherwise. see also our faq.