We heard A.I. was coming for writing. We didn’t know that writers or journalists would accept it.

Between AI-assisted pieces of journalism and AI-generated novels, some writers aren’t even trying to hide their use of AI. What does it mean for creativity, journalism, and the publishing industry?

A few months ago, I had coffee with a reporter at the New York Post looking to meet people on Substack and other “content creation” platforms.

“Oh, you’re missing out if you don’t use AI,” he told me, after I told him that I wasn’t really using AI for my writing or research because I didn’t want to become lazy. (To be clear: it’s not so much that I have a holier than thou approach to this–I just can see it becoming a crutch in a way that might impede my ability to develop original thoughts and ideas, something I presently enjoy.)

In any case, since that coffee–and in the last week or so–a lot has happened in the world of writing, journalism, and AI.

Just yesterday, the New York Times cut ties with a freelancer whose book review inadvertently incorporated a human writer’s review from The Guardian. Last week, the publisher Hachette had to cancel the release of a horror novel about “a desperate young woman who is held hostage by a man she met online and forced to live as his pet,” as they suspected it was AI-generated. And then, just a few days ago, the Wall Street Journal published a profile of Fortune editor Nick Lichtenberg’s usage of AI to help him churn out 600 stories using AI. (Lichtenberg contacted me a year ago to write an op-ed for Fortune; I reached out to him to ask about the response to the Wall Street Journal article and he couldn’t get back to me in time.)

The responses to this deluge of AI writing or AI-assisted writing have ranged from abject horror to “this is the way it’s going to be, so hop on the train before it’s too late.”

Some are bothered because these AI-writers are not leaving room for those who actually are doing it for the love of the game, with one poster on X writing: “Every professional writer who uses AI for anything except transcription should change careers right now. Give your spot to somebody who loves the work.” (Side note, this feels peripheral to AI and the prediction market’s infiltration into sports, where sports fans are pushing back against match results being too statistically pre-planned or even rigged.)

Another journalist who used to write for the Harvard Crimson posted: “Journalists and columnists are inevitably gonna use AI to write. That’s impossible to police. But we should, at least, make it deeply taboo to admit it publicly” – to which an editor quote-tweeted “Could not agree less.” (It’s not entirely clear whether they were disagreeing with the inevitability of AI or whether it is taboo to admit you’re using it.) Meanwhile, someone with a humanities PhD offered themselves up to the New York Times as an actually honest and original freelance book reviewer.

This week’s posts are sponsored by Parch, which makes non-alcoholic spirits and cocktails.

Writer and tech anthropologist Jasmine Sun seems to have found somewhat of a middle ground, recognizing AI’s limitations and its “sophomoric” writing style while also seeing its potential to be somewhat of an editor. Writing in The Atlantic, she describes how she took her writing archive and notes about what worked and what didn’t, and put it into Claude. While she did that, she made sure to remind Claude that it was not a perceptive co-writer, but a tool to make her smarter. Fellow Atlantic writer Derek Thompson agrees that while “people who outsource the full writing process to AI will find their screens full of words and their minds empty of thought,” but also that all writing involves “outsourcing” and drawing their ideas from “other people, books, and articles” – similar to what I’m doing now, in fact. Further, he says, “anybody who says AI transcription of long interviews obliterates the identity of the writer is being a little silly.”

What’s interesting here is not only that non-writers are getting caught using AI, but that journalists and writers themselves are either touting AI as an advantage, or succumbing to its time-saving temptations. I have to imagine that Fortune Magazine’s PR team greenlit the WSJ profile about Lichtenberg, who admitted that 20% of his work is seriously AI-assisted. And even in the face of the criticism of their colleagues, journalists are now talking openly about the nuances of how they might use it, whether for copy-editing, research, or something meatier.

So why is this happening? I can imagine that for some people in the media who have experienced countless forms of job loss, AI may just feel inevitable to keep working in the media. Perhaps it makes their lives significantly more frictionless–it’s not like writing and journalism is the highest-paid job out there, nor is there job security. And then of course there are people who probably previously could not write at all, yet believe they have the next big romance novel idea or horror story they can feed into an LLM and monetize.

I still strongly believe that the differentiating element for many creative and journalistic works in the future will be people’s abilities to be unpredictable and not so formulaic. In journalism, it might be to source well and connect with humans who have never had a voice before for AI to draw from. In creative writing, it might be to think of the strangest possible permutation of events or character traits that may have not been put together before–and therefore might be out of reach of large language models looking for past patterns. And beyond the New York Times, traditional publishers of books agree, with Hachette for instance putting out a statement reiterating their “[commitment] to protecting original creative expression and storytelling.”

While writing and art are derivative, there are still so many random creative connections and ideas that haven’t been made or scraped by the LLMs, as well as preferences they could not conjure unless they had some individual human with individual preferences guiding them. And on that point, I’ll leave you with an excerpt from novelist Catherine Lacey’s Substack:

“...an individual writer will always write better than an LLM precisely because they haven’t read everything— a writer chooses to read this book and avoid that book, she sees this film, that play, listens to this album obsessively, etc. A writer is limited by his life, and in that limit he makes choices about what to let in and it’s that limitation of influences that, over time, leads him to produce something precise and human. When you’re using an LLM, you’re choosing the blandness of the mean.”

What do you think, readers and writers? I’d love to hear your thoughts. :)

What else we’re reading this week…

From The New Yorker, Kanye West makes an album for the A.I. era: “Bully” is perhaps the first major album of the artificial-intelligence era—the first, that is, to be evaluated primarily in terms of how much it does or doesn’t use A.I.

An Irish man deployed an AI agent with a Northern Irish accent named “Rachel” to call thousands of pubs and inquire about the price of Guinness. Then, with the data assembled, he used Anthropic’s Claude to devise “Guinndex”—a “living, breathing” consumer price index for a pint of Guinness across Ireland.

California Governor Gavin Newsom signed an executive order giving the state four more months to develop AI policy frameworks in the face of President Trump’s deregulation. And California is not the only one—states like Utah and Minnesota are defying Trump too.

Elon Musk claims humanoid robots are going to take over manual labor. From the Washington Post: “Musk argues that Optimus will usher in a world of ‘amazing abundance’ that will make everyone fantastically wealthy, though it’s unclear how those gains would filter down to the masses.”

From Bloomberg, how a 35-year-old crypto bro helped Pakistan broker a relationship with Trump’s Middle Eastern envoy.

A Tennessee woman spent five months in a North Dakota jail because of a facial recognition error and a police department’s failure to corroborate a Clearview AI report.