Developer · Unicode
Unicode confusable & homograph detector.
Paste a username, domain label or identifier to see which characters have lookalikes, whether it mixes scripts, and what its UTS #39 skeleton is. Add a second string to test whether the two can be confused. Runs in your browser; nothing is sent anywhere.
Input
Loading confusables data…
Result
Enter a string and press Analyse.
Worked example
The string pаypal looks like paypal, but its second letter is U+0430 CYRILLIC SMALL LETTER A, not U+0061. The tool reports two scripts (Latin and Cyrillic), flags the string as mixed-script, lists U+0430 as mapping to a, and shows the skeleton paypal. Compared with the real paypal, the two strings are different text with identical skeletons, so they are confusable.
How the skeleton is computed
UTS #39 defines skeleton(X) in four steps: convert to NFD; remove characters with the Default_Ignorable_Code_Point property (zero-width characters, for example); replace each character with its prototype from confusables.txt, leaving it alone if it has no entry; convert to NFD again. Two strings are confusable when their skeletons are equal. The prototype is a canonical stand-in, not necessarily the letter you would expect, so a skeleton is for comparing, not for display. For example, m has the skeleton rn, because the data treats the pair rn as a lookalike of m.
Mixed-script detection
Each character is looked up by its Unicode Script_Extensions value. Digits, punctuation and combining marks are ignored. A string is flagged when no single script fits every remaining character. Han with Hiragana and Katakana (Japanese), Han with Hangul (Korean), and Han with Bopomofo are allowed combinations, as in UTS #39 section 5.1. The tool recognises 42 common scripts; a character from any other script is shown as “Other”.
Data and as-of date
The mapping is a trimmed copy of confusables.txt from Unicode 18.0.0, file dated 2026-08-06, with 6,712 source characters, taken from unicode.org/Public/security/latest/confusables.txt. The page shows the version loaded. Unicode revises the file with each release, so a newer release may add or change entries.
What this does not do
It does not reorder right-to-left text (the bidiSkeleton step), does not apply the identifier-status restrictions of UTS #39, and does not know how your fonts draw a character. It only knows the pairs in the data, so lookalikes that are not listed are not found, and it cannot tell you which real name a string is imitating. In the data the digit 1 and the letter l share a skeleton, so paypa1 and paypal compare as confusable.
Sources: Unicode confusables.txt; Unicode UTS #39, Unicode Security Mechanisms.
What “look-alike” means in Unicode
Unicode defines well over 100,000 characters, and many of them are drawn almost identically in common fonts. The Cyrillic letter “а” (U+0430) and the Latin “a” (U+0061) are separate characters with separate code points, yet they look the same, so a string can read as paypal to a person while containing a different character. Security people call such strings homographs or confusables. They matter for domain names, usernames, package names, file names and source code identifiers, anywhere a person decides whether two names are the same by eye.
Three ways to read the result
Mixed-script. If one string uses Latin and Cyrillic together, or Latin and Greek, that is the strongest warning sign, because a real name in a single language rarely does that. Confusable characters. The list shows which characters in your string have a listed lookalike and what they map to. A plain ASCII word can appear here, since the digit 1, the capital I and the lowercase l all map to each other in the data. Skeleton. The skeleton is a normal form for comparing strings; if two different strings have equal skeletons, they are confusable with each other.
More worked examples
All results below come from this tool’s data (Unicode 18.0.0).
раураl, typed with five Cyrillic letters (р, а, у, р, а) followed by a Latin “l”, has the skeletonpaypaland is flagged mixed-script (Cyrillic and Latin). Compared with the realpaypal, the two are confusable.аррӏеis written entirely in Cyrillic (including the letter ӏ, U+04CF). It is not mixed-script, so a mixed-script rule would pass it, yet its skeleton isapple. This is the whole-script confusable case, and it is why the skeleton comparison exists alongside the script check.googIe, with a capital I where the lowercase l belongs, is plain Latin: one script, no non-ASCII characters. Its skeleton isgooglebecause I and l share a prototype.mhas the skeletonrn, so a name with “rn” in it can be confusable with the same name using “m”. Comparernwithmand the tool reports them confusable even though both are lowercase ASCII.日本語abcis flagged as mixed-script (Han and Latin). That is only a signal: mixing Latin letters or digits with Japanese or Chinese text is normal in real names, and the check is a prompt to look closer, not a verdict.
