What to do with a borderline score
The grey zone is where a score carries the least information and invites the strongest conclusions.
Borderline scores are uncomfortable because they feel as though they should settle something, and they do not. The number is close enough to the line that either reading looks defensible, and that ambiguity pushes people toward the answer they can defend rather than the one that fits. A short procedure helps: slow the conclusion down, gather evidence the score cannot provide, and make the outcome reviewable by someone who was not there.
What this page is for: reading a detection score in context. It is not offered as a way past a checker, and no figure or example here should be taken as a promise about what any detector will report; scores move, and services disagree.
Why the grey zone is the hardest place
- Near the threshold, small differences flip the reading. A fraction of a point can move a passage from one side to the other, which is a property of the cut point rather than of the writing.
- The score is doing the least work there. Far from the line a score adds information; close to it, it mostly adds uncertainty dressed as a number.
- Both error types are live at once. Tighten the threshold and you catch more generated text and more real people; loosen it and the reverse. Near the line you feel both effects together.
- The passage is not the only evidence. Version history, notes, the writer's other work and the task itself all speak to the question a score cannot answer.
- Deciding early feels efficient and costs more later. A decision made to escape ambiguity is harder to defend than one made after looking for other evidence.
- A borderline score is a prompt, not a verdict. It is the moment to gather more, not the moment to conclude more.
A short procedure
- Name the threshold you are using, and why. If it was chosen after seeing this text, restart the step.
- Check the passage before checking the person. Quoted material, boilerplate, headings and very short passages all behave unusually, and none of them reflect on the writer.
- Ask for a second piece by the same writer. Variation between passages shows how much of the reading is about the writer and how much is about the text in front of you.
- Collect one piece of evidence that is not a score. Draft history, notes, an earlier version, or a short conversation about how the piece was written.
- Write the decision and the reasoning in one paragraph. If it cannot be written without leaning on the number alone, there is not yet enough to decide.
- Leave the door open to revise. A decision made on a borderline reading should be easy to revisit if new evidence arrives.
Questions people ask
If the score is just under the line, can I treat it as a pass?
You can, but say so plainly and record it. Treating a borderline reading as a pass is a choice, not a finding, and it should be as reviewable as any other choice.
Should I re-run the same text to see if it moves?
Only to check whether the tool is stable on that passage. If it moves, that instability is itself a useful reading and belongs in your notes.
Is it fair to ask the writer directly?
Often it is the most informative step available. A short question about how the piece was written usually says more than a second run of the same number.
What if I have to decide today?
Then decide with the threshold you stated in advance and record the uncertainty alongside the outcome. A decision made under a deadline is still allowed to be written down as uncertain.
Do borderline cases need a different rule?
They need the same rule applied honestly. The value of a pre-stated threshold is that it does not bend simply because one case feels close.