About fifteen years ago, I was working for a small but influential weekly paper in New York that was under assault. On the business side, our ads were disappearing to the internet; on the editorial side, scrappier, savvier online-only publications were siphoning away our readers.
In an effort to save our sinking dinghy of a paper, the owner occasionally invited experts to come talk about the still mysterious digital frontier and how we might succeed there. I remember sitting in a conference room with Craig Newmark, just about the humblest person I’ve ever met with a globally-known company named after him, and another gathering with some hot shots making a presentation larded with ye olde Web 1.5-era jargon. During the latter meeting, one digital visionary who probably referred to himself as some kind of “rock star” or “ninja,” told us that we needed to add more links within our stories since “Google likes lots of links.”
Because I was younger then and loved the sound of my own voice, I recall saying something like, “Why would I do that? I work for the New York Observer, not Google.”
His response still echoes in my mind: “We all work for Google now.”
The above is what we in the journalism racket call an anecdotal lede, a bit of throat-clearing that’s meant to help illustrate some of the larger points of the piece you’re reading (or more likely skimming). While a decent editor would probably scrap it, I’m using it to lightly establish my authority while offering a bit of my humanity. It lets the reader know that I’m someone worth listening to as I gear up to make my point.
Was my story self-indulgent? Maybe even a bit self-aggrandizing? Sure, but then again, so am I, and so is nearly every other writer you might ever hope to meet and thank for our hard work promoting Truth for the benefit of the World. (You’re welcome.)
It’s exactly those sorts of human qualities that disappear when artificial intelligence programs — like the one that spewed out a summer reading list full of hallucinatory recommendations for the Chicago Sun-Times — are employed to create what I can still barely bring myself to call “content.”
It’s extremely impressive that, with a few prompts, programs developed by San Francisco companies like OpenAI, Anthropic, and others can cobble together sentences and paragraphs and even mimic the tones and styles of publications and individual writers, much less come up with semi-plausible plots for Rebecca Makkai or Rumaan Alam novels. But even as the words these AI bots generate are already appearing on screens and in print, even the best are destined to be hollow at their core. A bot’s output will always lack the thing that brings the best writing to life: humanity.
Could a bot — regardless of how many pieces of prose it scans — write with the cool intelligence of Joan Didion, or achieve the moral clarity of James Baldwin, or capture the misanthropy in Tom Wolfe’s exclamation point-riddled pensées, or, despite the number of jokes it analyzes for impact and structure, replicate the wit of Dorothy Parker or, should it let its digital freak flag fly, replicate the “gonzo” journalism of Hunter S. Thompson, or even, should some nostalgist dream it to, resurrect the dilated, syntactically limber sentence structure of Murray Kempton, whose switchback-filled, 106-word-long, copy editor-infuriating sentences I have just poorly taken a stab at emulating?
No. It cannot. Because what each of these writers brought to their work was themselves, something even the most cutting-edge bot can never hope to emulate because a bot has no self to bring.
And, yet, we have to sit with the reality that we writers have been training ourselves to think and write like bots.
In the decade and a half since I learned that I worked for Google (minus the salary, stock options, and snack-filled grab-and-go kitchens of actually working for Google), journalists have internalized the lessons of that web expert. We’ve flattened our tics to attract clicks.
We’ve forced ourselves to write shorter pieces with punchier ledes, the better to keep up with social media’s constant refresh. Our headlines are now packed tightly with proper nouns and shorn of extra filigree in hopes of “optimizing” for search engines. We’re letting algorithmically-determined trending topics serve as our assigning editor, prompting us to write about whatever is happening right now. No, now. Editors spend their days looking at stories that achieved high traffic in order to replicate them, more or less, exactly in hopes of similar results.
Need more proof of the bot-ization of writing? Look no further than X (if you’re still there), where many writers spent the last decade training themselves to write within the company’s artificial constraints: limited word count, no editing (except with a premium membership), and arguments so shorn of nuance, they might as well be binary code. The short, sharp shock of X makes the construction of sentences, the building of paragraphs, and the overall effort of writing just too difficult to bother with. X’s visible metrics taught writers what worked and what didn’t, allowing the canniest to occasionally “win” the platform, albeit on the platform’s own narrow terms.
How is this reconfiguration of our writing — and our thinking — any different from feeding petabytes of text to large-language models to mimic AP or New York Times styles in order to generate mathematically precise simulacra of articles and editorials? We turned ourselves into bots in order to create content (there’s that word again) to appeal, in large part, to bots. Now the real bots are coming for our jobs and all we can do is look on in horror.
The only thing writers can do, then, is to write more like ourselves, to become, in a sense, more human.
We may not be able to beat the bots at their own game, but we can choose to play among humans again. Writers need to dig deeper into their personalities, write with exuberance and pretentiousness, deploy ever more self-indulgent anecdotal ledes, and reclaim the written word for the sake of all of our humanity.
