Unicode Lookalike Detector
Preparing workspace…
A homoglyph detector examines text for characters that may resemble familiar letters while belonging to different Unicode writing systems. Paste a word, domain spelling, filename, or paragraph to see selected confusable code points, their positions, script families, and possible Latin lookalikes. The checker also identifies certain invisible and bidirectional formatting characters. It does not decide whether the author intended deception, and it does not contain the complete Unicode confusables database. Multilingual writing can legitimately contain many scripts, while sophisticated spoofing may use characters outside the mapping. Treat the results as an inspection aid rather than an automated accusation or a guarantee of uniqueness.
Homoglyph Detector At a Glance
| Aspect | Details |
|---|---|
| Input | Unicode text, names, identifiers or domain spellings |
| Modes | Original code points, NFC, or NFKC normalization |
| Output | Selected mapped lookalikes, invisible controls, code points and script families |
| Position units | Unicode code points rather than display grapheme clusters |
| Boundary | Finite mapping; multilingual text and unmapped lookalikes need human review |
Why Different Code Points Can Look the Same
Unicode assigns a unique code point to each encoded character, but typography is designed for readable visual forms rather than an infallible identity system. A Cyrillic letter may resemble a Latin one when rendered in a common font even though their byte sequences and meanings differ. The text pаypal can be written with a Cyrillic small a in place of the Latin a, and casual readers may not notice the change. A program comparing strings by exact code points will notice it immediately. The detector displays the code point and script name so you can identify the actual character rather than judging from a screenshot or a typeface that may conceal the distinction.
Script Families and Mixed-Script Text
Writing systems are organized into scripts such as Latin, Cyrillic, Greek, Arabic, Han, or Devanagari. The analyzer labels characters from a selection of these script families and records when more than one recognized family appears in the supplied text. Mixed-script content is common in academic writing, translation, place names, international customer support, and bilingual communication. It is not a security finding by itself. Suspicion depends on context: a domain that resembles a brand might deserve scrutiny, while a sentence quoting two languages does not. Review the word boundaries and intended language before deciding whether a mixed-script notice represents an actual usability or identity concern.
The Meaning of a Confusable Mapping
A mapping such as Cyrillic о to Latin o means the two glyphs can appear visually similar in some fonts. It does not assert that the characters are interchangeable in language, pronunciation, search, or identification. Some pairs look nearly identical only at certain sizes, weights, or screen resolutions. The inspector uses a finite set of curated, frequently confusing characters instead of claiming full coverage. A result may miss historical scripts, mathematical letter forms, ligatures, or font-specific ambiguities. For important account identifiers, combine code-point inspection with a broader identity and naming policy rather than relying on one small lookalike list.
Unicode Normalization Versus Visual Similarity
Normalization converts selected equivalent Unicode representations into a standard sequence. NFC commonly combines compatible sequences where the Unicode standard defines canonical equivalence, while NFKC also applies compatibility transformations that can change the presentation of some symbols. Neither normalization method universally turns Cyrillic letters into Latin ones, because those letters remain distinct encoded characters. The workspace allows the original text, NFC, or NFKC to be inspected and reports the analyzed version separately. This helps reveal cases where normalization changes character count or representation. It should not be used blindly on legal names, programming identifiers, or documents where the precise original byte sequence matters.
Invisible and Bidirectional Formatting Characters
Some Unicode characters affect line breaks, joining, or display direction without producing an obvious printed glyph. Zero-width marks can appear in normal language processing, and bidirectional controls can be needed in mixed right-to-left and left-to-right documents. In other contexts they can make filenames, labels, or code appear different from their stored order. The checker flags a chosen group of zero-width and bidirectional formatting characters so their existence becomes visible in a report. Removal is not always the right response. Before deleting a flagged character, determine whether it serves a legitimate language or formatting purpose, especially in Arabic, Persian, or multilingual material.
Examining a Domain Name or Email Address
A suspicious domain might substitute one visually confusable character into a familiar name. Copying the text directly into the detector can reveal whether the spelling contains mixed scripts or mapped lookalikes. However, an internationalized domain can also be entirely legitimate, and the browser may display its ASCII-compatible punycode encoding instead of the Unicode form. The detector does not query ownership records or verify that a domain resolves to a safe server. If you are evaluating an unexpected message, compare the complete domain through an independently verified contact channel. Do not visit the destination just because the detector found no mapped substitutions.
