How to Check Whether a Word Is a Real Word

Someone insists that irregardless is not a word, or that emoji is, or that cwtch cannot possibly count. Before arguing, decide which of four questions is actually being asked, because each has a different answer and a different place to look. Is the word in a general dictionary? What label sits beside it? Is it on the list your game, spellchecker or publisher uses? Does the context accept it? We put 20 borderline words through all four, and the table below is the whole method; the sections that follow explain each row with what the test showed.

Question 1: is it in a general dictionary?

This is “real” in its plainest sense: has a reference work with editors, or a large community of them, recorded the word in use? Two free references answer it well, and they are worth using together because they disagree in instructive ways. Wiktionary (en.wiktionary.org/wiki/ followed by the word) is the widest net: written by volunteers, it covers slang and regional words that edited dictionaries take years to admit. In our test it drew the line between real and invented exactly where we had: entries for all 17 genuine words, nothing for quorify, brindlement and tressify, the three we made up to look plausible. Make sure you are reading the English section, though: a page can exist because the spelling is a word in another language, and a word can have an English section that says only “alternative form of” something else, as wifi does for Wi-Fi.

Cambridge Dictionary (dictionary.cambridge.org/dictionary/english/ followed by the word) is a professionally edited dictionary aimed at learners, and it was more conservative: 14 of the 17 real words, with unthaw, conversate and embiggen missing, and wifi held only under the trademark spelling Wi-Fi. Check the page heading before trusting it. When Cambridge has no entry it shows a search or suggestion page instead of an error, and for wifi it quietly sent us to Wi-Fi, so a page loading proves nothing until the headword at the top matches what you typed. Merriam-Webster is the standard American reference and is free to search, with usage notes on the contested words; we did not fetch it by script because the site does not allow that, so its verdicts are not in our table.

A real rare word gets at least one entry with a definition; an invented one gets none. That is the whole test at this level, and it is the one our real-or-fake word quiz is built on, with the fakes chosen to look as plausible as brindlement. Two independent references agreeing is a much stronger result than one, because dictionaries lag (a word can be in daily use for years before an edited dictionary admits it) and because Wiktionary is community-edited: an obscure entry may have been written by one person, and a page can exist for a word with very little real use behind it. If you check words often, ten lines of Python will pull the English labels from Wiktionary for any word. Run it as python3 realword.py irregardless; on our test it reported irregardless: English entry found; labels: [‘nonstandard|proscribed|sometimes|humorous’], petrichor: English entry found; labels: none, and quorify: no Wiktionary page.

import re, sys, urllib.request
word = sys.argv[1]
url = f"https://en.wiktionary.org/wiki/{word}?action=raw"
req = urllib.request.Request(url, headers={"User-Agent": "word-check/1.0"})
try:
    raw = urllib.request.urlopen(req).read().decode("utf8")
except urllib.error.HTTPError:
    sys.exit(f"{word}: no Wiktionary page")
m = re.search(r"==\s*English\s*==(.*?)(?:\n==[^=]|\Z)", raw, flags=re.S)
if not m:
    sys.exit(f"{word}: page exists, but no English section")
labels = re.findall(r"\{\{lb\|en\|([^}]*)\}\}", m.group(1))
print(f"{word}: English entry found; labels: {labels or 'none'}")

Question 2: what label does it carry?

An entry is not an endorsement. Dictionaries record ain’t and irregardless precisely so they can tell you that they are not standard, and skipping the label is how people end up citing a dictionary in support of a word the dictionary is warning them about. Wiktionary puts its labels in brackets at the start of each definition: (nonstandard, proscribed, sometimes humorous) for irregardless, (Wales) for cwtch. Cambridge’s labels are unusually clear because they are written for learners: informal, not standard, slang, specialized, Welsh English. Each family of label answers a slightly different question: how formal is it, where is it used, is it correct, and is it current.

informal, colloquial, slang

Real, current and correct in speech and casual writing; out of place in formal prose. Slang is the strongest of the three and usually the youngest: Cambridge and Wiktionary both give it to yeet, while selfie has settled down to informal.

