Background & Context§
On April 1, 2025, security researcher Michał Zalewski (lcamtuf) published a Substack article titled "AI children's books, body horror edition," showcasing a collection of Amazon-published children's books that appear to be entirely generated by AI with minimal human oversight. The books, such as "The Adventures of Captain Fluffernutter in The Land of the Snugglebugs" and "100,000 Whys: A Journey into the World of Fun and Discoveries," feature illustrations that are not merely cartoonish but deeply unsettling: characters with jaws protruding at unnatural angles, tangled limbs, and grotesque facial distortions. The article quickly went viral on Hacker News, where commenters expressed a mix of shock, dark humor, and existential concern about the state of AI-generated content targeted at children.
This incident is not an isolated anomaly but a symptom of a broader trend: the commercialization of AI slop—low-quality, mass-produced content generated by large language models and image generators with little to no human curation. As AI tools become cheaper and more accessible, the barrier to publishing a "children's book" has effectively dropped to zero, leading to a flood of titles that prioritize quantity over quality. For developers, data scientists, and founders building on AI platforms, this case raises critical questions about content moderation, ethical deployment, and the unintended consequences of generative models when applied to vulnerable demographics.
The News: What Happened Exactly§
Zalewski's article catalogs a series of AI-generated children's books that are not merely mediocre but actively disturbing. For instance, one book features a character whose jaw appears detached from its skull, sticking out sideways in a manner reminiscent of the manga "Attack on Titan," as one Hacker News commenter noted. Another image shows a child-like figure with a limb twisted into a pretzel-like shape, while a third depicts an adult holding a baby with a finger bent backward at an unnatural angle. These are not subtle flaws; they are gross anatomical impossibilities that any human illustrator would immediately correct. The text, meanwhile, is grammatically passable but devoid of narrative logic, often rambling through nonsensical plot points or abruptly ending mid-page.
Hacker News users reacted with a mix of horror and black humor. One commenter wrote, "My kids and I certainly had a good laugh looking at the pictures in the blog. That second picture of the jaw sticking out had my son ROFL-ing." Others were less amused: "The lack of real effort bothers me more than the content itself. We can't be bothered to even proofread children's books anymore?" Several users noted that these books are likely created by scammers using Amazon's Kindle Direct Publishing program, which allows anyone to upload a book in minutes. The "100,000 Whys" title, which claims to pack 100,000 questions into a 120-page book, was ridiculed for its mathematical impossibility—that would require over 833 whys per page.
Beyond the amusement lies a deeper concern. As one commenter articulated: "I've been wondering what the long term result on people's perceptions of reality will be after all this AI slop. ... Someone who grew up seeing AI slop from the start doesn't have that firm grasp on reality to spot fake content from." This echoes a key anxiety: children exposed to these distorted images and incoherent stories may internalize abnormal representations of human anatomy and narrative structure, potentially impairing their ability to recognize quality content in the future. The books are not just weird—they are pedagogically toxic. As another user put it, "There are some things that I feel really shouldn't be enshitified by AI and this is one of them."
Historical Parallels & Similar Incidents§
This phenomenon is not new; it is a direct descendant of the "shovelware" industry of the 1990s. Shovelware referred to low-quality software bundles—often CD-ROMs containing hundreds of games, clip art, or utilities—that were rushed to market with little testing or curation. Companies like CDAI (a German publisher) were notorious for releasing discs titled "700,000 Games" or "10,000 Fonts," which contained mostly freeware, demos, or non-functional executables. The parallel to "100,000 Whys" is striking: both employ inflated numbers to imply value, both rely on automated or minimal-effort production, and both target unsuspecting consumers who assume a product is legitimate because it is sold on a reputable platform (then retail stores, now Amazon). The difference is that today's AI shovelware can be produced at near-zero marginal cost and distributed instantly, amplifying the scale of the problem. As one commenter noted, "Slop aimed towards children has practically always existed. The '100,000 whys' naming reminds me of an old '700000 games' CDAI slop—AI is just a more complete reimplementation of the 'shovelware' from the 90s."
Another historical parallel is the proliferation of "vibe coding," a term for AI-generated code that appears functional but is often riddled with security vulnerabilities and logical errors. Just as vibe coding has led to a surge in insecure software deployed to production, AI-generated children's books represent vibe illustration and vibe storytelling, where the creator relies entirely on the model's output without verification. The consequences, however, are more insidious: bad code can be patched, but a child's developing perception of reality is not easily corrected. A commenter on Hacker News summed up the dilemma: "Just like Vibe coding, I think this article is trying to convey that, when children’s books important not to rely solely on AI, but to consciously strive to create high-quality works." This marks a critical inflection point: what happens when generative AI is used to create content for the most impressionable audience with zero human oversight? The answer, as Zalewski's article demonstrates, is body horror.