Last updated · August 2026
Integrity Policy
What We Check
Silvertext runs four checks on one upload: plagiarism matching against published pages on the open web, including journal and publisher pages that are openly reachable; grammar, spelling and punctuation; citation completeness and format; and an AI-likelihood estimate. Every finding is returned with the correction, the source, or the citation that resolves it.
What a Score Means
A plagiarism percentage describes overlapping words, not intent. An AI likelihood describes resemblance to model output, not authorship. Both are reported with the evidence behind them, and in bands wide enough not to imply precision we do not have. Neither is a determination about a person, and we will not present them as one.
Your Text and Your Data
Your document is processed to produce your report. It is not added to any archive that other people are checked against, and we do not use it to train models. We configure our providers not to train on it wherever they offer that control. A file you upload is deleted once its text has been read; the text itself is saved in your report so you can come back to it. Your report stays in your account until you delete it, and deleting your account (confirmed by email, carried out 30 days later unless you cancel it) deletes your reports with it.
What We Will Not Do
We do not sell writing for you, ghostwrite submissions, or offer a service whose purpose is to defeat a detector. We do not report you to your institution. We do not publish accusations, and we do not share your findings with anyone you have not authorized.
Evidence, not verdicts
Any integrity conversation should start with a human reading the paper. Our output is evidence for that conversation, not a substitute for it, and it is yours to bring, not something we send anywhere on your behalf.
Questions about how this applies to your institution? Talk to us. A person answers.