Turnitin AI Detection Languages: English, Spanish, Japanese, Arabic

Turnitin's AI writing detection does not read every language. As of 18 August 2026 it reads four: English, Spanish, Japanese, and Modern Standard Arabic. Arabic is the new one. The entry sits on the Turnitin release notes page, dated 2026 August 18, under the heading "New AI detection writing detection capabilities for Modern Standard Arabic submissions." The doubled word is theirs, not a typo here. The entry itself reads: "Turnitin is excited to release its AI writing detection capabilities in Arabic to help educators uphold academic integrity for Modern Standard Arabic language submissions."

Then comes the thing you will hit the moment you go check for yourself: Turnitin's own pages have not all caught up. We opened eight of them in a browser on 25 August 2026. File requirements for an AI Writing Report, all 695 characters of it, still prints the three-language line, and the string "Arabic" appears zero times on that page while "Spanish", "Japanese" and "English" each appear once. Using the AI Writing Report (6,484 characters) repeats the same three-language line inside its own file requirements section. The AI writing detection model release-notes page (9,233 characters) logged Spanish in September 2024 and Japanese in April 2025, has no Arabic entry at all, and its newest item is dated 2026 May 5. The AI writing detection capabilities FAQ has moved halfway: its opening block now links out to a Spanish, a Japanese and an Arabic FAQ, while its own answer under "What languages are supported?" still names three.

So four is the current number, and the four are not one multilingual model. The release-notes entry says "our Arabic AI writing model is a separate model from our English AI writing, Spanish AI writing, and Japanese AI writing models." Which also means a page that still says three is not wrong about how detection works. It is just older than the Arabic release.

So what happens if you submit in a language that is not on the list? Two Turnitin pages answer that, and they agree. The AI writing detection capabilities FAQ: "If a paper in a non-supported language is submitted, the detector will not process the submission." The Turnitin product updates page, inside its 18 August 2026 Arabic entry: "If a paper in an unsupported language is submitted, our model will not process the paper for AI writing detection."

The wording is worth slowing down on. It is not that the detector runs and returns a low score, or returns something ambiguous. It does not run. The FAQ describes what shows up instead: "The indicator will show an empty/error state with 'in-app' guidance that will tell users that this capability only works for supported languages. No report will be generated if the submitted content is not in one of the supported languages." The product updates page describes the step that comes before any of that: the language of the submission is identified first, and then "the appropriate model will process the submission and provide a report." A language outside the four has nothing to route to.

All of which is a statement about the mechanism's coverage, not a verdict on the paper. The text is not being analyzed and scored low. It is not being analyzed.

The language line is not the only condition in that requirements list. The same two lines also fix the file types: "File must be written in a supported language: English, Spanish, Japanese. Accepted file types: .docx, .pdf, .txt, .rtf." That is the page as it stood on 25 August 2026. The format half is current. The language half is one release behind.

So a submission needs two properties before the AI detector will analyze it: a language it covers, and a file type it accepts. Take the language list from the release notes instead of from this page and the current set is English, Spanish, Japanese, and Modern Standard Arabic. A file with an accepted extension written in a language outside that set does not meet the requirements, and neither does a supported-language paper saved in a format that is not on the list.

When two Turnitin pages disagree, prefer the dated one. Release notes and product updates stamp every entry with a date, so you can see when something changed and whether the page you are reading predates it. The guide pages carry no such stamp on the line you care about. For anything language-related right now, that means reading the 2026 August 18 entry rather than the requirements list, at least until the requirements list is updated.

If the main body of your paper is in Chinese, the answer today is that Turnitin's AI writing detection will not process it, because Chinese is not one of the four. The Arabic release is a good reminder of what that kind of sentence is worth: on 17 August 2026 a Modern Standard Arabic paper was outside the detector's scope, and on 18 August it was inside. The list describes what has shipped, not what is checkable in principle.

There is a second half to this that non-English writers should know about. The Arabic release-notes entry says: "Any previously submitted assignments can be checked with our AI writing detector when they are re-submitted to Turnitin, and you have AI writing enabled for your account." The same provision appeared when Spanish shipped in September 2024 and when Japanese shipped in April 2025. So a paper that was out of scope on the day it was handed in does not stay out of scope permanently. It becomes checkable if someone resubmits it after the language goes live.

Which is why "not processed" should be read narrowly. It describes the detector's coverage on a particular date, and the dates keep moving: Spanish in September 2024, Japanese in April 2025, Arabic in August 2026. It is not a safety statement. It says nothing about what a course, an instructor, or an institution will accept, and nothing about any check that sits outside AI writing detection.

The language limit belongs to AI writing detection, and it does not carry over to everything else Turnitin does to a submission. Similarity checking runs on different machinery, and its language coverage is not remotely the same list.

Translated Matching is the clearest case. It translates the non-English text in a submission into English and compares that against Turnitin's database, and the guide page for it in the enhanced Similarity Report lists 52 entries, Albanian through Vietnamese, with Chinese (Simplified and Traditional) counted as one. Four for AI writing detection, 52 for Translated Matching, same submission. That is why a Chinese-language paper can come back with a similarity result while the AI writing indicator shows nothing at all.

So when you read that "the detector will not process the submission," ask which detector. For AI writing detection on a language outside the four, nothing is processed and no report is generated. That says nothing about similarity, and nothing about what your institution does with the report afterwards. Keeping the two apart is the difference between reading your report and reading things into it.