Compilatio Magister+ AI detector
Vendor Compilatio SAS Official site compilatio.net/en/magister How this card is sourced
Updated · AI Detector Index editors
Compilatio's AI detector classified 40 of 54 documents correctly, 74% binary accuracy and second of 14 tools, with 0 false accusations in 18 human-written texts in Weber-Wulff 2023 (arXiv 2306.15666, Tables 5 and 9, read 9 November 2026). The vendor is Compilatio SAS of Chavanod, France (legal notices, read 9 November 2026), which sells the detector to institutions inside Compilatio Magister+ and to students inside Compilatio Studium; the free public checker the 2023 study used no longer exists. The card joins the detector index beside the other tools sold to institutions, and how this card is sourced explains why the vendor's own 7,400-text test stays a vendor figure.
What Compilatio detects
| Field | Value | Source |
|---|---|---|
| Content types detected | Text, handwriting | compilatio.net/en/magister |
| Generators covered | ChatGPT, Gemini, Claude, Other generative AI systems (vendor wording); the v4.6.2 model was trained against 2026 versions of ChatGPT, Claude and Gemini | compilatio.net/en/whats-new/introduction-... |
| Languages | French, English, Spanish, Italian, Portuguese, Russian, Arabic, Hindi (35 listed) | compilatio.net/en/ai-detector-info |
| Input limits | Documents are uploaded to Magister+ (teachers, via the web interface, LMS or API) or Studium (students); the detector works passage by passage and returns a score for the estimated proportion of AI text with the passages highlighted. Studium meters analysis in credits, 1 credit = 250 words analyzed, valid across documents for 6 months. No document size limit is published; the 2023 free tool took 200 to 2,000 characters (Weber-Wulff 2023 Table 2). | compilatio.net/en/product/studium |
| Free tier | None for the detector. Magister+ is sold to institutions by demo and quote; Studium's first free partial analysis shows the top 3 plagiarism sources in a document, and a full analysis needs a credit pack | compilatio.net/en/product/studium |
| API | Yes, compilatio.net/en/plagiarism-detection-api | compilatio.net/en/plagiarism-detection-api |
| Integrations | Moodle, Canvas, Brightspace (D2L), Blackboard (Learn and Learn Ultra), Microsoft Teams, Open LMS, Cineca, LTI custom integration, SSO, REST API (plagiarism, AI and altered text detection) | compilatio.net/en/lms-integration |
| Data retention | Privacy policy (support centre, Compilatio SAS as controller for Studium, Magister and Magister+): Account data includes documents you have submitted. Delivery of the service of Detection and Measurement of Similarities in texts: Identification / Account / Technical, duration of storage 5 years from the end of the contractual relationship. Your data is kept and stored for the duration of the processing on SAS COMPILATIO's servers, located in France at 47, rue Cassiopee, Chavanod (74650). GDPR page: Submitted documents are not used for commercial purposes or reused outside the terms of the agreement with your institution. | support.compilatio.net/hc/en-us/articles/... |
| Ownership | COMPILATIO, a French simplified joint-stock company (SAS) with share capital of EUR 234,400, registered 512 216 847 in Annecy, 47 rue Cassiopee, 74650 Chavanod, France; president Frederic Agnes; the general terms may name COMPILATIO SRL of Turin, Italy as contracting party (legal notices). The AI detector is sold inside Compilatio Magister+ for institutions and Compilatio Studium for students; the public free checker at ai-detector.compilatio.net that Weber-Wulff 2023 tested now redirects to an information page. The vendor is a disclosed conference sponsor in Weber-Wulff 2023. | compilatio.net/en/legal-notices |
| Independent studies | 1 record | Studies index |
The Magister page lists "Detection of content generated by generative AIs such as ChatGPT, Gemini, Claude" as a Magister+ feature that the Standard tier lacks, beside semantic and cross-language plagiarism detection, altered text detection and OCR on handwritten documents (compilatio.net/en/magister, read 9 November 2026). The detector returns "a score indicating the estimated proportion of text that may have been generated by AI and highlights the relevant passages in the report" (AI detector page, read 9 November 2026). The current model, version 4.6.2, has been live for Magister+ and Studium users since October 1, 2026 and was trained on "authentic human texts, compared with texts generated by the 2026 versions of ChatGPT, Claude, and Gemini in several languages" (what's new, 5 November 2026, read 9 November 2026).
