What input languages does Turnitin's AI detection cover? Seven, added from April 2023 to 2026, with paraphrase detection in English only
Seven, by Turnitin's AI writing detection FAQ read 9 November 2026: "long-form English, long-form Spanish, long-form Japanese, long-form Modern Arabic, long-form Dutch, long-form German, and long-form Turkish", each with its own model (guides.turnitin.com, read 8 November 2026 for the card and re-read 9 November 2026). Three of the seven carry a launch date on Turnitin's own pages: English on 4 April 2023, in the CEO's launch post (turnitin.com, read 9 November 2026), Spanish on 12 September 2024 and Japanese on 2 April 2025, both in the model release notes (guides.turnitin.com, read 9 November 2026). Arabic, Dutch, German and Turkish appear in the FAQ and in their own localised FAQ pages with no release note, and the same FAQ states that "AI paraphrasing and bypassing detection capabilities are only available for English submissions".
The question matters because the Turnitin card records one language count, 7, while the study record behind it is English-only, and because the vendor's sibling product covers fewer languages than the indicator itself. Rows list each language with the date Turnitin's pages attach to it, the generators the vendor names for that model where it names any, and the page the date comes from, read on the date shown. This page is one of the questions answered from the index; the languages ranking orders every card by stated language count, the non-English evidence ranking counts the records that tested each card outside English, the Spanish detectors question covers the one language most of them share, and the AI detector register carries every card.
| Language | Date added, as Turnitin's pages state it | Generators Turnitin names for the model | Source | Read date |
|---|---|---|---|---|
| English | 4 April 2023 (launch post, Chris Caren); model updates 24 May 2023, 6 December 2023, 16 July 2024, 27 August 2025, 14 October 2025, 12 February 2026, 14 October 2026 | 33 named models from GPT-4o and GPT-5.x to Claude Opus 4.6, Gemini 3.1 Pro, DeepSeek V3.2, Grok 4.1, Llama 4 and Mistral Large 3, per the FAQ | turnitin.com; guides.turnitin.com | 9 November 2026 |
| Spanish | 12 September 2024 (release note); Spanish model updated 5 May 2026 | GPT-3.5 and GPT-4 | guides.turnitin.com | 9 November 2026 |
| Japanese | 2 April 2025 (release note) | GPT-4 (version 0613), GPT-4o, GPT-4o-mini | guides.turnitin.com | 9 November 2026 |
| Modern Arabic | Listed in the FAQ; no release note on the model page; Arabic FAQ page last edited 18 August 2026 | Arabic FAQ lists GPT-5.2 Pro, GPT-5.3 Chat, GPT-5.4, Claude Opus 4.5 and 4.6, Claude Sonnet 4.6, Gemini 2.5 Pro, Gemini 3 Flash, Gemini 3.1 Pro, DeepSeek R1, V3.1 and 3.2, Grok 4.1 Fast | guides.turnitin.com (Arabic) | 9 November 2026 |
| Dutch | Listed in the FAQ; no release note on the model page; Dutch FAQ page last edited 8 October 2026 | Dutch FAQ lists Claude Opus 4.6 and 4.8, Claude Sonnet 4.6, DeepSeek R1 and V3.2, GPT-5.2 Pro to GPT-5.5, Gemini 2.5 Pro to 3.5 Flash, Grok 4.1 Fast, Llama 4 Maverick | guides.turnitin.com (Dutch) | 9 November 2026 |
| German | Listed in the FAQ; no release note on the model page; German FAQ page last edited 7 October 2026 | German FAQ lists the same set as the Dutch page, with Grok 4.1 Fast marked n/a | guides.turnitin.com (German) | 9 November 2026 |
| Turkish | Listed in the FAQ; no release note on the model page; Turkish FAQ page last edited 6 October 2026 | Turkish FAQ lists the same set as the Dutch page | guides.turnitin.com (Turkish) | 9 November 2026 |
Dates and model lists as Turnitin's pages state them, read 9 November 2026; the "last edited" dates for the four localised FAQ pages are the edit timestamps the help centre reports for those articles, which bound when each page existed and not when the model launched.
