How much weight should writers give an AI detector score?

[ethanjamescolez]

Question for writers and people who give feedback on drafts.
If a chapter, critique, review, synopsis, or writing exercise receives a high AI detector score, how much weight should anyone give that result?
The report may look precise because it includes a percentage and highlighted sentences, but the output is still probabilistic. It can produce false positives and false negatives, and it cannot prove who wrote a passage, whether it was plagiarized, or whether misconduct occurred.
For a writing discussion, I would rather look at a broader evidence trail:

• notes, outlines, drafts, and revision history
• the writer's sources and citations
• whether voice and structure are consistent across versions
• whether feedback can be answered with specific revision choices
• whether formulaic wording comes from genre conventions, prompts, translation, or editing
• whether the text sample is long enough and preserved in context

A detector might point to sentences worth rereading, but it should not replace a conversation about craft. If a passage feels vague, repetitive, or unlike the rest of the work, those are concrete revision topics regardless of what generated the score.
Disclosure: I work on a small detector/reporting workflow, but I am intentionally not naming or linking it here. I am asking about writing feedback and fair interpretation, not promoting a site or cross-posting an announcement. https://detector-de-ia.net/
Would you use this kind of report only as a private prompt for closer reading, show it to the writer with strong caveats, or leave it out of critique altogether?