I was shocked: my husband was using AI to write our children’s bedtime stories

The other night, from the hallway outside my second-grader’s bedroom, I heard her ask my husband for a very specific bedtime story – so specific I could have sworn she was choosing prompts just to screw with him. She wanted one that was about “mowing lawns in a place called Bananaland, and a festival, and monkeys, but make it funny”.

G’luck, I thought, as I started to tiptoe away, having completed my read-aloud portion from The Swiss Family Robinson, which we’re making our way through together each night.

I stuck around to hear him be eaten alive. But then he cleared his throat and began.

There was a protagonist, a “clever and adventurous girl, who was not your typical Bananaland resident”. There was a lawnmower, named the Banana Blade, that featured prominently. There were aromas of fresh-cut banana leaves, and mischievous monkeys, who threw banana peels into her path. By the time he got to the riddle the adventurous girl needed to solve, delivered in rhythm and rhyme, my entire world order had heaved on its axis.

He works in venture capital. I’m the writer. And all I have the bandwidth to do at the end of the day is read aloud to her from books other creative people have written, even if their plots feature a “good little mother” who spends her days cheerfully whipping up delicious meals of roasted penguin and being praised for her frugality and Christian values.

To not only come up with a brand-new story, but also add a rhyming riddle into the mix? Just whom had I married? I started to cycle through the stories of other women who’d also found themselves living with men who harbored deep secrets, from Rosemary’s Baby to Jane Eyre. Was he communing with the devil? Did he have a first wife hidden in the coat closet?

“It’s ChatGPT,” he whispered, with a shrug, after he tiptoed out and found me slack-jawed and panicky. “I just feed in her insane prompts and it spits out a story.”


My panic then took a left turn. He wasn’t communing with the devil, exactly, just a robot – but in my profession, the one in which I string together words for a living, the line between artificial intelligence and Satan gets fuzzy. Another thought nagged at me: was this yet another product of a consumerist culture that promises instant gratification to children, who can be heard yelping up at their parents from strollers all over the world: “Just go to Amazon and buy it”?

While I fondly remember the serialized stories my father used to tell me at bedtime, often featuring a wise-cracking pigeon name Lou (“I’m walkin’ here!” he’d squawk as taxis swerved to avoid hitting him), I long ago gave up the idea that I could tap into some creative fount at bedtime.

It’s a combination of world events, the obligations of a busy life with three little kids, and the relentless stream of illness that comes along with that – we just recovered from a month-long period in which our household boasted three RSV infections, two strep throats, one case of croup, one ear infection and one pneumonia diagnosis (that one was mine). Like the homemade three-course meals that pop up on my Instagram feed, shot by the Parisian mothers who’ve somehow infiltrated my stream, spinning yarns for a rapt audience is a goal to strive towards. This means that the preschooler, who’s going through a Mommy phase and only wants me to put her to bed, hasn’t been told an on-the-spot story in ages.

“Oh yeah, we do robot stories,” one friend told me when I started asking around. “For sure,” said another. “I just plug in prompts with the kids’ names in them.” Heading online, I found that articles abounded exalting the welcome assist of having personalized bedtime stories at the fingertips of even the most tapped-out parent (including Alexis Ohanian, the co-founder of Reddit, who is apparently a fan).

A few years before AI started to dominate the headlines, I published a book that investigated the intersection of parenting and technology, and came to the conclusion that despite various marketers’ claims to the contrary, tech by and large only serves to transfer a parent’s focus from their own child to a device. Its promise is to “hack” a moment, make it more efficient – but rarely is that the right goal for parents, who shouldn’t be trying to optimize moments with their children, just doing their best to be present, without a scrim of blue light in between. I wondered: could the introduction of AI change that?

The next morning, having checked the closets for wives, I ventured over to ChatGPT and started feeding it some prompts.

“Tell me a story in the style of Goodnight Moon about trucks” (for the two-year-old).

