Can we spot GPT-written content?