Essay plagiarism checker.
Check your essay before you submit. Free up to 2,500 words — covers a typical class essay in one paste. Catches the common essay-specific mistakes (uncited paraphrase, missing in-text reference, common-knowledge over-flagging); AI-paragraph detection with the ESL-adjusted band is coming soon. No signup.
Four mistakes the report catches before your teacher does.
Each one classified separately. Properly-cited material excluded from the headline number; the genuinely-problematic stuff stays flagged.
Reworded source, no citation
You re-phrased something from Wikipedia or a textbook in your own words and forgot the in-text reference. Tags as UNCITED-PARA. Counted in the headline score. The fix: add the (Author, year) or footnote — paraphrase WITH citation is legitimate scholarly work.
Verbatim phrase, no quotation marks
Three or more consecutive words copied from a source without quotation marks (even if the citation is there). Tags as QUOTED-NO-MARKS. The fix: add the quotation marks OR paraphrase fully into your own words.
Standard phrasing flagged as plagiarism
“Water boils at 100°C” or “World War II ended in 1945” — facts that appear in 2,000+ sources. Tags as COMMON-KNOWLEDGE. Excluded from the headline score automatically. Many basic checkers over-flag these; we don't.
ChatGPT / Claude / Gemini output (coming soon)
Per-paragraph AI-likelihood with ESL-adjusted band is on the roadmap. Common pattern: you drafted with AI, didn't disclose, didn't rewrite. When it ships it tags as AI-LIKELY (or ESL ADJ if your lexical signature reads ESL). The fix will depend on your school's policy — disclose, rewrite, or leave.
Three sentences from a typical first-year essay. Three different classifications.
Each example shows what the highlighting looks like in the report and which classification fires. The headline similarity score reflects categories 2 and (sometimes) 3 — not category 1.
01 · Cited quote — excluded
“All happy families are alike; each unhappy family is unhappy in its own way” (Tolstoy, 1877, p. 1). Quotation marks ✓ · in-text citation ✓ · author and year present ✓ — tagged QUOTED-CITED. Excluded from the headline similarity score. Properly-cited literary quotation is the canonical use of source material; the engine rewards it.
02 · Uncited paraphrase — counted
Tolstoy argued that every happy family looks the same, but each unhappy family has its own particular misery — a famous opening line. No quotation marks · no citation. Tagged UNCITED-PARA. Counted toward the headline score. The fix: add (Tolstoy, 1877) at the end of the sentence.
03 · Common knowledge — excluded
Anna Karenina was published in serialised form between 1875 and 1877. Matched against 2,400+ encyclopedia and reference sources. Tagged COMMON-KNOWLEDGE. Excluded from the headline score automatically. Standard biographical facts don't need a citation, and the engine doesn't penalise their inclusion.
Six categories. Every matched interval lands in exactly one.
The headline similarity score reflects categories 4 and 5 only. The others are expected scholarly work — properly attributed, common knowledge, or your own.
01 · QUOTED · CITED
Quotation marks present, in-text citation (Author, year) or footnote within the same sentence, author appears in bibliography. Excluded from headline. The canonical use of source material; the engine treats it as legitimate scholarly work.
02 · PARA · CITED
Paraphrase with attribution: “Smith (2023) argues...” or trailing (Smith, 2023). Counted at reduced weight. Most lit reviews are 60-80% this category — paraphrasing peers' work with attribution is what an essay IS.
03 · COMMON KNOWLEDGE
Phrases appearing in 2,000+ independent reference sources (“Water boils at 100°C”, standard biographical dates, well-known scientific formulas). Excluded from headline. Adjustable threshold for fields with denser technical terminology.
04 · UNCITED PARAPHRASE
Paraphrase with no attribution in the surrounding paragraph. THE category that should drive your headline similarity score. Counted at full weight. The fix is almost always adding the (Author, year) reference.
05 · QUOTED · UNCITED
Quotation marks present but no citation, OR no quotation marks at all but verbatim 3+-word phrase from a source. The most plagiarism-ish category — both academic and ethical concern. Counted at full weight.
06 · AI-LIKELY (coming soon)
Per-paragraph AI-likelihood (Binoculars-based, Hans et al. 2024) is on the roadmap. ESL-adjusted band when the lexical signature suggests non-native English. Reported separately from the plagiarism score — AI use is a different question with a different school policy.
What you see after 47 seconds on a 2,500-word class essay.
Headline similarity number that classifies what's actually problematic, per-paragraph AI verdict with ESL band, and click-through to every matched interval with the source URL.
