Snook tagged me so after re-reading this many times it finally dawned on me what you were talking about. Maybe that's just my poor reading comprehension.

According to https://www.zerogpt.com/ the following lines got pegged as AI:

- The next morning, he had shown up at her house bright and early. He needed all the time he could get. 
- Just rest, and it will be here before you know it."
- "I was thinking, maybe we could meet at the stables at daybreak?
- He didn't imagine that he'd sleep much as his mind was racing with dilemmas. She seemed to enjoy him, but she didn't know him. If she didn't know him, how would she love him?

And some other lines which likely appear in a whole bunch of stories which ChatGPT has stolen to 'learn' from.

Another AI detector at https://contentatscale.ai/ai-content-detector/ found entirely different lines for the most part:

- He needed all the time he could get.
- He held out his arm, hoping that she wouldn't hesitate.
- Just rest, and it will be here before you know it."
- It was so contagious that he found himself wheezing with laughter too, but he had no idea why.

Anyways, my verdict is AI content detectors are bullshit. Like those 'original content' detectors that told everyone their plagiarism was original. AI content is blatantly obvious without any detector. Its so obvious it would be a stretch to say it's not AI. In those cases the detectors serve a purpose of identifying which generator possibly created it or just supporting what people already know, that it's AI.