When did Turnitin add each language?
English first, then one language a year for two years, then four at once without a dated note. The English indicator launched on 4 April 2023 in Turnitin Feedback Studio, Turnitin Originality, Turnitin Similarity, Simcheck and Originality Check, per the launch post by Turnitin's CEO (turnitin.com, read 9 November 2026). Spanish followed on 12 September 2024, "a different model than our [English] model ... trained with Spanish language writing and to detect text likely generated with the GPT-3.5 and GPT-4 large language models", and received an update "while minimizing false positives" on 5 May 2026 (guides.turnitin.com, read 9 November 2026). Japanese followed on 2 April 2025, trained to detect "GPT-4 (version 0613), GPT4-o ... and GPT-4o-mini", with a note that "our Japanese and English AI detection capabilities differ" (same page). Modern Arabic, Dutch, German and Turkish are named in the English FAQ's list of supported languages and each has a localised FAQ page whose model list runs to 2026 releases, but the model release notes page, which lists every English, Spanish and Japanese change from 24 May 2023 to 14 October 2026, carries no entry for any of the four (guides.turnitin.com, read 9 November 2026). The Turnitin claims record, written from the FAQ read 8 November 2026, records the same seven models and the English update of October 2026.
Does Turnitin detect paraphrased AI text in every language?
No, in English only. The FAQ states: "our AI paraphrasing and bypassing detection capabilities are only available for English submissions" (guides.turnitin.com, read 9 November 2026). Those two capabilities were added to the English model on 6 December 2023, for text "paraphrased using an AI word spinner", and on 27 August 2025, for "the likely use of AI bypasser tools" (guides.turnitin.com, read 9 November 2026), and the paraphrased text question lists Turnitin among the 16 cards with such a statement on that basis. A Spanish, Japanese, Arabic, Dutch, German or Turkish submission gets the AI-generated score for its language and no paraphrase or bypasser category. The same FAQ sets the floor for processing "depending on the language" at 300 words in evaluated sentences, and says a submission in a language outside the seven "will not process", showing "an empty/error state" instead of a score.
Has any study tested Turnitin outside English?
None in the seven languages, and two at the edge of English. Turnitin has 5 records in the index, and the non-English evidence ranking counts 2 of them: Weber-Wulff 2023, whose class 02-MT held 9 human documents written in Bosnian, Czech, German, Latvian, Slovak, Spanish or Swedish and machine-translated into English, on which Turnitin made 0 false accusations (DOI 10.1007/s40979-023-00146-z, read 8 November 2026), and Van Vlasselaer 2026, whose 40 human papers were written in English by non-native graduate students and all classified correctly (DOI 10.1007/s40979-026-00226-w, read 9 November 2026). Neither is a test of the Spanish or Japanese model. Orenstrakh 2023 did test 10 human and 10 ChatGPT submissions in Spanish, but its roster held CopyLeaks, GPTZero, GLTR, the GPT-2 Output Detector, CheckForAI and the OpenAI classifier, not Turnitin (Orenstrakh 2023 record). Turnitin's own published evidence for the non-English models is the claim the card quotes for every language: a document false positive rate under 1% for documents with over 20% AI writing, validated on over 700,000 pre-ChatGPT papers before each release, with no per-language figure and no true positive rate (guides.turnitin.com, read 9 November 2026). The ELL figures the card carries, a 0.014 false positive rate for ELL writers against 0.013 for native English writers, are for English-language writing by second-language learners, not for another language.
The two Turnitin products and their language scope
| Card | AI detection languages, as the card states them | Paraphrase and bypasser detection | Source | Read date |
|---|---|---|---|---|
| Turnitin (AI writing indicator in the Similarity Report) | 7: English, Spanish, Japanese, Modern Arabic, Dutch, German, Turkish | English only | guides.turnitin.com | 8 November 2026 |
| Turnitin Clarity (Writing Report, with the Originality add-on) | 3 inside Clarity: English, Japanese, Spanish; students may write in any language and the process data is language-agnostic | Clarity's FAQs say only that the AI assistant is "trained and tested in English only" | guides.turnitin.com | 9 November 2026 |
Scope as each card records it from the page linked, read on the date shown.
