An abstract compresses: no examples, no digressions, every noun nominalised. Those constraints overlap with what detectors read as machine-like, so a perfectly correct abstract can score higher than a messy body full of specifics.
The fix is one concrete anchor: the population, the number, the intervention, the site. Keep the compression, add the detail. Then check the body separately, because the flag usually tracks the paragraph texture rather than the topic.
This piece on detector accuracy explains why abstracts are a known weak spot, and this academic guide covers the balance.