How to use it for a domain, username or package name
Paste the candidate in the first box and, if you have the legitimate name, paste it in Compare with. If the two skeletons are equal and the strings differ, treat them as lookalikes and decide which one you own. For domain registration and display, browsers and registries apply their own policies on top of this data, and internationalized domain names are encoded in Punycode (an ASCII form beginning xn--) on the wire, so a lookalike can often be spotted by viewing that form. This tool does not convert to Punycode.
Invisible characters
The skeleton step removes characters with the Default_Ignorable_Code_Point property, such as zero-width spaces and joiners. A string with a hidden zero-width character therefore matches the visible one when compared here, although a database or file system would treat them as different. The result lists ignorable characters it found so you can see them.
Limits worth remembering
Confusability depends on the font. The data records lookalikes that Unicode considers visually close in common fonts, not what your screen draws; a pair can be identical in one typeface and distinct in another. And the data is a snapshot: a new Unicode release can add or change entries, so a clean result today is not a guarantee about the future.
Sources
Each source below was opened and checked in October 2026 (retrieved 2026-10).
- Unicode Technical Standard #39, Unicode Security Mechanisms: confusable detection, skeleton definition and mixed-script rules
- Unicode confusables.txt: the character mapping data this tool uses
- Unicode Technical Report #36, Security Considerations: background on visual spoofing and homographs
- RFC 5890, Internationalized Domain Names for Applications: Definitions and Document Framework: U-labels, A-labels and the xn-- prefix
Frequently asked questions
Why does an ordinary word get flagged?
The data lists lookalikes of many everyday characters, including the digit 1, the letters l and I, and the letter m. A skeleton difference only means the string contains a character that has a lookalike, not that anything is wrong. Mixed-script is the stronger signal; confusable characters inside a single script are usually harmless.
Why was nothing flagged when two strings looked the same?
The skeletons only match when every difference is covered by the data. Characters can look alike in your font without being listed, and some lookalikes are listed only in one direction through a shared prototype.
Are real names in other languages false positives?
Often. A Greek name written entirely in Greek is one script and is not flagged as mixed, but its letters may still have Latin lookalikes in the data. Mixed Latin and Cyrillic text is legitimate in some languages and in brand names.
Does the tool send my text anywhere?
No. The data file is downloaded once and all analysis runs in your browser.
What is a homograph attack?
It is an attempt to deceive people with a name that looks like another one, usually by substituting characters from a different script, such as using a Cyrillic а in place of a Latin a in a domain or username. The two strings look the same to a person and are different to a computer.
What is a confusable skeleton?
It is the normal form defined by Unicode Technical Standard 39. The text is normalised, ignorable characters are dropped and each character is replaced by its prototype from the confusables data. If two different strings have the same skeleton, they are considered confusable.
Why is the mixed-script check not enough on its own?
Because a string written entirely in one script can still imitate another. The Cyrillic word аррӏе looks like apple but is one script only. The skeleton comparison catches that case; the script check catches strings that blend scripts.
Can plain ASCII text be confusable?
Yes. The data treats the digit 1, the capital I and the lowercase l as lookalikes of each other, and the letter pair rn as a lookalike of m. That is why googIe (with a capital I) has the skeleton google.
How do I check two names against each other?
Enter one in String to check and the other in Compare with, then press Analyse. The tool reports whether they are identical, different but confusable, or unrelated, and shows both skeletons.
Does this tool convert text to Punycode?
No. It analyses characters, scripts and skeletons only. Internationalized domain names are carried in DNS as ASCII Punycode labels beginning xn--, but that conversion is outside this tool.
Is a flagged name necessarily malicious?
No. Many legitimate names contain lookalike characters or mix scripts, such as Japanese text with Latin letters. A flag means look closer; a clean result does not mean a name is safe.
No security guarantee
This tool reports what the Unicode confusables data and script properties say about a string. A clean result does not mean a name is safe, and a flag does not mean it is malicious. It is not a substitute for a registry policy, a browser’s own IDN display rules, or a security review.