Languages are stated as a count: "the AI checker is multilingual. It can detect AI in more than 35 languages" (AI detector page, read 9 November 2026). The support article names eight, French, English, Spanish, Italian, Portuguese, Russian, Arabic and Hindi, and lists the 24 languages its reliability test covered (support article 17408769118993, read 9 November 2026). The Magister page's stat strip says "+40 Supported languages" for the product as a whole. The card records 35 with eight named, and the 40 is filed as a contradiction.
Input is by document upload: teachers submit through the Magister web interface, an LMS or the API, students through Studium, and the detector "typically analyzes texts passage by passage" (what's new, read 9 November 2026). Studium meters analysis in credits, "1 credit = 250 words analyzed", valid across documents for 6 months (compilatio.net/en/product/studium, read 9 November 2026). No document size limit is published; the free tool Weber-Wulff tested in 2023 took 200 to 2,000 characters (arXiv 2306.15666, Table 2). Integrations are Moodle, Canvas, Brightspace (D2L), Blackboard Learn and Learn Ultra, Microsoft Teams, Open LMS and Cineca over LTI, SSO, and a REST API that returns "a probability score that can be used directly in your interface" (LMS page and API page, read 9 November 2026). The nearest cards are Turnitin, which ranked first on the same 54 documents, and ZeroGPT, which also recorded 0 false accusations in that study.
Limits and plans
| Plan | Price | Limits | Checked on |
|---|---|---|---|
| Compilatio Magister+ (institutions) | AI detection, semantic and cross-language plagiarism detection, altered text detection, grammar and spelling checker, priority processing; price by demo request. Magister Standard excludes AI detection. | ||
| Studium 20 credits | EUR 4.99 | 5,000 words analyzed; plagiarism and AI detection for students. | |
| Studium 80 credits | EUR 12.99 | 20,000 words analyzed (marked Best seller). | |
| Studium 160 credits | EUR 19.99 | 40,000 words analyzed. | |
| Studium 400 credits | EUR 34.99 | 100,000 words analyzed. |
There is no free detector. Magister+ is sold to institutions by demo request with no published price, and the AI detector FAQ says "the price of the AI detector varies depending on the solution you choose" (AI detector page, read 9 November 2026). Studium's "first free partial analysis shows you the top 3 sources detected in your document", a plagiarism preview, and a full analysis needs a credit pack (compilatio.net/en/product/studium, read 9 November 2026). Figures below are as displayed in euros, read 9 November 2026.
Retention is stated in years. The privacy policy names Compilatio SAS as controller for Studium, Magister and Magister+, counts "documents you have submitted" among Account data, and sets the storage period for "Delivery of the service of Detection and Measurement of Similarities in texts" at "5 years from the end of the contractual relationship" (support article 4416953631505, read 9 November 2026). Data "is kept and stored for the duration of the processing on SAS COMPILATIO's servers, located in France", and the GDPR page adds that "submitted documents are not used for commercial purposes or reused outside the terms of the agreement with your institution" (GDPR page, read 9 November 2026). The card's retention rank field carries 1,825 days from the five-year clause, a stated period the data retention ranking can sort, read 9 November 2026 in the data file.
What the vendor claims
High-Performance AI Detector: Reliability Between 94% and 99%. (Magister page.) The Compilatio detection engine achieves a reliability rate of 94 to 99% on academic content, depending on document length, language, and the nature of the analyzed content. These measurements were conducted on approximately 7,400 texts in 24 languages. The sample consisted of 3,700 texts written by humans and 3,700 texts written by Artificial Intelligence. Performance measurements in this article are valid for the Compilatio AI text detection system version 4.5.3, in use since September 4, 2025. (Support article; the v4.6.2 model has been live since October 1, 2026 with no new figures.)
The vendor states a reliability rate of 94% to 99%: the Magister page reads "Reliability Between 94% and 99%" and "Fewer than 1% false positives" (compilatio.net/en/magister, read 9 November 2026). The support article behind those figures is more specific than most vendor pages on the index. It says the detection engine "achieves a reliability rate of 94 to 99% on academic content, depending on document length, language, and the nature of the analyzed content" and "displays a false positive rate below 1%", that "these measurements were conducted on approximately 7,400 texts in 24 languages", 3,700 human and 3,700 AI, and that the figures "are valid for the Compilatio AI text detection system version 4.5.3, in use since September 4, 2025" (support article 17408769118993, read 9 November 2026). The v4.6.2 announcement of 5 November 2026 claims fewer incorrectly flagged passages but publishes no new figures, so the 7,400-text numbers describe the previous model.