“Have the characters of Swiss Family Robinson visit New York City” (for the second-grader).

“Tell me a short bedtime story for a four-year-old who likes Frozen, and make it funny” (for the preschooler).

I quickly realized that anything AI-generated for the two-year-old was missing the point. His experience of Goodnight Moon has more to do with curling up in my lap and looking at its comforting illustrations than it does to do with the prose itself, sleeping trucks in the garage be damned.

ChatGPT wildly bungled the Swiss Family Robinson prompt, even as I continually refined it, plopping the characters in Times Square, where, in a sequence that would make Johann David Wyss roll over in his grave, they “danced whimsically in the glow of neon signs and laughed under the twinkling lights, creating memories that would last a lifetime” (right up there with eating roast penguin).

It did the best for my four-year-old, introducing that bulletproof comic preschool element, a fart, but making it “frosty”. Even so, it was moralistic (“sometimes the silliest moments create the most magical memories of friendship”), formulaic and relied heavily on that cardinal sin of writing, telling and not showing. You can deploy mischievous eye twinkles only so many times before the frosty-farting little snowman starts to appear as deranged as Jack Nicholson in The Shining.

The algorithm is honest – in its robotic way – about its shortcomings as a bedtime storyteller. It offered that, if used appropriately, it could provide a “time-saving” tool (again with the optimization!) to “co-create” with your child while “teaching morals and values” (Lou never once taught me morals, except that jaywalking is acceptable), and “fostering of love of imagination” (though I’ve yet to come across a child so pragmatic that they need help with that).

Until we merge more fully with the machine, until AI can mimic my voice, my history, my personal creative instincts, the entire act remains a simulation

But it cautioned that “it’s essential to use ChatGPT as a supplement to, rather than a replacement for, your own storytelling”, and pointed out that overreliance on AI at bedtime might “encourage passivity”, both in the tellers and the listeners. Well, duh. Bedtime is arguably my most defenseless parenting moment of the day. The idea that I’d be able to use the text as a jumping-off point for my own creativity and not simply read it word for word is as realistic as serving my kids pâté en croute for dinner.

The real issue, I concluded as I mulled it over more, has less to do with the skill of the AI model, which will no doubt improve, and even less to do with the issue of feeding into the demands of our children. It’s a fundamental misunderstanding of how much joy a kid is taking from a story, tailored so specifically to him, and how much he’s taking from the hearing of that story, any story, filtered through the medium of a beloved parent who’s looking him in the eye, or stroking his head, and not craning towards the blue light of a phone like a dystopian sunflower.

Until we merge more fully with the machine, until AI can mimic my voice, my history, my personal creative instincts, the entire act remains a simulation, from the way the words are put together to the feelings they are supposed to evoke. Which is likely why my second-grader stopped asking my husband for AI stories a couple weeks in, pronouncing them “repetitive”.

So, as we wrap up this AI-fueled year, I plead with fellow parents to resist the pull of AI’s tentacles at bedtime. Everything in moderation, sure – there are a lot of bedtimes. But the idea that my child will grow up with a profound lack of the kind of whimsical bedtime joy that can only come from co-creating morals-driven, hyper-specific stories with me is ludicrous. She’s evolutionarily programmed to find that joy and whimsy, even if it takes us grownups a moment to catch up.

On the way to school the other day, we spied a group of pigeons congregating on a street lamp.

“What are they doing there?” she asked. After a beat, she answered her own question: “Going to school, right?”

In the remaining blocks, the two of us created a parallel world, one in which squirrels and pigeons and sparrows were off to schooldays of learning how to hide acorns, and fly, and eat worm snacks. We’ve kept it going every morning since. As I dragged her scooter hurriedly behind me one morning, she forced me to slow down, pointing to a squirrel scurrying over the stone wall separating the street from the park.

“He was late to school, just like us,” she said. “Too bad he forgot his scooter at home.” I could swear I saw her eyes actually twinkle.

Leave a Comment