Introduction
Stuck On A Word aims to provide useful, transparent word-finding and puzzle tools. This page documents where the site’s word data is loaded from, how the datasets are processed in the published code, how the tools work at a high level, and how editorial updates and corrections are handled.
It is not a claim that any list is complete or that every result is accepted by every game.
Word data sources
Bundled dictionary chunks
The live tools load local JSON files listed in the dictionary-length manifest. The manifest currently identifies itself as version 1 and lists 172,820 entries across lengths two to 28. It is used by the general finder, pattern, ladder and other dictionary-backed tools.
Source attribution pending verification. The repository includes the processed files, but does not identify an upstream publisher, source date, or reuse licence. No licence or attribution claim is made here until that provenance is recorded and verified.
Purpose-specific dictionary chunks
The use-case manifest lists separate local chunks for wordle-5, anagram-general and long-words. The Wordle Solver loads the wordle-5 chunk; the published manifest currently lists 8,636 entries for it. These files are used to limit a tool to a relevant subset rather than as a statement of a game’s official answer list.
Source attribution pending verification. The upstream publisher, licence and release date for these processed files are not recorded in the repository.
Legacy CSV fallback
Some SEO listing pages can fall back to the local words.csv file when dictionary chunks are unavailable. Its upstream source and licence are also pending verification.
Dataset processing and filtering
The browser code trims entries, converts them to lower case and retains only entries matching a–z. The length-based loader selects chunks by the requested word length; the generic finder code can remove duplicate lengths before loading them.
Because the published filter accepts only letters A–Z, entries with hyphens, apostrophes, spaces, numerals or other unsupported characters do not pass that filter. The code does not publish a separate rule that identifies or removes proper nouns, abbreviations, offensive terms, archaic words or regional spellings; their presence depends on the bundled data. Repeated letters are retained as part of a word’s spelling and are evaluated by the relevant tool.
Tools can use different chunks: for example, the Wordle Solver uses the five-letter use-case list, while general search pages request entries by length. This is why the same spelling may not appear in every tool.
Tool methodology
Word finder, unscrambler, anagram and crossword-style tools filter loaded local words against the letters, positions, patterns or lengths supplied in the interface. Anagrams compare letter counts so repeated letters must be available in the input. Pattern and letter-bank tools apply their visible constraints to the selected list.
The Wordle Solver filters its five-letter list using green, yellow and grey tile constraints. When a letter is also marked green or yellow, a grey tile is treated as a position restriction rather than a blanket exclusion. Word-ladder tools search the available dictionary for one-letter changes between words. Pages that show letter frequencies or starting-word advice present editorial analysis rather than a universal puzzle rule.
Results can vary with the selected dictionary, the puzzle’s rules and the word list accepted by the game being played.
Testing and quality checks
The repository currently does not include an automated test suite, documented known-answer cases or a browser-test harness. The published code includes input validation and explicit handling for Wordle duplicate-letter positions, but this is not a substitute for comprehensive testing.
Changes should be checked in supported desktop and mobile browsers, including normal searches, no-result states, pattern constraints and repeated-letter behaviour. Reported errors can be reviewed against the relevant input, dataset and puzzle rule before a correction is published.
Editorial standards
Articles and guides should prefer primary or reputable sources for factual claims, distinguish facts from recommendations, and cite sources where appropriate. Original calculations or analysis should explain their inputs and assumptions. Pages should use clear publication and review dates, and should not make misleading or unsupported claims.
Where AI or automation is used to draft, analyse or format material, it is reviewed and edited by a human before publication.
Updates and review schedule
Datasets, tools and editorial pages may be reviewed when source data or puzzle rules change, an error is reported, a new check identifies an issue, or an article becomes outdated.
Published:
Last reviewed:
Known limitations
- No word list is universally complete.
- Different games accept different words and apply different rules.
- Proper nouns, abbreviations, slang and regional spellings may be handled differently by different lists.
- External puzzle data may occasionally be delayed or unavailable.
- Check each result against the rules of the specific puzzle you are playing.
Corrections and feedback
You can report missing words, incorrect results, outdated information, broken tools, or licensing and attribution concerns through the contact page or at abstractcs65@gmail.com.
Please include the tool or article URL, the input you used, the result you expected and the result you received.