A language on Turnitin's supported list has a model and a release date at best, and a per-language accuracy figure from Turnitin or from any study in this index at no point.
Questions about Turnitin's language coverage
How many languages does Turnitin's AI detector support?
7 at 9 November 2026: long-form English, Spanish, Japanese, Modern Arabic, Dutch, German and Turkish, each with its own model, per Turnitin's FAQ (guides.turnitin.com). A submission in any other language is not processed.
When did Turnitin add Spanish AI detection?
12 September 2024, per the model release notes, with a model "trained with Spanish language writing" to detect GPT-3.5 and GPT-4 output, and an update on 5 May 2026 (guides.turnitin.com, read 9 November 2026). Japanese followed on 2 April 2025.
Does Turnitin detect paraphrased AI text in Spanish?
1 language, English: "AI paraphrasing and bypassing detection capabilities are only available for English submissions", per the FAQ read 9 November 2026 (guides.turnitin.com). The English capability dates from 6 December 2023 for paraphrase and 27 August 2025 for bypasser tools.
Has Turnitin's Spanish or Japanese model been tested by a study?
0 of Turnitin's 5 records in the index tested either model, read 9 November 2026; the 2 records counted for non-English evidence used machine-translated English (Weber-Wulff 2023, 9 documents, 0 false accusations) and English by non-native writers (Van Vlasselaer 2026, 40 papers, all correct) (non-English evidence ranking).
Does Turnitin Clarity cover the same languages?
3 of the 7: AI writing detection inside Clarity, when the Originality add-on is licensed, covers English, Japanese and Spanish, while students may write in any language and the Writing Report's process data is language-agnostic (guides.turnitin.com, read 9 November 2026).
Sources
- Turnitin, AI writing detection capabilities FAQs (English): https://guides.turnitin.com/hc/en-us/articles/28477544839821-Turnitin-s-AI-writing-detection-capabilities-FAQs (read 8 November 2026 for the card, re-read 9 November 2026)
- Turnitin, AI writing detection model release notes: https://guides.turnitin.com/hc/en-us/articles/28294949544717-AI-writing-detection-model (read 9 November 2026)
- Turnitin, localised AI writing detection FAQs: Arabic https://guides.turnitin.com/hc/ar/articles/28477544839821, Dutch https://guides.turnitin.com/hc/nl/articles/28477544839821, German https://guides.turnitin.com/hc/de/articles/28477544839821, Turkish https://guides.turnitin.com/hc/tr/articles/28477544839821 (read 9 November 2026, with the help centre's edit timestamps for each article)
- Caren, C. (4 April 2023). The launch of Turnitin's AI writing detector and the road ahead: https://www.turnitin.com/blog/the-launch-of-turnitins-ai-writing-detector-and-the-road-ahead (read 9 November 2026)
- Turnitin Clarity FAQs for instructors: https://guides.turnitin.com/hc/en-us/articles/37198804513165-FAQs-for-Instructors-using-Turnitin-Clarity (read 9 November 2026)
- Weber-Wulff et al. 2023: https://doi.org/10.1007/s40979-023-00146-z (read 8 November 2026)
- Van Vlasselaer 2026: https://doi.org/10.1007/s40979-026-00226-w (read 9 November 2026)
Revision log: 2026-11-09: page created from the Turnitin and Turnitin Clarity cards, the Turnitin claims record and Turnitin's FAQ, release notes and launch post read 9 November 2026, at index version 2026.11.8.
Related pages: Which AI detectors let you delete your submitted text? The 17 that state a deletion route, the 10 that say nothing is kept, and the 15 that state neither, Which AI detectors offer sentence-level highlighting instead of a single score? The 17 cards whose pages state it, and what each highlights, AI detectors ranked by study records on non-English and non-native text and Questions about AI detectors, answered from the index.