dialect, regional, or a place name (Wales, Scottish English, US)

Standard within a region and unfamiliar outside it. Cwtch and bairn are not wrong anywhere; a reader in Texas may simply not know them. A place label on a disputed word, as with Cambridge’s “US, not standard” on irregardless, tells you where the dispute mostly happens.

nonstandard, not standard

In use, and widely regarded as an error or as uneducated. This is the label that settles most arguments: the word is real (it has an entry) and it is not accepted in careful writing (it has this label). Ain’t, irregardless, conversate and the “supposedly” sense of supposably all carry it.

proscribed

Wiktionary’s stronger version: usage guides actively advise against it. Unthaw gets it because un- should reverse the action and here does not; irregardless gets it because the ir- adds nothing to regardless.

dated, archaic, obsolete

Real but no longer current, in increasing degrees. Obsolete means no living writer uses it except deliberately; Wiktionary’s older verb yeet carries this label, which is a reminder to read the label next to the sense you actually mean.

formal, specialized, technical, literary

The opposite direction: the word is fully standard but belongs to a register. Petrichor is marked specialized at Cambridge; defenestrate is formal. Nobody will call these errors, but they may call them showing off.

humorous, jocular

Used with a wink. Wiktionary gives embiggen “nonstandard, usually jocular”, which is exactly how it is used.

trademark

A proprietary name that has become a common word. Cambridge marks Wi-Fi this way, which also explains why the capitalised hyphenated spelling is the headword.

Labels attach to senses, not to words, and that produced the result in the test we least expected. Most people who hear supposably assume it is a mangled supposedly, and in that sense Wiktionary labels it nonstandard. But it has an older literal sense, “able to be supposed”, which Wiktionary lists first without a label and which Cambridge records with no warning at all (one of its senses is even marked formal). Reading the first line of an entry is not the same as reading the entry; the label you need is the one beside the meaning you intend. Wiktionary’s labels can also be more opinionated than an edited dictionary’s (“sometimes humorous” reflects the view of whichever editors last touched the page), which is one more reason to read two references rather than one.

Question 3: is it on the list that governs your purpose?

A word game, a spellchecker and a publisher each run on a fixed list, compiled at some date for some purpose, and being real does not put a word on it. This is where the test produced its strangest column. We checked the 20 words against a 172,446-word list built from the public-domain ENABLE word list with offensive words removed, the list this site uses for word-game checks (how we check words explains it). It held four: bairn, defenestrate, irregardless and supposably. Two of those carry “not standard” labels in the dictionaries, while emoji, podcast and selfie, which nobody disputes and which have no labels at all, are missing. That is the difference between “compiled for games in a particular decade” and “correct”, laid out in one column, and we did not expect it from a list we know well.

None of it is a mistake in the list; it falls out of when and why the list was made. ENABLE was assembled for word games before emoji, selfie and podcast existed in English, and by design it contains no capitalised words, nothing with an apostrophe or hyphen, and nothing under three letters. So ain’t and Wi-Fi could never be in it whatever their standing, and cwtch, a word with no vowel letters, is not there either, although the older cwm and crwth are (see words with no vowels). It keeps irregardless and supposably because a game list records what players might play, not what an editor would allow. The reasons a real word can be missing from a game list are set out in the words you can’t play and why not; if a game rejects emoji, the word is not the problem, the list’s age is, and the game’s own rules are the only place to appeal.

A spellchecker is a different list again, tuned to produce few false alarms in ordinary documents rather than to judge usage. Hunspell, the free engine behind LibreOffice and Firefox, accepted 8 of the 17 real words with its American dictionary: it passed ain’t, gonna and irregardless, which occur constantly in real text, and rejected petrichor and cwtch, which hardly ever do. (Its British dictionary additionally accepted hangry and embiggen.) On a computer with it installed, echo hangry | hunspell -d en_GB -l prints the word back if that dictionary rejects it and nothing if it accepts it. A spellchecker is a frequency list with a spelling attached, and it should not be mistaken for a usage guide; the same goes, in the other direction, for a publisher’s style guide, which may ban a perfectly standard word for reasons of house taste.

