Free plagiarism checker — no signup
Check up to 2,500 words for plagiarism against 150M+ web pages and academic papers. No signup required. Open-source engine — audit the code yourself.
Five corpora. One free check.
Every check on the free tier scans against all five — toggle individually on paid tiers if you want to scope.
Google + Brave crawl, scraped fresh per check. Catches what indexed search would find.
Every paper with a DOI, including paywalled abstracts. Updated weekly from the OpenAlex bulk dump.
Snapshot of the open web. Includes older content the live SERP no longer surfaces.
Open-access academic full-text where the publisher allows it. Indexed monthly from the CORE bulk export.
Citations + metadata for everything published with a DOI. Continuous via the Crossref API.
Every matched passage shows its source.
Source URL, similarity score, and the exact overlapping text — for every match. Export PDF, share a read-only link, or save to a folder.
The detection engine takes a submitted document and decomposes it into 4-sentence overlapping windows. Each window gets hashed into 64-bit fingerprints using winnowing — Schleimer's local-minimum k-gram selection from the 2003 SIGMOD paper. The fingerprints query a GIN-indexed Postgres column to find candidate matches in roughly logarithmic time. From there, the cascade narrows: cardinality scoring trims candidates to top-K, then seed-and-extend alignment produces the actual passages shown in this report.
An engine that can be read.
Every plagiarism checker before us has been a black box. A score comes back, and you don't know how. We're the first that isn't. The engine — winnowing fingerprints, alignment, and report assembly — is Apache 2.0 on GitHub, with MinHash LSH and vector reranking landing there as they ship. Read the algorithms. Fork the repo. Run your own copy.
This matters because checking is consequential. When the score gets a student suspended or a paper retracted, the people involved should be able to audit how the score was generated. We made that possible. The code that powers noplag.com and the code you can self-host are the same code; the diff between hosted and self-hosted is purely operational, never algorithmic.
What we can't catch — and how we tell you.
Heavy paraphrasing — text rewritten beyond roughly 60% semantic distance from the source — escapes detection on every tool, including this one. We're explicit about it. Common-knowledge phrases get over-flagged sometimes; you'll need to review the highlighted passages yourself and judge what's actually plagiarism. We surface a confidence interval on every match so you know which to trust and which to verify.
Ten questions, ten honest answers.
More questions than answered here? Email hello@noplag.com — we add to this list when patterns repeat.
- Is Noplag's free plagiarism checker really free?
- Yes. Up to 2,500 words per check, no signup, no credit card. The Pro tier ($23/mo) raises the limit and adds the Noplag Database + API access; the Premium tier ($79/mo) adds Live Web search + branded reports.
- What sources does the free check scan?
- The web (via Google + Brave SERPs), Wikipedia, OpenAlex (250M academic works), arXiv, PubMed Central, Common Crawl, and our growing Noplag Database (anonymized fingerprints from prior checks).
- Will the free checker find AI-generated text?
- It will detect verbatim and near-verbatim matches to AI output already published online. Standalone AI-generated-text detection is coming soon as a Premium-tier add-on (see /ai-content-detector).
- Do you keep my submitted text?
- Free tier: documents purged 30 days after check completion. Paid plans: configurable 30–90 days, unlimited on Enterprise. We never use your text to train models. EU residency available.
- How accurate is plagiarism detection?
- The engine ships with a published evaluation set; current results are at /how-it-works. Like every checker, we under-detect heavy paraphrasing and over-flag common-knowledge phrases. We're explicit about this — see the honest-limitations note above.
- Can I integrate this into my own product?
- Yes — the Pro tier exposes a REST API. The engine is also Apache 2.0 open-source — self-host the whole thing if you prefer.
- How is Noplag different from Turnitin?
- We're open-source (you can audit the code), browser-first (no LMS required), and built for individuals + small teams as the wedge. See /vs-turnitin for an honest comparison.
- Is there a word limit?
- 2,500 words per check on the free tier. Pro: 10,000 per check. Premium: 50,000 per check. Enterprise: unlimited.
- Do you support languages other than English?
- Yes — English, Spanish, Portuguese, Polish, and Ukrainian at launch. Multilingual coverage is a Premium-tier capability gate; see /pricing.
- What's the Noplag Database?
- A shared fingerprint store that helps detect copy-paste across our user base. By default, every check contributes anonymized fingerprints — not full text — to this index across all tiers; you can opt out and turn it off anytime. See /privacy.
Check 2,500 words — free, no signup.
The free tier runs forever. Paste the text, get the report. Open-source engine — audit how the score was made.
Check for plagiarism