Why a 2,500-word essay needs a different report than a 60,000-word thesis.
Free tier was built for this length
A typical class essay is 500–2,500 words. The free tier covers 2,500 in one paste — fits a class submission without splitting. Longer formats (research papers, theses) need Pro for the full document in one go.
Short essay → references count more
Class essays usually have 5–15 sources in the bibliography. The auto-exclude rule for the reference list matters more here than for a 100-source dissertation: getting it right is what makes the headline number honest at this length.
Shorter text = less AI signal
AI-detector confidence drops on short text — the Stanford 2023 paper noted this. The verdict on a 200-word paragraph is less defensible than on a 5,000-word chapter. We report per-paragraph likelihood with the caveat band so you don't over-trust a single paragraph's score.
Your self-check is yours. Your school's check (if they run one) is theirs.
When you run this self-check on noplag.com, the result is stored in YOUR account, encrypted at rest, accessible only with your passphrase. We don't have an institutional dashboard that surfaces self-checks to teachers. Your teacher can't search for you, can't see how many times you re-ran, can't see what the score was before you fixed things. The only check a school sees is one the school itself initiated through its LMS-integration channel — and you'll know when that's happening because it's tied to an assignment in your Canvas / Moodle / Blackboard. The two are separate by design. Self-check what you want, fix what you want, hand in what you want. Your school's report on your hand-in is a different report.
Read the privacy policyThe questions students ask before they paste.
- Will my teacher see that I ran this self-check?
- No. Self-checks live in your own account, encrypted at rest. We don't have an institutional dashboard that surfaces student self-checks to teachers — the only check a school sees is one it initiated through its LMS integration, and you know when that's happening because it's tied to an assignment. Run a self-check the night before. Nobody at your school can search for it.
- Is the free tier really enough for a typical class essay?
- A typical class essay is 500–2,500 words; the free tier covers 2,500 words per check. For a standard 1,000–2,500-word essay that fits without splitting. For longer assignments (3,000+ words, research papers, theses), you'll need Pro — but 30 days of Pro is free for student email addresses.
- What if I used ChatGPT to brainstorm?
- Depends on your school's policy. The AI detector (coming soon) will report a likelihood score — it doesn't make the call about whether using AI was allowed. Check your assignment instructions. If AI assistance is allowed with disclosure, the report will tell you which paragraphs to disclose. If it's not allowed, it will tell you which paragraphs to rewrite as your own.
- I'm ESL — will the AI detector unfairly flag me?
- Stanford 2023 measured a 61.3% false-positive rate for ESL writers on standard AI detectors. Noplag's AI verdict — coming soon — will ship with an ESL-adjusted calibration that recalibrates against an ESL-validation set when the lexical signature suggests non-native English. Residual error ~3% (documented, not zero). You'll see ESL ADJ next to your verdict if the recalibration kicked in.
- What happens to my essay after the check?
- Stored in your own encrypted folder and added to the Noplag Database by default to strengthen everyone's matches — you can opt out with one toggle. Never used to train anything. Right-to-erasure: delete your account and every fingerprint we hold for you is deleted within 24 hours. We don't sell, share, or use student essays to train AI.
- Will it catch a paraphrased Wikipedia passage?
- Partly, and we would rather be precise about it: light rewording that leaves five-word runs intact is caught today by L1 winnowing. Thorough paraphrasing is what the L2 (MinHash LSH, v1.1) and L3 (vector embeddings, v1.2) layers are being built for, and they are not shipped yet. The classifier separates a paraphrase WITH citation (legitimate) from a paraphrase WITHOUT citation (uncited). If your Wikipedia paraphrase has a (Wikipedia, n.d.) reference, it doesn't count toward the headline score.
- My essay is in a non-English language. Does it work?
- Detection is calibrated for English, Spanish, Portuguese, Polish, and Ukrainian at launch, with more languages rolling out. Detection quality is benchmarked on PAN-PC-11 and published at /docs/developers/benchmarks; per-language evaluation is on the roadmap. Strongest on European languages.
- What if I disagree with the similarity score?
- The engine is Apache 2.0 — open the algorithm at github.com/NoplagLabs/noplag-engine, read the threshold logic, point to the line of code that produced your score. Every interval is reproducible against the engine commit stamped on the report. If you find a bug, open an issue; we treat student bug reports the same as anyone else's.
Check your essay before you turn it in.
Free covers a typical class essay. No card, no signup, no “your school will see this” dialog. Your self-check stays yours.