Question 4: does the context accept it?

The last check is not a lookup at all; it is a judgement about the reader. Gonna is fine in dialogue and wrong in a covering letter. Petrichor is fine in a nature essay and showing off in a text message. Cwtch is exactly right for a Welsh audience and needs a gloss for anyone else. Keeping two words apart makes this judgement easier. Real means attested: people use the word and a reference has recorded that they do. Irregardless is real; so is ain’t; so, far more recently, is hangry, a blend of hungry and angry. Announcing that one of these “is not a word” is a claim about the dictionary that the dictionary itself contradicts. Accepted means allowed by a particular authority for a particular purpose: a game’s word list, a newspaper’s style guide, an examiner’s marking scheme, a spellchecker’s dictionary. Irregardless is real and is not accepted in formal writing; emoji is real and is not accepted by our game list; cwtch is real, accepted by Cambridge, and rejected by the spellchecker.

The practical rule that follows is to name the authority before arguing. “Is irregardless a word?” has no single answer. “Is irregardless in the Cambridge Dictionary?” (yes, labelled not standard), “would a careful editor let it through?” (no) and “will this game score it?” (depends on the game; ours does) all have one.

Three words taken through all four questions

supposably. Question 1: yes, in both dictionaries. Question 2: nonstandard in the sense “supposedly”, unlabelled in the older sense “able to be supposed”. Question 3: on the game list; rejected by the spellchecker. Question 4: for a piece of writing, use supposedly unless you mean the rare sense and are prepared to be misread.

cwtch. Question 1: entries in both dictionaries. Question 2: Welsh English; no warning label. Question 3: fails the spellchecker and the game list, for reasons that have nothing to do with correctness (rare in ordinary documents; no vowel letters and no history in a list compiled elsewhere). Question 4: real and accepted wherever the reader knows the word.

brindlement. Question 1: no Wiktionary page, no Cambridge entry. The other three questions do not arise; this is what a fake looks like, four blanks in a row, and the spellchecker and the list both rejected it too.

How we checked these twenty words

The 20 words were chosen to cover the kinds of word people argue about: recent coinages (emoji, selfie, wifi, podcast, hangry, yeet), informal contractions (ain’t, gonna), regional words (cwtch from Wales, bairn from Scotland and northern England), disputed forms (irregardless, unthaw, conversate, supposably), a technical term (petrichor), a jocular coinage (embiggen), a formal rarity (defenestrate), and three invented look-alikes (quorify, brindlement, tressify). Each was looked up by script on the same day in four places: the English section of Wiktionary, fetched from en.wiktionary.org/wiki/<word>, recording whether an English entry exists and every usage label attached to a definition; the Cambridge Dictionary at dictionary.cambridge.org/dictionary/english/<word>, checking the page title to confirm a real entry rather than a search page and reading the labels in the entry; Hunspell 1.7.2 with its American dictionary; and the 172,446-word game list.

Two traps in the automated checks are worth passing on, because a reader doing this by hand can fall into both. A page loading is not an entry: Cambridge redirected wifi to Wi-Fi and served a search page for the missing words, so the test has to be “does the headword match”, not “did the page load”. And a label can attach to one sense but not another, as with supposably and yeet. Our word list is also one list among many; other games use others, and we make no claim about what any commercial game accepts beyond what its own published rules say.

Grounded in what each check caught, the order we would use is: Wiktionary first, because it separated real from invented perfectly and shows the labels in the first line; Cambridge (or Merriam-Webster) second, as the edited second opinion, checking the headword and reading the label on the sense you mean; then the list that governs your actual purpose, taking its verdict as a rule of that house and nothing more; and context last, which no lookup can do for you. If the purpose is a word game, a rejection is usually the list’s age or its rules rather than a comment on the word; on the list this site uses, bairn and defenestrate score and emoji does not, and anyone curious can see how that feels in sixty seconds.