Skip to content
Matched against published research

Check your writing against the real literature

Every passage is searched across live academic databases and matched to published, open-access work — with a DOI on each source so anyone can verify the overlap. It even catches paraphrased reuse that word-for-word checks miss.

The similarity check runs as part of an AI Detector analysis — one upload gives you both the AI verdict and the source matches.

app/ai-detector · similarity
3 sources · 4,120 words
  • Chen et al. — Learning Analytics Review8.4%
    doi:10.1016/j.lar.2023.04.012
  • Osei & Park — AI in Higher Education6.1%
    doi:10.1007/s11423-022-1018
  • arXiv preprint 2304.098174.2%
    doi:arXiv:2304.09817

// where it searches

Live search across published research

Your text is chunked and searched across many academic databases in parallel, downloading open-access full text where it's available — so matches point to real, citable work.

CrossrefOpenAlexEurope PMCarXivUnpaywallCORESpringerDOAJSemantic ScholarPubMedHALWeb search

This checks your writing against published and open-access literature with DOI evidence — it is not a private student-paper database, so it's built for grounding and citing your work against the real record.

// two scoring lanes

Catches copying and paraphrasing

Two independent layers run over every chunk. The semantic lane is strictly additive — it only ever finds more, and degrades gracefully if a source can't be embedded.

Lane 1 · lexical

Word-for-word overlap

Word and n-gram Jaccard plus bigram containment catch copied and lightly-edited passages with high precision.

Lane 2 · semantic

Paraphrase rescue

A SPECTER2 embedding layer compares meaning, not just words — surfacing reworded reuse that lexical checks alone would miss.

Matched %one word-weighted overlap score, with every contributing source listed.

// evidence you can verify

Every match points to a real source

Overlap is reported chunk by chunk, and each contributing source is listed with a DOI — so a reviewer can open the original and check it in one click.

Your document · chunk 731% overlap

Recent work has explored transformer-based approaches to text classification. The integration of machine-learning methodologies facilitates unprecedented analytical capabilities across academic domains, particularly where labelled data is scarce.

matches Chen et al. — Learning Analytics Review0.62 ratio

Primary sources

  • Chen et al. — Learning Analytics Review8.4%
    via Crossrefdoi:10.1016/j.lar.2023.04.012
  • Osei & Park — AI in Higher Education6.1%
    via OpenAlexdoi:10.1007/s11423-022-1018
  • Preprint — Detecting generated text4.2%
    via arXivdoi:arXiv:2304.09817

Illustrative sources. When nothing meaningful matches, the report simply says so.

// FAQ

Questions, answered

How the similarity check searches, scores and cites its matches.

  • It runs as the second half of an AI Detector analysis — one upload gives you both the sentence-level AI verdict and the source matches. There's no separate document to submit.

// verify your writing

See what your writing matches

Upload a document to get an AI verdict and DOI-linked source matches together. Free to start, no card required.