Similarity detection that reads like an examiner
One number is easy to produce. A finding you can defend takes more: exact matches, reworded passages, masking tricks and legitimate quotation — each detected, labelled and filterable on the original document, so the final judgment stays human.
Where we look before anything is scored
A similarity score is only as good as the ground it searched. Every submission is checked against four source pools — consolidated into one report with one source list.
The open web
Phrase-level search across live web pages, with results consolidated by host — so one website becomes one line in the source list, not twenty scattered rows.
Open scholarly sources
Open-access journals, preprints and academic repositories, searched through scholarly APIs — with full-text body matching where sources provide it, not abstracts alone.
Your own repository
Every submission your institution processes joins a private index, so this term's essays are checked against every term before them — visible to your institution only.
Student-to-student collusion
Cohort-level matching surfaces shared and recycled work between students — even when the text never appeared anywhere public.
Copying has layers. So does detection.
Verbatim copying is the easy case. The report separates four kinds of evidence, so an examiner can see not just how much matched — but how.
Exact matching
Verbatim and near-verbatim runs are matched at sentence level, so mosaic copying — a sentence here, a clause there — accumulates into the score instead of slipping under it.
Paraphrase-aware semantic matching
Passages rewritten to dodge word-for-word checks are caught when they keep the source's structure and meaning — and labelled as paraphrase matches, distinct from exact ones.
Anti-masking
Homoglyph substitutions, invisible characters and similar masking tricks don't quietly defeat the check — they are detected and flagged on the report itself, as evidence for the examiner rather than a blind spot for the checker.
Citation & quote awareness
Quoted passages and bibliography entries are recognised for what they are — so properly referenced work can be filtered out of the score instead of counting against the student.
A score you can interrogate
The headline number is a starting point, not a verdict. Live filters let the examiner apply judgment — and watch the score respond.
Filters that recompute the score as you read
Every finding is annotated on the original document layout — the same view in the browser and in the PDF. Toggle a filter and the similarity score recomputes immediately, so the number you defend is the number after judgment was applied.
- Exclude quotes — recognised quotations drop out and the headline score updates instantly.
- Exclude bibliography — reference lists are removed from the calculation without hand-editing.
- Small-match threshold — screen out incidental short matches that add noise, not evidence.
- Per-source exclusion — rule out the assignment template or a permitted source; the score follows.
What we cover — and what we don't
Similarity questions
Is a high similarity score proof of plagiarism?
Can students beat the check with character tricks?
Does paraphrased text actually get caught?
What happens to our students' work after a check?
Read a real report before you decide
Book a demo and walk through a similarity report with our team — or start a free 14-day evaluation and put 50 real submissions through the full pipeline.