How to use the mixed-script and look-alike word checker
- Paste the text, names or web addresses you want to check into the box.
- Leave the replace option on if you want a copy with look-alike letters swapped for Latin ones, or turn it off to get the text back unchanged.
- Read the counts, then the list of mixed-script words with their line and column.
- Check the second table for each letter that does not fit its word, with its code point and the Latin letter it imitates, then copy or download the checked text.
Formula
Split the text into words made of letters, marks and digits
Find the script of each letter: Latin, Greek or Cyrillic
A word with letters from two or more of these scripts is mixed
The main script is the one with the most letters in the word; the others are listed
Replacement maps a Greek or Cyrillic letter that is drawn like a Latin letter to that letter. Letters with no look-alike in the table are kept.
A fake payment address
Take the text "Sign in at paypal.com" where the first a of the word is the Cyrillic letter U+0430 and not the Latin a. The word looks the same, but it has letters from two scripts, so it is listed as Cyrillic + Latin, with the odd letter U+0430 that looks like a.
A second word that starts with the Greek capital Alpha (U+0391) is listed the same way. A word written wholly in Cyrillic, such as the Russian name of Moscow, is not flagged because it does not mix scripts. With replacement on, the checked text reads "Sign in at paypal.com or Apple.com".
Why look-alike letters matter
Many letters in different alphabets are drawn almost identically. The Cyrillic small letters that look like a, e, o, p, c, x and y, and Greek capital letters such as Alpha, Beta, Epsilon and Omicron, can take the place of Latin letters without any visible change. The strings are different to a computer, so a name can pass a quick look while pointing somewhere else, and a search for the Latin spelling will not find it.
Browsers and registries limit mixing in web addresses for this reason, but names in documents, usernames, file names and source code are often not checked at all. Looking for words that mix scripts is a simple and effective test, because a real word in ordinary text almost never contains letters from two alphabets.
Reading the results
A flagged word is a prompt to look, not proof of a problem. Some brand names, technical terms and mathematical notation do mix scripts on purpose, for example the Greek letter mu or a variable named with a Greek letter next to Latin ones. The tool shows the letters that stand out and what they imitate so you can decide.
The check covers Latin, Greek and Cyrillic, which account for most look-alike tricks. It does not find invisible characters, which the hidden character detector handles, and it does not know about look-alikes inside other scripts. Replacement is a convenience for the letters in its table. Review the result before you use it.
Assumptions and limits
- Only Latin, Greek and Cyrillic are compared. A word with letters from other scripts is not flagged for that.
- Words are runs of letters, marks and digits, so a hyphen, dot or apostrophe ends a word.
- The look-alike table covers common letters and is not the full Unicode confusables list.
- Nothing is uploaded. The text stays in your browser and up to 200,000 characters are read.
Frequently asked questions
Does it flag a normal Russian or Greek sentence?
No. A word written only in Cyrillic or only in Greek has one script and is not flagged. Only words that mix scripts are listed.
Why is a number plus a letter not flagged?
Digits have no script of their own here, so a word with digits and Latin letters is a single-script word.
Which letter counts as the odd one out?
The tool counts the letters of each script in the word. The script with the most letters is the main one, with Latin winning a tie, and the rest are listed.
Does it find invisible or zero-width characters?
No. Use the hidden character detector for those. This tool looks only at visible letters from different alphabets.