AI in The New York Times “Modern Love” Column

LLMs keep us at the most superficial level of experience

Yesterday for The Atlantic, Vauhini Vara reported on a New York Times essay, published in the paper’s “Modern Love” column, that appears to be at least partly AI generated. The essay in question is by writer Kate Gilgan, who lost custody of her son during her active alcoholism. Now years sober, Gilgan compresses years of her complex relationship with her son in a short, highly narrative essay. To be fair, that approach—narrating decades in 1500 words—is fairly typical for “Modern Love” and the author was limited by word count. Still, the subject is one that calls for a great deal of introspection, and it seems Gilgan outsourced some of that to LLMs. Vara reports:

“Kate Gilgan, the author of the column, told me that she hadn’t copied and pasted language from an AI model into her work. ‘However, I did utilize AI as a tool,’ she added, seeking ‘inspiration and guidance and correction.’ She said she’d prompted various products (including ChatGPT, Claude, Copilot, Gemini, and Perplexity) to help her stay on topic in a paragraph, for example, or stick to a theme.”

I write personal essays, sometimes on subjects that are painful to revisit: my mother’s early death from cancer, my tumultuous adolescence, the end of my marriage. Writing well on these topics often requires me to unearth, to some extent, the pain I felt in those moments. In the first drafts, I rarely go deep enough.

In Body Work, memoirist Melissa Febos argues that the initial versions of the stories we tell about ourselves are often simplistic narratives, part of a “theater of types in which we cast ourselves in the role we wish we had played, or the one we were prescribed.” Through writing and revision, Febos argues, we can reach a truer, more complex version. To do so, the writer must nearly always overcome a great deal of resistance—to be able to question the stories she’s always told herself and others, to recall the messiness of the situation and feelings she’s relating. That requires writing and rewriting and being uncomfortable. It requires turning inward. Gilgan, the author of the AI-assisted Modern Love essay, turned outwards.

I don’t know what Gilgan felt while writing her essay, but I strongly suspect she turned to AI in the moments of greatest discomfort. She glosses over the events that must have been most difficult to recreate. She describes, for instance, the moment her son, then age twelve, told her he did not love her and did not want to see her again. I can only imagine how painful it must have been for Gilgan, sober and clear-headed by that point, to have her child cut contact. As a reader (and as a mother), I wanted the writer to sit with that moment, to describe its devastation. Instead, she immediately turns to her recovery, described in the most general terms. At that moment “didn’t give up,” she writes, but rather “shifted.” She stops trying to force a result and instead “started showing up in quiet, invisible ways.” It’s never quite clear what those “invisible ways” are. The strongest parts of her essay reference specific details: the “Franklin the Turtle” books she read her son, for instance. Details, that is, that an LLM could not access.

I’ve started writing essays thinking I was going to make one specific point and discovered—through writing—that the truth is far messier. For example, after Roe v. Wade was overturned, I felt compelled to write about an abortion I had when I was a teenager (here you have a very small window into what my adolescence was like). The initial impetus for the essay was: to make a political/ideological statement about how I had no regret terminating a pregnancy during a time I had zero support and wasn’t taking care of myself very well (that’s an understatement). I thought it would be a straightforward essay. When I started dredging up those memories, however, I realized that my feelings are far more mixed than I imagined. The experience is tangled up with the years of failed fertility treatment I underwent in my thirties and my (absolutely joyful) experience of motherhood now. I still did not regret terminating that pregnancy, but the story was not as straightforward as I’d initially thought.

AI would have allowed me to polish the straightforward narrative I’d planned to tell, not discover a new one. The story I’d planned to write was: X happened, which led to Y. Instead, I was drawn into something much muddier and it took me months (and a lot of crying) to finish the essay. (That essay was slated to be published in a literary magazine earlier this year, but I pulled it from publication for various reasons. It is part of a book manuscript on motherhood that will appear at some point).

AI cannot access the messiness of our memories, nor determine the ways in which those memories are filtered through our current experience. LLMs can smooth things out, focus the paragraphs—but that is the least interesting part of memoir. Gilgan’s “Modern Love” essay is highly linear. X led to Y. Again, she was restricted by the form of the column. However, she repeats many generalities (“a wound,” “heartbreak”) that could have been replaced with concrete detail had she been willing to sit with the memories longer.

Turning to AI for “inspiration and guidance,” as Gilgan did, allows writers to polish the stories they’ve probably been telling themselves (and others) for years rather than moving deeper into the material.

I’ve always seen writing as a process of excavation. The specificities of memory are unearthed by digging, not by staying on the surface as LLMs force us to do.

---

I have so much more to say about this topic, but I need to stop here so I can attend to my day job (as an editor) before my daughter goes on spring break next week, at which point my life will be joyfully chaotic. Maybe she and I will do some non-metaphorical digging in our garden and plant some new flowers.