Writing in the Time of AI
It’s an odd time to be a writer.
I open my newsfeeds to see stories about authors being decried for secretly writing books with AI and people vowing to boycott Barnes & Noble over their declaration that they will stock AI-generated books. Then I load LinkedIn and every post I see is either sponsored advertisements for the newest, most cutting edge generative AI features or chatty professionals talking about how of course they don’t need AI to do their work—but here are the 12 ways they’ve found to use it to streamline their process. College graduation speakers are praising AI as the future to crowds of booing students and every upper level professional is gushing about how they can do they can do the work of multiple employees with AI while readers on social media scoff at any post using an em dash as “obviously AI.”
Over a year ago now (time flies), I was working on a freelance writing and editing gig for a client who wanted me to check over influencer submitted recipes. I would log in to their accounts and carefully fill out spreadsheets with measurements all converted to consistent units, missing ingredients filled in after researching similar recipes, instructions rewritten to sound consistent from recipe to recipe, and cheerful, brand-appropriate copy telling people how each recipe could be a perfect weeknight pick-me-up. After reviewing dozens of recipes already on the site and cross referencing them with the tone of voice being used across their web copy and social media, I wrote myself a style guide to follow because nobody had bothered to write one before. It was quiet and meticulous work, the kind that generally goes unnoticed when it’s done right.
It wasn’t a surprise when my contract manager cheerfully came to me asking if I thought AI could take over my job, but it did still sting. I drafted a calculatedly friendly and non-hostile email back, explaining what was obvious to me as a writer: not only could a large language model not do this job, but it’s precisely what they’re worst at. Attention to detail, consistency across pieces of copy, an understanding of what ingredients make sense in a recipe, and of course the ability to understand and viscerally describe how good food makes you feel are all beyond the capabilities of LLMs that will confidently reaffirm that “October” has an “M” in it. My manager listened politely and didn’t believe me. Instead, he asked me to test several to see which did it best.
I had never intentionally used generative AI, but there I was with 6 different models open, running drills with them in a job interview for my own replacement. It made me feel nauseated. I tested them diligently, ranking them on how they performed according to the style guide I had written for my own use, thinking it would one day be a kind thing to hand over to the next person who took over my job. Instead, I watched loading wheels spin and the LLMs fail at every hurdle that I had anticipated, churning out work that required just as much (if not more) editing as the original recipes I had fed into the system. I noted down my results, comparing each model’s performance (universally bad) and reporting back on the pros and cons of each. I showed my report to another writer before submitting it to my manager. She sympathized with me having had to take so much time away from my normal work to do this task, but said I had done so dutifully and should be commended. When I submitted the final report, my manager thanked me for my work and expressed that it was a shame that nothing worked quite right but he appreciated me checking.
Contract renewals came around within a month or two and my contract was not renewed.
This wasn’t a surprise either but this one stung worse. A few months later, I talked with the writer I had shared my draft with. We chatted about old clients and with a laugh she said how funny it was that she had been pulled to do my old job—recipe editing. Some bitter part of me was happy that at least in some way I had been proven right: an AI couldn’t do the job.
Freelance jobs are few and far between now and half of the job listings I find are positions for writers to help train LLMs. I sigh every time and tell LinkedIn not to show me those listings anymore but there are always two more reappearing for each that I close like some kind of digital hydra. When I get tired of looking at the listings I go back to my main feed and see another tech bro talking about how he’s cracked the code on generating more better social content than ever before and how easy it is with AI. Every job I actually apply for say that applicants need to be in tune with Gen Z and Gen Alpha in the same breath as they emphasize how important it is to be AI-savvy to streamline writing while in my life I see that same “youth demographic” gleefully calling out AI posts and users as “clankers,” the new slur they’ve made up to mock those they see as bootlickers for genAI. Most of those applications go unanswered, not even bothering to send a rejection letter.
I’m writing this post from an airplane, looking out at the sunset through the clouds. On the upside, I didn’t have to check my once busy work calendar when I booked my flights. On the downside, the tickets were another charge on a credit card when I don’t know when my next gig will be. I’ve started looking at jobs in other similar fields but few industries seem more promising right now. AI might not be able to do our jobs well, but it sure doesn’t stop upper management from trying to make it work.
On the car ride to the airport, my mother said, “You know, I keep imagining you writing a book and just self-publishing it on Amazon, selling it for a reasonable price and making a little living that way. I’ve seen so many people do that.”
“I would like to do more creative writing,” I replied while I thought about the hundreds of hours I would put into writing a book I loved only to toss it into the abyss of Amazon’s book marketplace currently drowning in self-published AI slop. “Maybe.”
