Greek Alphabet

What is the Greek alphabet?

The Greek alphabet has 24 letters, running from alpha to omega. It has been in continuous use for roughly three thousand years, which makes it the oldest alphabet still written today that records vowels as letters in their own right — the innovation it is remembered for. Modern Greek uses it as an ordinary writing system, the way English uses the Latin one.

Almost everyone else uses it as a supply of symbols. Pi is a circle constant, sigma a sum, delta a change, lambda a wavelength and a function, mu a millionth, omega an ohm. Statistics leans on alpha, beta and chi; particle physics has run through most of the alphabet twice. American fraternities and sororities name themselves with it, and hurricanes were once named with it after the Latin list ran out.

The table above is generated from the Unicode Character Database, so the letters, their names and their code points come from the file that defines them rather than from a copy of a copy. Each letter shows its capital and small forms, the two code points, and whether Unicode marks it as confusable with a Latin letter. Click any letter to copy it.

How to use this page

1. Find the letter and click it. Clicking either the capital or the small form copies that character to your clipboard. The code point column shows exactly which character you are getting, which matters more than it sounds — several Greek letters have Latin twins that look identical and are not the same character at all.
2. Paste a suspicious string into the box. If a username, domain or filename contains Greek letters standing in for Latin ones, the checker lists them with their position, their Unicode name and the letter each is imitating, then shows you what the text reads as once they are swapped back.
3. Read the lowercasing comparison. The box lowercases your text twice: once as a whole string, and once one character at a time. For every language but Greek those two results are identical. When they differ, you are looking at the bug described below.

Why lowercasing Greek is not a per-character job

Sigma is the only Greek letter with two lowercase forms. Ordinary lowercase sigma is σ, U+03C3. At the end of a word it is written ς, U+03C2, the final sigma. Which one is correct depends entirely on where the letter sits in the word, and that makes Greek the one script where converting case is not a lookup table.

You can watch it happen. The Greek word for street, ΟΔΟΣ, lowercases to a word ending in ς — final sigma, U+03C2 — because the software can see that the sigma is the last letter. Lowercase the same four characters one at a time and the result ends in σ, U+03C3, because a single character has no word to be at the end of. Two spellings of the same word, from the same input, differing by which function you reached for.

We swept all 24 letters to see how special this is: sigma is the only one where the two methods disagree. Put any other Greek letter at the end of a word and both approaches give the same answer. The rule is Unicode's Final_Sigma condition, and it is the only context-sensitive case mapping in the default algorithm — everything else really is a table.

What makes this bug durable is that it is invisible everywhere else. We checked English, Spanish, Portuguese, French, German and Russian: per-character lowercasing is correct for all of them, accents and all. So a hand-rolled loop, a tokeniser that lowercases as it splits, or a slug generator that works character by character passes every test its author is likely to write, and then quietly produces a second spelling of every Greek word ending in sigma.

There is a related trap in round trips. Uppercasing Greek and then lowercasing it returns the original for all 24 letters. Going the other way does not: σοφοσ, written with an ordinary sigma at the end, uppercases to ΣΟΦΟΣ and comes back as σοφος with a final sigma. Both sigmas uppercase to the same Σ, so the distinction is destroyed before the lowercase step ever runs. Case-folding Greek to compare two strings is safe. Case-folding it and storing the result is not.

Letters that are not the letters they look like

Fourteen of the 24 capitals are marked by Unicode as confusable with a Basic Latin letter: alpha with A, beta with B, epsilon with E, zeta with Z, eta with H, iota with l, kappa with K, mu with M, nu with N, omicron with O, rho with P, tau with T, upsilon with Y and chi with X. Every one of those pairs is two different characters that your eyes cannot tell apart.

The pairs are worth reading slowly, because the shapes do not follow the sounds. Rho is an r, but it is shaped like a P. Eta is a long e, but it is shaped like an H. Chi is a k-ish sound written X, and upsilon is a u written Y. Anyone transliterating by shape rather than by sound gets these backwards, which is its own small reason to keep the table handy.

That list is not our judgement. It comes from confusables.txt, the Unicode data file behind UTS 39, the security mechanisms that domain registrars and browsers use to catch spoofed identifiers. So the ten capitals not on the list — delta, theta, lambda, xi, pi, sigma, phi, psi, omega and gamma — are absent because Unicode does not consider them confusable, not because nobody looked. Eight lowercase letters are flagged too, including the mildly alarming sigma-for-o and gamma-for-y.

The practical part: NFKD normalisation folds none of the fourteen. Normalising is the usual reflex for comparing strings that might be written differently, and it genuinely fixes accents, ligatures and fullwidth characters — but a Greek capital eta standing in for an H survives it untouched, on both sides, still unequal. A lookalike check built on normalise-and-compare will not catch a single one of these. That is exactly the gap the checker above fills.

Honest limits

The names in the table are Unicode's names for the characters, kept in their Unicode spelling rather than translated. Greek letters have perfectly good names in every language on this site, and if you are writing prose you should use them — but the point of this column is to say which character you are looking at, and the Unicode name is the one that identifies it unambiguously.

Whether two letters really look alike depends on the typeface. Unicode's confusable list is a general judgement, and any given font can widen or narrow the gap: some faces give Greek capitals slightly different proportions from their Latin twins, and a few make normally distinct letters harder to tell apart. Treat the column as a warning about which pairs to check, not as a claim about the font on your screen.

Most importantly, finding a confusable letter does not mean anything is wrong. An ordinary Greek word is full of them — ΑΘΗΝΑ is almost entirely letters that Unicode marks confusable, and it is simply the name of a city. What actually signals trouble is mixing scripts inside a single run, because a word that is part Greek and part Latin is not a word in either language. The checker reports that separately, and it is the line to pay attention to.

Why is it free?

Because it is a table and a few string comparisons. Everything on this page runs in your browser: nothing you paste into the checker is uploaded, nothing is stored, and no server is involved in reading a list of 24 letters.

So there is no account, no sign-up and nothing held back. The script that generates the table from the Unicode Character Database and confusables.txt is in the repository beside the page, along with the 128-assertion test suite that checks the sigma behaviour, the round trips and the homoglyph claims — including the negative controls that stop those tests passing for the wrong reason.