Comparing scientific abstracts generated by ChatGPT to real abstracts with detectors and blinded human reviewers - npj Digital Medicine