How a check works
Similarity checking with a declared scope
You choose the scope. The local corpus check runs on this application. Web discovery is a separate choice and it is off until a search provider's storage terms allow a durable report.
What the number means
Similarity is coverage inside a named scope.
Verified token coverage equals the unique word tokens that sit inside verified passages, divided by the eligible word tokens in the submission. Overlapping sources are counted once. Semantic similarity and AI signals are not added in.
Searched in this demo
- Four original English and Arabic sample documents written for this project
- Licensed CC0, with stable source identifiers
- Verbatim and near-verbatim character alignment only
Not searched
- The public web, journals, and books outside the sample
- Any private academic database, including Turnitin
- Other customers’ files, which are never corpus sources by default
- Paraphrase, translation, or AI authorship
Method
Evidence before a figure.
- 1
You submit text
The development path accepts pasted text, or a plain-text file read locally. It does not upload PDF or DOCX and it does not keep the submission.
- 2
The named scope is compared
A pinned local engine aligns the text with the sample corpus. Fingerprints narrow attention; only aligned passages count.
- 3
You see what was and was not checked
The result lists verified excerpts, character offsets, source identifiers, and the searches that did not run.
How a check works
Local corpus
The check extracts text, finds fingerprint candidates in the licensed development corpus and your authorized private sources, then aligns passages. Only verified spans enter the percentage.
Web discovery
When the provider is enabled, you must consent again. The application sends a small number of excerpts, fetches allowed pages, and verifies them independently. A search snippet does not raise the percentage. If the web stage fails, the report says the coverage is partial.
What the figure is
Overlapping matches are counted once. A passage shorter than 50 characters is below the verifier minimum. AI writing scores are disabled. This page does not offer a live internet check while the provider is switched off.