Licenses & attributions

Fonts

Icons

Inspiration & references

• 3Blue1Brown: Solving Wordle using information theory, a visual explanation of the approach
• Claude Shannon: A Mathematical Theory of Communication, the source of the entropy measure used here

Word lists

The Wordle guess list comes from an archived public copy of the game's list. The possible-answer list begins with the original public answer array and adds answers observed through the New York Times daily-answer API. Practice and retained lookup pages also draw on SCOWL and other public word-list sources. For the Wordle lists, the upstream source links and per-file checksums are on the data page, and provenance notes ship inside the current research release.

The independent Spelling Bee-style tools use a separate 62,717-word lexicon generated from the hash-pinned Ubuntu wamerican 2020.12.07-4build1 package (SCOWL 2020.12.07), then filtered to lowercase ASCII entries of 4–15 letters and screened with this site's exclusions. The package's source, copyright and licence notices are preserved in the project; any reuse or redistribution decision requires the owner to review those exact notices. This corpus does not represent The New York Times Spelling Bee dictionary or acceptance policy, so its counts and matches can differ from a published puzzle.

Data sources

• Word etymologies: sourced from Wiktionary, available under the Creative Commons Attribution-ShareAlike (CC BY-SA) license. Every displayed etymology links back to its Wiktionary source.
• Definitions and lexical details: selected and lightly formatted from an offline English Wiktionary extract produced by Kaikki. Wiktionary text is available under CC BY-SA 4.0. The import selects up to four senses, collapses whitespace and may shorten long entries; every displayed entry links to its source and identifies the adaptation. Unsourced glosses are omitted, and definitions are never treated as evidence that a word is accepted or can be an answer.
• Opt-in dictionary API lookups: the separate definition API may request a missing word server-side from dictionaryapi.dev. Word-page HTML never waits on that service.
• Exact optimal-play evaluations and daily optimal routes: derived from Alex Selby's Wordle solver results and decision tree (MIT License).
• Pronunciations: derived from the CMU Pronouncing Dictionary (BSD-2-Clause). Words absent from CMUdict carry machine-generated estimates from the g2p_en model (Apache-2.0) and are labeled as estimates wherever shown.
• Synonyms, antonyms and categories: derived from Open English WordNet, 2025 edition, under CC BY 4.0.
• Example sentences: from the Tatoeba Project, licensed CC BY 2.0 FR; each displayed sentence is credited to its contributor's Tatoeba username and linked to its sentence page.
• How Sure? comparison facts: quantities, labels and as-of dates from Wikidata, released under CC0 1.0; each round names its source and the date its figures were true.
• Learner-difficulty levels: derived from the CEFR-J Vocabulary Profile ver 1.5, redistributed by Open Language Profiles; free for research and commercial use with citation; copyright Tono Laboratory at TUFS. Cited as: "The CEFR-J Wordlist Version 1.5. Compiled by Yukio Tono, Tokyo University of Foreign Studies. Retrieved from http://www.cefr-j.org/download.html on 1/20/2020."
• Psycholinguistic norms: imageability, familiarity and age-of-acquisition values from the Glasgow Norms (Scott et al., 2019), and perceptual-modality values from the Lancaster Sensorimotor Norms (Lynott et al., 2020), both under CC BY 4.0; values are rescaled and adapted as described on the pages that show them.
• Word-frequency scores: derived from wordfreq by Robyn Speer (code MIT; word-frequency data CC BY-SA 4.0), used for the site's rarity flags and frequency bands.

Full license texts

Follow the project links above for license texts and notices. Python packages may also install transitive dependencies; their metadata and license files are included with the installed packages.