AI Children’s Books and the Body Horror of Automated Publishing

Summary

The essay follows an earlier visual demonstration arguing that AI-generated content often looks recognizably alike. The author had collected about 220 AI-generated children’s books but initially avoided judging their quality, saying debates about quality usually end with the promise that a future model will improve. After buying one Amazon bestseller as a parent, the author focuses on children’s encyclopedias, arguing that they are attractive targets because they likely sell well, are often bought as gifts by adults who are not the readers, and can be produced more cheaply than traditional books without directly copying closely protected fictional properties. The article stresses that encyclopedias influence children and therefore deserve a high standard. Against claims that frontier models had reached PhD-level intelligence by mid-2025, and based on publication dates and the artwork’s apparent connection to a major laboratory’s flagship model, the author presents a series of grotesque visual failures: distorted reflections, an unfamiliar planet, an altered cat, fused animals and trees, and other unsettling scenes. The author notes that rankings and reviews can be manipulated, but says similar material is widespread in relevant Amazon categories and likely generates real sales. The conclusion is qualified rather than absolute: future models may produce flawless children’s encyclopedias, but current AI-generated books can expose children to badly formed and disturbing content.