Anti-plagiarism software — Turnitin, iThenticate, and Crossref Similarity Check are the three names used across research administration and scholarly publishing — is text-matching software that compares a submitted document against a reference corpus and flags overlapping passages. Despite the name, none of these tools determines whether plagiarism occurred; a human reviewer still interprets what the similarity means.
What This Kind of Software Actually Does
These tools work by text-matching, not by judging originality of ideas. The system segments a submitted document into phrases, discards common stop words, and checks each phrase against a reference database — typically current and archived web content, previously submitted student papers, and licensed periodicals and journals. For the publishing-industry tools, the corpus also includes content contributed by participating publishers. The output is a similarity index: the percentage of the submitted text with a detectable match somewhere in that corpus. For a full breakdown of how that percentage is calculated and formatted into a report, see originality report.
Turnitin vs. iThenticate vs. Crossref Similarity Check
These three names are not competing, unrelated products — they are the same detection engine, offered through three different channels to three different markets.
- Turnitin is the education-market product: universities license it primarily to screen student coursework and, in many graduate schools, electronic thesis/dissertation (ETD) deposits, usually through an integration with the institution’s learning management system or repository software.
- iThenticate is built by the same company (Turnitin owns iThenticate) but marketed specifically to the research and scholarly-publishing side: individual publishers, journal editorial offices, and research institutions checking manuscripts, grant applications, or pre-submission drafts, independent of any classroom LMS.
- Crossref Similarity Check is a service Crossref operates on top of the iThenticate engine, giving Crossref member publishers and journals integrated similarity-checking access — often built directly into their editorial-management system — without licensing iThenticate separately. Its reference corpus additionally draws on published content contributed by Crossref member publishers.
In short: if the context is a university checking student or thesis work, it is almost always Turnitin. If the context is a journal or publisher screening manuscripts, it is almost always iThenticate or, more specifically, Crossref Similarity Check.
A Similarity Score Is Not a Plagiarism Verdict
This is the single most important operational fact about this category of software, and it comes from the vendors themselves, not just from critics of the tools. Turnitin’s own product guidance states plainly that the Similarity tool “does not determine whether plagiarism has occurred” — it is built to support a reviewer’s own judgment, not replace it. A high similarity index can come from properly quoted and cited material, a shared bibliography or reference list, standard methodological language common to a field, or a manuscript matching the author’s own previously published work (see self-plagiarism) — none of which is plagiarism in the research-integrity sense. Conversely, a low score does not guarantee a clean manuscript: the tools can only match against text already in their corpus, so translated passages, heavily reworded (but still uncredited) borrowing, and copied ideas expressed in new words can all slip through undetected. Interpreting a report is a matching-and-review exercise, not a single pass/fail number.
Where This Fits in the Editorial and Academic Workflow
In journal and publisher settings, a similarity check is typically run as a screening step early in the submission process — before or during initial editorial review — and the resulting report is one input the handling editor considers alongside the manuscript itself, not an automatic accept or reject rule. Editorial offices generally follow the same underlying guidance from the Committee on Publication Ethics (COPE) that governs how any research-integrity concern raised during review should be investigated, rather than acting on a similarity score in isolation.
In university settings, the equivalent step is usually built into thesis and dissertation submission (an originality report reviewed by the supervisor or graduate office alongside the candidate’s own declaration of originality) or, separately, into coursework submission through the institution’s LMS. When a similarity report raises a substantive concern rather than a routine false positive, it typically escalates into the institution’s or journal’s research misconduct process rather than being resolved by the software itself.
What Anti-Plagiarism Software Cannot Do
- It cannot judge intent. The software returns a percentage; deciding whether an author acted in bad faith is a human editorial or institutional-integrity function.
- It is limited to its reference corpus. Content that has never been indexed — paywalled sources the vendor doesn’t hold a license for, non-digitized material, or text in a language the corpus doesn’t cover well — can be copied without triggering a match.
- It does not evaluate data, images, or figures. These tools check text strings; they are not built to detect duplicated or manipulated images, data fabrication, or plagiarized code or datasets.
- AI-writing detection is a separate feature, not the same thing. Some vendors, including Turnitin, now offer a distinct AI-generated-text detection capability alongside their traditional similarity checking. It is marketed and evaluated separately from the similarity index, works on different signals, and should not be assumed to carry the same reliability characteristics discussed above — treat it as a distinct tool with its own evidentiary limits, not an extension of plagiarism detection.
Choosing Between These Tools
For most institutions, the practical choice isn’t really a choice: universities screening coursework and student theses use Turnitin because that is the product built for, and licensed into, the education market. Publishers and journals screening manuscripts use iThenticate or Crossref Similarity Check because those are the products built for scholarly publishing — and if the journal or publisher is already a Crossref member, Similarity Check is usually the more integrated and cost-effective route into the same underlying engine rather than licensing iThenticate directly. A number of other standalone text-matching and plagiarism-checking products exist on the market outside this ecosystem, but Turnitin, iThenticate, and Crossref Similarity Check are the tools most deeply embedded in university and scholarly-publishing workflows specifically.
Frequently Asked Questions
Is anti-plagiarism software the same thing as an AI detector?
No. Text-similarity checking and AI-generated-text detection are separate features built on different underlying methods, even when the same vendor (such as Turnitin) offers both. A document can score low on similarity and still be flagged by an AI detector, or vice versa.
What similarity percentage counts as plagiarism?
There is no universal numeric threshold, and vendors themselves warn against treating one as a verdict. A short, well-cited quotation and a shared reference list can both push a score up without indicating misconduct, while a low score doesn’t rule it out. Journals and institutions that publish their own internal thresholds treat them as a trigger for closer human review, not an automatic decision.
Is Turnitin the same company as iThenticate?
Yes. Turnitin owns iThenticate; both run on the same core detection engine but are packaged and sold to different markets — Turnitin to education, iThenticate to research and publishing.
Do all academic journals check manuscripts for plagiarism?
It is common, especially among established and Crossref-member publishers, but not universal. Practice varies by publisher, journal, and field, and even where a check is run, editorial offices treat it as one screening input rather than a guaranteed step in every submission’s evaluation.
Can these tools produce false positives or false negatives?
Yes, both. False positives commonly come from properly cited quotations, shared boilerplate or methodological language, and an author’s own previously published text. False negatives occur when matching source material simply isn’t in the vendor’s reference corpus, or when text has been substantially reworded rather than copied verbatim.
