The evidence behind every check.
Credence runs a similarity check against reference sources and an AI-generation likelihood estimate on text and images - then shows the matched excerpt, the score, and the reasoning behind every call. A text-matching report, not a verdict.
Originality
93%
AI likelihood
12%
"...the matched excerpt appears here, alongside its source and similarity score."
Contents
Similarity detection, with sources
Cross-checked against reference texts, your own prior submissions, and the open web. Every match ships with its source excerpt and a similarity score - a text-matching report, not a verdict.
AI-generation likelihood
Burstiness, word-entropy, and phrasing-pattern analysis, reported as a percentage with an honest confidence band - never a bare yes/no.
Image provenance
EXIF metadata, edge-artifact analysis, and C2PA content-credential lookups, cross-checked in a single pass over the file.
Writing feedback
Passive voice, vague language, repetition, and overlong sentences - flagged with the reasoning attached, never silently rewritten.
Who it's for
Students
Check a draft before submission - see exactly which passages need a citation or a rewrite, before a grader does.
Educators & graders
Review submitted work with the matched source and similarity score attached, instead of a bare percentage to take on faith.
Writers & researchers
Verify originality and screen for AI-generated passages before publishing or submitting for review.
Frequently asked
It's the percentage of your submission that matches text found in reference sources, your own prior submissions, or the open web. A high match isn't automatically plagiarism (quotes and common phrasing match too), and a low score isn't automatic proof of originality. It's evidence to review, not a verdict.