Co-authors and whose pattern shows up in the score
A detector scores the document you give it, not either author individually — a co-written piece is a genuinely different object than either author’s solo writing.
Blended writing can score differently from either writer alone
Two distinct sentence rhythms alternating through a piece can itself read as unusual to a detector — not because either author’s writing is generated, but because the combination doesn’t match the more consistent voice a detector expects from a single writer.
Decide the writing process before the editing process
A cleaner draft-to-final process — one author drafts a full section, the other edits it, rather than sentence-by-sentence collaborative rewriting — tends to preserve a more legible, attributable structure than constant back-and-forth editing, which is also just easier to explain later if it matters.
If a co-written piece scores mid-range
That is weak evidence either way for the reasons above, more than it would be for a single-author piece. Read the specific reasons behind the score, same as any other borderline result, rather than treating a joint byline as an automatic red flag.
Keep both authors’ drafts, not just the merged version
The same version-history habit that protects a solo writer protects a co-writing team, and is often the fastest way to show which sections came from which author’s own drafting process if that question ever comes up.