With the development of technology, the quality of AI-generated text has improved. This is relevant in the AI art field, where AI generates literature or poetry that is appreciated. This study compared human-made and AI-generated haiku poetry, which is composed with 17 syllables and the world's shortest and clearest rules, to examine aesthetic evaluations of AI art and people's beliefs about it. AI-generated haiku were divided into those with and without human intervention. Two tasks were completed by 385 participants. The first involved evaluating human-made and AI-generated haiku on 21 items, such as beauty. The second involved determining whether the haiku were human-made or AI-generated. The results showed that the beauty rating of the AI-generated haiku with the human intervention was the highest, and those of the human-made and AI-generated haiku without human intervention were equal. Furthermore, participants could not distinguish between human-made and AI-generated haiku. These results suggest that human–AI collaboration has better creativity in haiku production. Moreover, a negative correlation was found between discrimination performance and beauty rating in AI-generated haiku, suggesting that high-quality AI-generated work is believed to be human-made. This study indicates the potential of human–AI collaboration in haiku and the underestimation of AI art due to algorithm aversion. •Human-made and AI-generated poetry were examined using haiku poetry.•The beauty rating of the AI-generated haiku with human intervention was the highest.•Participants could not distinguish between human-made and AI-generated haiku.•AI-generated haiku showed a negative correlation between beauty score and hit rate.•State and trait factors that explain the beauty and accuracy of author distinction were examined.