Is Translating a Text Plagiarism?

Yes. Translating someone else's text and presenting it as your own writing is plagiarism, even though not a single word survives from the original. Translation replaces the words and keeps everything a citation exists to credit: the ideas, the argument, the structure, the evidence, the order in which the points arrive.

The cleanest way to think about it is that a translation is a paraphrase in another language. That is not a metaphor, it is literally how the style guides classify it, and paraphrase without attribution has been plagiarism all along. What changes across the language boundary is not the ethics. It is only the odds of detection, and less than you would think.

The three shapes it takes

The blunt version is whole-cloth translation: take an essay published in Spanish, render it in English, put your name on it. Nobody who does this is confused about whether it is cheating, and no serious tribunal treats "but I translated it" as a defense.

The common version is quieter. A student researching in two languages translates a paragraph from a French source into an English paper and cites nothing, half-believing that the act of translation made the passage theirs. The labor is real, and the labor is irrelevant. Effort spent rewording is not authorship, in any language.

The third shape is laundering: running your source text out to another language and back so the wording comes home rearranged, then submitting the round trip. This one is worth naming precisely because it is a detection trick, not a misunderstanding. The text that comes back still contains none of your ideas, and the intent to disguise is the aggravating factor, not a mitigating one.

Why it feels undetectable, and why that is thinner protection than it looks

Plagiarism checkers are string matchers at heart. They find sequences of words shared between your text and an indexed source, and a translated passage shares no sequences with its original. Cross-language matching is genuinely the hardest case for detection tools, harder than any synonym-swapped rewrite, and pretending otherwise would be selling you something.

But software is not what catches most translation plagiarism. People are. Instructors read the literature of their field in more than one language, and the foreign-language sources a student is most tempted to translate are the prominent ones, which are exactly the ones an examiner knows. Voice does the rest: a translated argument arrives in prose that does not sound like the rest of the paper, organized in a way the student cannot defend in conversation. One question about why the argument runs in that order tends to settle it.

And once suspicion exists, the language barrier works against you, because the match, once found, is undeniable. From there the process is the ordinary one, with the same escalation and the same record as any other plagiarism finding. No institution has a discount tier for plagiarism that crossed a border.

The rule when you translate a source yourself

Now the legitimate case, because there is one, and it is common in any research that reads beyond English.

You may absolutely quote a Spanish interview or a German study in your English paper by translating the passage yourself. APA's guidance on citing your own translations is specific and a little surprising: your translation counts as a paraphrase, not a quotation, so it takes no quotation marks. The reasoning is that it is not possible to codify how exact any given translation is, so quotation marks would promise a fidelity nobody can verify.

What it does take is the citation, exactly as any paraphrase does: author, year, and the location of the passage in the original. Skip that and your careful translation becomes the second shape from the list above. The quotation marks were never the load-bearing part. The citation is.

Citing a translation someone else made

When you read a published translation rather than translating yourself, the translator is part of the record, and APA's format for translated works makes both dates visible. The reference entry looks like this:

Freud, S. (2010). Civilization and its discontents (J. Strachey, Ed. & Trans.). W. W. Norton & Company. (Original work published 1930)

And the in-text citation carries both years, earlier one first:

(Freud, 1930/2010)

This is not bureaucratic fussiness. The double date tells your reader which text you actually held, and it credits the translator, whose choices shaped every sentence you read. Quoting a published translation while citing only the original author erases a real contributor, which is its own small version of the problem this whole article is about.

What about machine translation?

Nothing changes. Whether the rendering was done by you, a paid translator, or a translation tool in a browser tab, the ideas in the output belong to the original author, and the citation obligation rides with the ideas. A tool cannot launder authorship for the same reason it cannot confer it: it never had any to give.

The one genuinely new wrinkle is speed. Machine translation makes the whole-cloth shape effortless, which mostly means more people now try it against examiners who have seen it before. The pattern it produces, fluent paragraphs with a faint wrongness in idiom and a structure imported wholesale from elsewhere, is familiar to anyone who grades.

One boundary note: reusing your own earlier work by translating it into another language is a reuse question rather than a translation question, and the answer belongs to your instructor, asked before you submit.

How to actually check work across languages

If you are verifying a text, your own or someone else's, the practical rule is to scan in the language of the suspected source, because that is where matchable strings live.

A translated-to-English draft will scan clean against the English web no matter how derivative it is, since the matching text exists only in the source language. So when a paper leans on Spanish or German sources, scan the passages you can identify in those languages. PlagiarismCheck.io scans in Spanish, French, German and other languages, free and without an account, so checking a suspect paragraph in its own language takes about a minute. What no checker can honestly promise is catching a competent translation by scanning only the output language, and a tool that claims otherwise is describing a product that does not exist.

For editors commissioning multilingual work, the same logic slots into the acceptance checks you already run: scan each language against its own web, and treat a suspiciously fluent section the way you would treat any other structural red flag.

The unit of authorship is the idea

Every version of this question resolves the same way once you stop treating language as the boundary of ownership. The words were never the thing you owed credit for. What you cite is the thinking, and thinking survives translation perfectly, which is the entire point of translation.

So the working rule fits in one sentence: if the ideas came from a source, cite the source, in whatever language the ideas arrived and whatever language they left in. Get that right and translation goes back to being what it is supposed to be, one of the most useful things a researcher can do.