Was this text likely AI-generated?
Focus Language origin
Method Artifact-level inference
Why AgencyThread
Where the words came from is not the same as who shaped the work
AI Detection, process provenance, and AgencyThread ask related—but different—questions about AI-assisted work.
What happened during creation?
Focus Process history
Method Event/process record
What did the human materially contribute?
Focus Intellectual agency
Method Evidence-based interpretation
Continuity across the work
Follow the human thread through AI-assisted work
Meaningful human contribution is rarely one event. It can run through a project from the initial framing decision to a later critique, a rejected suggestion, a source check, a structural change, and a final judgment call.
That continuous thread is what AgencyThread is designed to interpret. Not where the language came from. Not how many times someone clicked, typed, or revised. But the thread of human decisions and intellectual participation that shaped what the work became.
1Frame
2Direct
3Evaluate
4Develop
5Verify
6Decide
Consequential judgment
Measure the thinking, not the typing
AgencyThread does not reward prompt counts, edit counts, keystrokes, time spent, manual retyping, cosmetic rewriting, or arbitrary disagreement with AI.
Those are activity signals. They are not reliable evidence of intellectual agency.
A person can exercise excellent judgment by reviewing an AI contribution and deciding that nothing needs to change. Another person can spend considerable time retyping the same contribution without materially shaping it.
AgencyThread is built to distinguish between activity and consequential judgment.
Not agency signalsPrompt count · Keystrokes · Time · Retyping · Cosmetic edits
Agency signalsFraming · Judgment · Reframing · Evaluation · Verification · Accountability
A different premise
This is not detection with softer language
AgencyThread does not produce a percentage of text that is “AI.” It does not attempt to catch someone using AI.
Its premise is different: once AI is part of ordinary knowledge work, the more useful question is not merely whether AI participated. It is what the human actually contributed within that collaboration.