AI detection, watermark checks, and character scans
If your question is “Why does this value fail to match?”, a code-point report can be useful evidence. If the question is “Who wrote this essay?”, the same report is insufficient. An invisible character may have entered through a document editor, a PDF export, or a manual formatting choice.
Authorship assessment, provider watermark verification, and text cleanup are different tasks. A system designed for one does not automatically support the others. Before acting on a result, identify what was measured, how the method was validated, and what alternative explanations remain.
How to interpret hidden-character findings
A finding identifies a supported character in the pasted input. The count tells you how often it occurs. A proposed cleanup tells you what the tool would remove or replace. None of those observations identify a person, model, or generation history.
An empty report is equally limited. It means the scanner did not find the patterns it covers in the supplied text. It does not certify that the passage is original, accurate, human-written, or free of a statistical watermark. Treating “nothing found” as a universal clearance would exceed the evidence.
How to assess AI text with supporting evidence
When authorship matters, examine relevant drafting records, sources, version history, and the writer’s explanation of their work. Apply the same expectations consistently and give people a chance to address specific concerns. A writing style, punctuation preference, or odd space alone is a poor basis for an accusation.
For the practical cleanup task, keep the original, inspect the findings, and verify the output in its destination. For a provenance task, look for an appropriate provider method or signed source records and understand what those records actually establish. This separation gives both workflows a clearer, more defensible result.
Sources & further reading
Primary references for the technical points in this guide.