The vendor's caveats are in the same article: "No AI detector can be 100% reliable"; "The reliability rates communicated by AI detectors are not comparable!"; and "it may happen that a human has a style similar to that of an artificial intelligence" (read 9 November 2026). The AI detector page says the score "is not proof, but rather an indicator that must be interpreted". The vendor also links Weber-Wulff 2023 from its own page as "this third-party study" that "validates the reliability of the Compilatio AI detector" (AI detector page, read 9 November 2026); the study's figures for Compilatio are below, including the 11 AI documents it missed.
What independent studies found
- AIDI-S-0001Weber-Wulff 20232023
Compilatio: 40 of 54 documents correct (74% binary accuracy, rank 2 of 14), 679 of 864 points (79%) on the logarithmic method (rank 2); 0 false accusations in 18 human-written documents; 11 of 36 AI documents undetected (30.6%), with 0 of 9 unedited ChatGPT texts in class 03-AI missed, 3 of 9 manually edited and 7 of 9 QuillBot-paraphrased texts missed. The tool once returned NaN% reliability on a text containing program code. Tested March 7 to 28, 2023 on the free tool (200 to 2,000 characters).
One record in the studies index tested Compilatio. Weber-Wulff 2023 ran 54 documents through 14 detectors between March 7 and 28, 2023, and Compilatio ranked second on every accuracy method: 40 of 54 correct in the binary count (74%), 41.5 of 54 in the semi-binary count (77%) and 679 of 864 points (79%) in the logarithmic count, behind Turnitin on each (arXiv 2306.15666, Tables 5, 6 and 8, read 9 November 2026). It made 0 false accusations in 18 human documents, including the 9 machine-translated ones (Table 9).
The misses were on edited text. Compilatio left 11 of 36 AI documents undetected, 30.6% (Table 10): 0 of 9 unedited ChatGPT texts in class 03-AI, 1 of 9 in class 04-AI, 3 of 9 manually edited texts and 7 of 9 QuillBot-paraphrased texts (arXiv 2306.15666, read 9 November 2026). By class in the binary table it scored 8, 9, 8, 8, 5 and 2 of 9 (Table 5). The authors also note that the tool once returned "NaN% reliability" on a ChatGPT text containing program code, and they disclose that two of them help organise a conference sponsored by Turnitin and Compilatio. The tool tested was the free public checker with a 2,000-character ceiling, not Magister+.
The gap between the vendor's 94% floor (compilatio.net/en/magister, read 9 November 2026) and the 79% logarithmic accuracy in Weber-Wulff 2023 (arXiv 2306.15666, Table 8) is 15 points on the one study available, a 2023 result on a 2023 free tool set against a 2025 measurement of a paid one. The card's lowest figure is the 2 of 9 paraphrased texts caught, 22%, in Table 5.
Where Compilatio appears in the index
Rankings
- AIDI-R-01AI detectors ranked by independent testing Two records added in the fifth wave on 9 November 2026, Pegoraro 2023 and Fleckenstein 2024, changed nine counts and none: Pegoraro 2023 tested ZeroGPT, GPTZero, Copyleaks, Originality.ai, Writer, Content at Scale (BrandWell), DetectGPT, the GPT-2 Output Detector and GLTR, lifting GPTZero to 12, the GPT-2 Output Detector to 10, Copyleaks to 7 and Originality.ai and ZeroGPT to 6; Fleckenstein 2024 tested 289 teachers and no software, so it counts for no card ([Fleckenstein 2024 record](/studies/fleckenstein-2024-teachers/)). Illuminarty, one of four cards added in the same wave, joined with its one image record from Ha and Passananti 2024, so 18 of 72 cards now qualify.
