Find lookalike letters from another alphabet

Latin P, Cyrillic Р and Greek Ρ look the same and are three different characters. This checker finds every word that mixes Latin, Cyrillic and Greek letters, shows which letters are the odd ones, and gives the word back in one alphabet. It lists invisible characters too.

Check a text

The example has four mixed words and a zero-width space; paste your own text over it. Nothing leaves your browser.

Why one odd letter matters

A reader sees nothing wrong with Pусский. Software sees another word, because its first letter is the Latin P:

Where the odd letters come from

Most are typed by shape: the Latin P key for Р, a Latin i for the Ukrainian і, a Cyrillic о in an English word typed with the Cyrillic layout still on. Others come from text recognition of scanned pages, from copying out of PDF files, and from editing a word of one language letter by letter into another. Some are put there on purpose, to get past spam filters and plagiarism checks, or to make a fake address.

How the checker reads a text

A word is a run of letters, with their accents and other combining marks. Anything else ends it: a space, a digit, a hyphen, an apostrophe, a tag. So USB-кабель is two words, USB and кабель, and neither is mixed.

A word is mixed when it has letters of two of the three alphabets. Its own alphabet is the one most of its letters are in; on a tie, the one most letters of its line are in. The letters of the other alphabets are the odd ones. When every odd letter has a lookalike in the word's own alphabet, the checker spells the word in that alphabet. When one has none, like N in a Cyrillic word, the word is listed for you to retype.

Invisible characters are listed apart: zero-width spaces and joiners, the word joiner, a byte order mark inside the text, soft hyphens, direction controls, control characters, and the replacement character � that a wrong text encoding leaves behind. Joiners next to emoji, and in scripts that need them, such as Persian and the Indian scripts, are not listed.

Letters that look the same

Each row is one shape in the alphabets that have it. The checker makes its corrections from this table; where one letter has two rows, the first one is used.

Lookalike letters in the Latin, Cyrillic and Greek alphabets, with their Unicode code points.
LatinCyrillicGreekNote
A U+0041Capital AА U+0410Capital AΑ U+0391Capital Alpha
a U+0061Small Aа U+0430Small A—
B U+0042Capital BВ U+0412Capital VeΒ U+0392Capital Beta
C U+0043Capital CС U+0421Capital Es—
c U+0063Small Cс U+0441Small Es—
E U+0045Capital EЕ U+0415Capital IeΕ U+0395Capital Epsilon
e U+0065Small Eе U+0435Small Ie—
Ë U+00CBCapital E With DiaeresisЁ U+0401Capital Io—Cyrillic Ё: Russian, Belarusian
ë U+00EBSmall E With Diaeresisё U+0451Small Io—
È U+00C8Capital E With GraveЀ U+0400Capital Ie With Grave—Cyrillic Ѐ: Macedonian, as in сѐ
è U+00E8Small E With Graveѐ U+0450Small Ie With Grave—
H U+0048Capital HН U+041DCapital EnΗ U+0397Capital Eta
h U+0068Small Hһ U+04BBSmall Shha—Cyrillic һ: Bashkir, Kazakh, Tatar
I U+0049Capital IІ U+0406Capital Byelorussian-ukrainian IΙ U+0399Capital IotaCyrillic І: Ukrainian, Belarusian, Kazakh
i U+0069Small Iі U+0456Small Byelorussian-ukrainian I—
Ï U+00CFCapital I With DiaeresisЇ U+0407Capital Yi—Cyrillic Ї: Ukrainian
ï U+00EFSmall I With Diaeresisї U+0457Small Yi—
J U+004ACapital JЈ U+0408Capital Je—Cyrillic Ј: Serbian, Macedonian
j U+006ASmall Jј U+0458Small Je—
K U+004BCapital KК U+041ACapital KaΚ U+039ACapital Kappa
—к U+043ASmall Kaκ U+03BASmall Kappa
M U+004DCapital MМ U+041CCapital EmΜ U+039CCapital Mu
N U+004ECapital N—Ν U+039DCapital NuCyrillic has no N: И is another letter
O U+004FCapital OО U+041ECapital OΟ U+039FCapital Omicron
o U+006FSmall Oо U+043ESmall Oο U+03BFSmall Omicron
P U+0050Capital PР U+0420Capital ErΡ U+03A1Capital Rho
p U+0070Small Pр U+0440Small Erρ U+03C1Small Rho
S U+0053Capital SЅ U+0405Capital Dze—Cyrillic Ѕ: Macedonian
s U+0073Small Sѕ U+0455Small Dze—
T U+0054Capital TТ U+0422Capital TeΤ U+03A4Capital Tau
X U+0058Capital XХ U+0425Capital HaΧ U+03A7Capital Chi
x U+0078Small Xх U+0445Small Ha—
Y U+0059Capital YУ U+0423Capital UΥ U+03A5Capital UpsilonCyrillic У is a little different in most fonts
Y U+0059Capital YҮ U+04AECapital Straight U—Cyrillic Ү: Kazakh, Mongolian, Tatar
y U+0079Small Yу U+0443Small U—
Z U+005ACapital Z—Ζ U+0396Capital Zeta
u U+0075Small U—υ U+03C5Small Upsilon
v U+0076Small V—ν U+03BDSmall Nu
—Г U+0413Capital GheΓ U+0393Capital Gamma
—П U+041FCapital PeΠ U+03A0Capital Pi
—Ф U+0424Capital EfΦ U+03A6Capital Phi
d U+0064Small Dԁ U+0501Small Komi De—Cyrillic ԁ, ԛ, ԝ: Komi and Kurdish letters, seen mostly in phishing addresses
q U+0071Small Qԛ U+051BSmall Qa—
w U+0077Small Wԝ U+051DSmall We—
Q U+0051Capital QԚ U+051ACapital Qa—
W U+0057Capital WԜ U+051CCapital We—

Questions

How do I find Latin letters in a Russian or Ukrainian text?

Paste the text above. Every word with a Latin letter among Cyrillic ones is listed with its line, and the corrected text is ready to copy. Find in a text editor cannot do it: the lookalike letter is what makes the word look right.

How do I find Cyrillic letters in an English text?

The same way. An English word with a Cyrillic а or о in it is mixed, and the checker gives it back in Latin letters.

Is every mixed word a mistake?

In running text, nearly every one. The exceptions are names styled that way on purpose, such as a brand that puts a Latin letter into a Cyrillic word, and codes or formulas. The checker lists those too; skip the ones you meant.

Does it find a whole word typed in the wrong alphabet?

No. OCA in Latin letters inside a Russian sentence reads as the Russian word ОСА, and no Cyrillic letter gives it away. Telling it from an abbreviation such as CEO or IT would take the language of every sentence, so the checker only marks words that mix alphabets.

What about digits that look like letters?

A digit ends a word here, so a 3 typed for З is not marked. Words with digits on purpose, like mp3 or 4K, are too common to list every one.

What happens to the text I paste?

Nothing outside your browser. The page uploads nothing; the check runs in its own script, and the text is gone when you close the tab.

More checks for translations

translation-checks, free under the MIT licence, fails a translation with a mixed word since version 1.5.0, in CI or as a pre-commit hook, along with lost placeholders, changed numbers and wrong plural forms. Run over the 613,570 page translations our own sites had published, it found 363 with such a word that our earlier checks had let through: 190 of them Macedonian сѐ and нѐ typed with the Latin è, the rest slips of the kind in the example above. Polyglot, one of our kits, translates catalogues and sites with a local model and puts every line through the same checks before it is written, and the Plural-Forms checker tests the plural rule of a gettext catalogue against CLDR for every language.

Updated: