Paste any text and find out. 98 languages, instant, and nothing you paste is stored.
Detection happens in two steps. First the writing system: text in Japanese kana, Korean Hangul, Devanagari, Arabic script, Cyrillic, Thai or Greek narrows the possibilities immediately, because only a handful of languages use each. That step alone is close to perfect.
Then, within those candidates, the text is compared against statistical fingerprints of each language — the most common three-letter sequences that appear in it. Every language has a distinctive pattern here, which is why -ción points to Spanish and -ung to German.
Short input is genuinely ambiguous, and no detector handles it well. “Hola” is Spanish, but it is also a word in several other languages, and four letters carry almost no statistical signal. One or two full sentences give a much more reliable answer, which is why the tool tells you when your text is on the short side rather than pretending to be certain.
Closely related languages are the other hard case. Spanish and Portuguese, Norwegian and Danish, Indonesian and Malay share a great deal — when the result looks uncertain, the alternatives listed underneath are worth reading.
Text is sent to our server to be analysed and is not stored, logged or passed to any third party. Detection runs entirely on our own statistical model — nothing is sent to an external service.