- AIDI-R-02AI detector false positive rates, ranked
- AIDI-R-04AI detectors ranked by languages supported
- AIDI-R-06AI detectors ranked by stated data retention
- AIDI-R-09AI detector false negative rates, ranked
- AIDI-R-10AI detector accuracy claims vs published results, ranked by the gap
- AIDI-R-11AI detectors ranked by named integration surfaces
- AIDI-R-15AI detectors ranked by the LMS platforms the vendor names
- AIDI-R-17AI detectors ranked by the generator models the vendor names
- AIDI-R-18AI detectors ranked by study records on non-English and non-native text
Fits these use cases, questions and comparisons
- AIDI-U-02AI detectors for school districts
- AIDI-U-08AI detectors for researchers checking their own manuscripts
- AIDI-U-09AI detectors for law firms and courts: an index fit check
- AIDI-U-10AI detectors for grant and proposal reviewers
- AIDI-U-12AI detectors for localisation and translation teams
- AIDI-U-13AI detectors for government and public sector buyers
- AIDI-Q-31Which AI detectors were in the Weber-Wulff 2023 study? The 14 tools, their result on 54 documents, and what each one is today
Compilatio appears in six rankings, read 9 November 2026 in the data file. AI detectors ranked by independent studies counts its 1 record; false positive rates sorts on the 0% of 18 human documents; false negative rates on the 30.6% of 36 AI documents; languages on the stated 35; data retention on the 1,825 days; and the claim versus evidence gap on the 15 points between 94% and 79%.
Compilatio also appears on detectors with an API and on LMS integrations with Moodle, Canvas, Brightspace and Blackboard, and the school districts and admissions offices views read the same fields. It is absent from the free tiers ranking, because no free detector exists, and from the robustness ranking, which takes no value from Weber-Wulff 2023's per-class counts. Every detector card on the register shows the same fourteen fields, so a vendor that publishes its sample size and model version sits in the same columns as the vendors that publish a percentage alone.
Since 9 November 2026 Compilatio leads the LMS coverage ranking with seven named platforms, appears on the Canvas without Turnitin question as one of seven non-Turnitin Canvas cards, and has one record on the non-English evidence ranking, the machine-translated class of Weber-Wulff 2023.
Questions about Compilatio Magister+ AI detector
How accurate was Compilatio in Weber-Wulff 2023?
40 of 54 documents correct (74%), rank 2 of 14, with 0 false accusations in 18 human texts and 11 of 36 AI texts undetected (30.6%), per Tables 5, 9 and 10 of the paper.
What accuracy does Compilatio claim?
The vendor states a reliability rate of 94% to 99% and a false positive rate below 1%, measured on about 7,400 texts in 24 languages for model version 4.5.3, per its support article read 9 November 2026.
Is there a free Compilatio AI detector?
No. The detector is sold inside Magister+ by institutional quote and inside Studium by credit pack from EUR 4.99 for 5,000 words; the free Studium analysis shows plagiarism sources only, read 9 November 2026.
How many languages does Compilatio's AI detector cover?
The AI detector page states more than 35 languages and the support article names eight and lists 24 tested; the Magister page says +40 for the product as a whole, read 9 November 2026.
How long does Compilatio keep submitted documents?
Five years from the end of the contractual relationship, on servers in France, per the privacy policy read 9 November 2026; submitted documents are not reused outside the institution's agreement.
Does Compilatio integrate with an LMS?
Yes: Moodle, Canvas, Brightspace (D2L), Blackboard Learn and Learn Ultra, Microsoft Teams, Open LMS and Cineca over LTI, plus SSO and a REST API, per the integration pages read 9 November 2026.
Sources and revisions
Sources
- compilatio.net/en/magisterRead 9 November 2026
- compilatio.net/en/ai-detector-infoRead 9 November 2026
- compilatio.net/en/product/studiumRead 9 November 2026
- compilatio.net/en/lms-integrationRead 9 November 2026
- compilatio.net/en/plagiarism-detection-apiRead 9 November 2026
- compilatio.net/en/whats-new/introduction-...Read 9 November 2026
- compilatio.net/en/gdpr-compliance-securit...Read 9 November 2026
- compilatio.net/en/legal-noticesRead 9 November 2026
- support.compilatio.net/hc/en-us/articles/...Read 9 November 2026
- support.compilatio.net/hc/en-us/articles/...Read 9 November 2026
- arxiv.org/pdf/2306.15666Read 9 November 2026
Revision log
- Card created2026-11-09: lms coverage ranking, canvas question and non-english evidence ranking linked (seventh wave)
). Illuminarty, one of four cards added in the same wave, joined with its one image record from Ha and Passananti 2024, so 18 of 72 cards now qualify.: register plate](/img/independently-tested-480.webp)
