Written and maintained by CASRAI Editorial Board
Published · Last updated
On this page
Compilatio is a French similarity-detection and AI-writing-detection service sold mainly to universities, schools and training organisations. This guide is for authors, supervisors and integrity staff who need to know what a Compilatio report is based on, how its plans differ, and how it differs in design from iThenticate, the tool most publishers use. Everything about the product below is what Compilatio itself states on its public pages (read 2026-11-08); accuracy figures are the vendor’s own and are labelled as such.
Ask CASRAI · free to tryQuestion about this guide? Ask CASRAI answers with linked sources. 2 free questions a day, no account.
What Compilatio is
Compilatio’s English site lists three products: Magister for educators (similarity and AI detection), Studium for students who want to self-check a draft, and Gingo for creating and grading assessments with AI help. It states support for more than 40 languages, monolingual and cross-lingual analysis, and integration through learning-management systems, single sign-on and an API. The vendor describes its users as universities, schools, training organisations, ministries and research centres, and claims 2,000+ institutions, 2 million+ users and 60+ countries (vendor claims, not independently confirmed). Its public product pages show no product aimed at journal publishers; the publisher link points to a separate site.
Magister, Magister+ and Studium compared
| Capability | Magister | Magister+ | Studium |
|---|---|---|---|
| Intended user | Educators | Educators | Students |
| Exact and lightly reworded text (syntactic rewording) | Yes | Yes | Similarity analysis included with a free account (initial analysis) |
| Deep (semantic) paraphrase detection | No | Yes | Not stated |
| Translated-text (multilingual) similarity | No | Yes | Not stated |
| AI-generated content detection | No | Yes | Not stated |
| Comparison with other institutions’ student papers | Yes (Compilatio network) | Yes (Magister base) | Never |
Source: Compilatio’s Magister+ comparison page and similarities-detection page, read 2026-11-08. Where the vendor pages did not state a capability, the cell says so rather than assuming it.
What Compilatio compares against
Per the vendor, the tool extracts raw text from PDF, Word and plain-text files and strips metadata, images and formatting before comparison. It reports four kinds of similarity: verbatim or lightly modified text, translated text, syntactic rewording (synonym swaps, reordered clauses) and semantic paraphrase. The comparison sources it names are:
- The web, described as hundreds of billions of pages, including archives of sites that no longer exist.
- Scientific publications, described as over one billion academic documents, including open-access material and repositories.
- Student papers, only for Magister users, drawn from other institutions in its network, and never for Studium.
Compilatio says it filters insignificant matches such as set phrases, legal terms and short quotations. It does not publish how the percentage is calculated, so two tools’ percentages for the same file are not directly comparable.
Reading the score
Compilatio’s own guidance reads below 10% as generally normal, 10 to 25% as warranting verification, and above 25% as needing careful review, and adds that the number alone never tells the whole story. Treat these as the vendor’s rule of thumb, not a field standard: they are not set by Crossref, COPE or any funder. A match can be a properly quoted passage, a reference list or the author’s own earlier work, so the useful step is always reading the highlighted sources. The originality report entry explains what such a report does and does not show, and plagiarism defines the underlying conduct that a similarity score can only suggest.
AI detection and what it cannot prove
AI detection is a Magister+ feature, and flags appear in the same report as similarity matches. Compilatio states that the result alone does not prove use of generative AI, and that no AI detector is fully reliable and a result should be interpreted rather than treated as absolute proof. It also reports reliability of between 94% and 99% and under 1% false positives, from tests it conducted itself. Those figures are vendor-reported and this guide has not seen independent testing, so do not use them as evidence in a misconduct decision. An AI flag is a prompt for a conversation and a review of drafts, not a finding.
Compilatio vs iThenticate
| Dimension | Compilatio | iThenticate |
|---|---|---|
| Designed for | Teaching and assessment: Magister (educators), Studium (students) | Publishers and research-output screening, offered to publishers through Crossref Similarity Check |
| Access route | Institution licence, LMS integration, SSO, API; free student account for Studium’s initial analysis | Crossref members actively publishing DOI-assigned content, at reduced rates |
| Scholarly corpus | Vendor states over one billion academic documents | Crossref states over 78 million full-text scholarly items, which members make available through full-text URLs in their metadata |
| Student-paper comparison | Magister only, within its network | Not described on Crossref’s page |
| AI detection | Magister+ (vendor-described) | Not described on Crossref’s page |
Sources: Compilatio pages above; Crossref Similarity Check page, read 2026-11-08. The corpus figures are each owner’s own descriptions and count different things (documents versus full-text scholarly items), so they should not be ranked against each other. For the publisher side, see the iThenticate pricing guide, the Crossref Similarity Check vs Turnitin comparison and iThenticate vs Turnitin; for the classroom side, the Turnitin guide. For how Turnitin, iThenticate and Crossref Similarity Check relate, see anti-plagiarism software explained.
FAQ
Is Compilatio the same as iThenticate?
No. They are separate products from different vendors. Compilatio is built around educators and students; iThenticate reaches publishers through Crossref Similarity Check for members who publish DOI-assigned content.
Does Compilatio detect AI-written text?
Per the vendor, AI detection is part of Magister+, not the base Magister plan, and it states the result does not prove generative-AI use by itself.
Does Studium store my paper in a database others are checked against?
Compilatio states that student papers are used as comparison sources only with Magister, and never for Studium.
What similarity percentage is acceptable?
Compilatio’s guide treats under 10% as generally normal and over 25% as needing careful review, but it stresses that the number is not decisive. Journals and institutions set their own policies.
Sources
- Compilatio, English home page, https://www.compilatio.net/en (read 2026-11-08).
- Compilatio, similarities detection information, https://www.compilatio.net/en/similarities-detection-info (read 2026-11-08).
- Compilatio, Magister+ comparison, https://www.compilatio.net/en/magister-plus (read 2026-11-08).
- Crossref, Similarity Check, https://www.crossref.org/services/similarity-check/ (read 2026-11-08).