Practical Example: Similar Product Identifiers
Imagine two inventory labels that appear as ITEM-001 in a spreadsheet, but one contains a non-Latin character that resembles an ordinary letter. A sorting or duplicate-removal routine may preserve both labels because their code-point sequences differ. Copy each string into the checker and compare the code points in the flagged positions. An exact-text comparison can then determine whether the identifiers are byte-for-byte or character-for-character equivalent. The underlying problem may be accidental copy-and-paste from translated material rather than malicious impersonation. Resolve it through a documented identifier format so people know which character set is valid for that business process.
Reviewing Code, Filenames, and Variables
Visually similar identifiers can confuse code reviews, error logs, and file listings. A variable declared with a non-Latin letter might resemble another variable in the same file and lead a reviewer to misunderstand which value is used. Formatting controls can also alter how a source line is displayed in some editors without altering the stored character order. When reviewing code from an untrusted source, examine exact bytes and Unicode escapes using a trusted development environment. The detector can expose selected problems in copied text, but it is not a compiler, static-analysis engine, or substitute for language-specific restrictions on identifier characters.
Fonts, Screen Readers, and Visual Differences
Two characters that look identical in one typeface may appear noticeably different in another. Screen readers can also pronounce letters or names differently based on language settings, but that behavior varies by software and context. A visual scan of a word is therefore not sufficient to establish identical meaning or rendering for all users. The tool reports code points precisely because the underlying identity is more stable than a screenshot. If users repeatedly confuse labels in a product interface, consider clearer typography, explicit identifiers, and accessible descriptions rather than depending on subtle glyph distinctions that disappear at smaller display sizes.
Understanding the Position and Count Report
The positions in the report count Unicode code points from the analyzed string, not bytes or extended grapheme clusters. A single visible grapheme may contain a base letter followed by combining marks, and an emoji sequence may include several code points joined into one displayed symbol. A positional count is therefore most useful for locating suspect encoded characters in a text editor that exposes code points, not for predicting cursor movement in every application. The report also summarizes recognized script families and gives a special label to invisible controls, making the output easier to read without pretending to reconstruct every grapheme boundary.
When Not to Remove a Flagged Character
Removing a character solely because it is outside ASCII can damage names, cultural context, mathematical notation, and legitimate localized content. Normal Arabic or Greek writing does not need Latin replacements to be authentic. Even invisible characters may be functionally important for right-to-left text or character shaping. Before normalizing or replacing data, identify the expected language and the technical requirement. If an identifier must be restricted to ASCII, validate that at data entry with a documented rule. For prose, preservation and clear language metadata are usually more important. A flagged result should initiate inspection, not indiscriminate transliteration.
Saving Evidence of an Encoding Problem
When an encoding mismatch causes a production defect, store the exact original string, the displayed text, code point values, and the application context. Include information about the input method and platform if the issue appears only on one device. This makes it easier to reproduce a sorting error, display mismatch, or identifier collision. The JSON report can be attached to an engineering issue without replacing the real source input. If the text includes customer information, sanitize the report before sharing it. An accessible palette builder can address color legibility issues elsewhere in an interface, but it cannot make two underlying Unicode identifiers identical.
Frequently Asked Questions About Homoglyph Detection
Is a mixed-script warning proof of fraud?
No. Multilingual text regularly mixes writing systems. The warning identifies a fact about the characters present; the intended language, context, and claimed identity must be reviewed before drawing conclusions.
Can NFC or NFKC remove every deceptive lookalike?
No. Unicode normalization handles defined equivalence relationships, not every visual resemblance between unrelated characters. Greek and Cyrillic letters often remain distinct from Latin letters after normalization.
Why does the report show more code points than visible letters?
Some displayed symbols contain combining marks, variation selectors, or joined sequences. The report counts code points for locating encoded characters, not the grapheme clusters seen by a reader.
Can I scan an entire website with this tool?
Not directly. This workspace analyzes pasted text only. It does not crawl web pages or retrieve remote domain registrations, and it does not make a security determination about sites.
Does the detector include every Unicode confusable character?
No. It uses a deliberately limited visual-similarity map and selected invisible controls. For identity-critical checks, use established Unicode security profiles and validated confusable data alongside context-specific rules.