AI Detection vs. Proof of Authorship

Two different questions

“Does this text look like AI wrote it?” and “Did this person write this?” sound similar, but they are different questions. A detector answers the first one, with uncertainty. Only evidence about the writing process can answer the second.

How proof of authorship works

• The writer works in a tool that records the writing process: typing, deletions, revisions, pastes, and time spent.
• The record is sealed as it is created so it cannot be edited afterward.
• When finished, the writer gets a certificate tied to the exact final text.
• A teacher, client, or editor opens the certificate to see a summary and, if the writer allows, a replay of how the document came together.

Side by side

• Works on: detection works on any text, even old text. Proof only covers writing done in the recording tool.
• Result: detection gives a probability. Proof gives a record of what happened.
• False accusations: detection can wrongly flag honest writers. Proof protects them.
• Evasion: detection is defeated by paraphrasing. Faking a natural writing record takes far more effort than writing.
• Who controls it: detection is run by the reviewer. Proof is held by the writer, who chooses what to share.

Can proof of authorship be fooled?

No system is perfect. Someone determined could retype AI text slowly into a recording editor. Good systems look for the difference between composing and copying: real composition has pauses, rewrites, and reorganization, while copying is steady and linear. The goal is to make cheating more work than honest writing and to flag the patterns that suggest copying.

When to use which

• Use a detector to screen text you already have, as a reason to look closer.
• Use proof of authorship when you know in advance that your writing may be questioned: graded essays, client deliverables, contest entries, and job applications.
• Teachers and clients can ask for proof up front instead of relying on detection after the fact.