Spelling Rankings

Spelling rankings collect the words that English writers struggle with most, the language pairs that spell-checkers cannot disambiguate, and the structural properties of words that make them hard to memorise. Every ranking on this hub is pre-computed from PlainSpell's underlying dictionary data, entries pulled from Wiktionary's kaikki.org JSON dumps and an open word-frequency list. We do not estimate, interpolate, or synthesize values. Each ranking's positions are stable between data refreshes and reproducible across visits. If a row looks wrong, the upstream Wiktionary page can be inspected to verify the underlying claim, every ranked entry traces back to a real dictionary record, not to a generated guess.

The five rankings on this page each measure a different dimension of spelling difficulty. The most-misspelled list counts how many distinct edit-distance misspelling variants each English word generates, a longer list of variants means more ways the word can go wrong in practice. The most-confusable pairs combine word frequency with visual edit distance: when both words are common AND look alike, substitution errors become almost inevitable, which is why their/there and than/then live near the top of that list. Longest-words filters Wiktionary's full lemma set to the words that also carry a documented frequency rank, distinguishing real-world long words from dictionary curiosities or chemical nomenclature. Largest homophone groups bucket Wiktionary's IPA field, when three or more lemmas share an identical pronunciation, those words form a group, and the group size is the rank value. Languages-by-size simply counts Wiktionary entries per language pack, exposing how much of each language's vocabulary the open-source corpus has captured.

Reading a ranking page is most useful when paired with the detail pages it links to. A ranked word like accommodate on the most-misspelled list is also reachable at its individual word page, where IPA pronunciation, etymology, part-of-speech tags, and the specific list of recorded misspellings appear together. A ranked confusable pair such as their / there / they're opens into a comparison page that shows each lemma's distinct meaning, the frequency gap between them, and the contexts in which substitution typically happens. The ranking is a discovery layer; the detail page is the evidence layer. Both are sourced from the same Wiktionary base, no editorial scaffolding is added on top.

Methodology specifics live on the methodology page, but the short version is: rankings are deterministic functions of public open-source data. We do not poll users, we do not weigh by editorial opinion, and we do not surface a row unless the underlying dictionary record supports it. When new Wiktionary dumps ship, the rankings are rebuilt, at which point a word may move up or down a few positions, or a new lemma may enter the list. Old positions are preserved in an internal change log for the current quarter so changes are traceable. If you spot what looks like an error, the about page documents the contact path for corrections; the source for every row is a public URL anyone can verify.

Beyond rankings, PlainSpell exposes browsable word lists at /en for English, plus parallel sections for the other supported languages. Spelling guides cover the patterns that drive the rankings above, silent letters, doubled consonants, Greek and Latin roots, vowel digraphs, and the homophone clusters that English inherited from centuries of borrowing. Together, the ranking hub, the language browse pages, and the editorial guides form one cross-linked surface where any ranked word can be followed back to its dictionary record, its phonetic class, and the spelling principle that makes it hard.

A practical note for writers and educators: the rankings are most useful as a curriculum input, not as an isolated list to memorise. Most-misspelled positions correlate with the words that produce the highest rate of spell-check overrides in published text, which makes that ranking a strong candidate list for spelling drills. Most-confusable pairs are the targets that grammar-checkers tend to miss because both members of each pair are valid dictionary entries; reviewing the top of that list catches a disproportionate share of substitution errors before they reach a reader. Longest-words and largest-homophone-group lists are more useful as exploratory surfaces, they reveal patterns in word formation and pronunciation history rather than prescriptive items to drill. Languages-by-size, finally, is a coverage report on the open dictionary corpus itself: it indicates which languages PlainSpell can answer the most detailed questions for and where the data is thinnest. Taken together, the five rankings give a complete picture of where spelling difficulty concentrates across the modern English lexicon and the four other languages PlainSpell currently indexes.