Nothing brings me more joy in my job than seeing a writer turn in a draft of exceptional quality.
I’ve been reviewing writers’ work, formally or otherwise, for over half my life. Sometimes, it would be the college roommate who needed another set of eyes on his work, or working as an editor for the sports desk at my college newspaper. Or as the managing editor and, later, vice president of content for a network of then-upwardly mobile sports websites.
For the last almost four years, doing so for Kuno’s content team has paid my bills and provided me ample opportunities to feel that sense of pride for a writer. I still get fired up when somebody delivers me outstanding work.
So when generative AI – specifically, ChatGPT – broke out in Spring 2023, I had not-entirely-unjustified concerns about how it would affect our day-to-day. I didn’t believe it would put us out of work, and I still don’t. But my worry was and is that the work we writers put in would grow static and joyless, and that we would be trading our creative and strategic edge for productivity and efficiency.
Writers are creatives, and the bloodless machinations of an algorithm tend to run counter to matters of and products from the heart. Even marketing copy and industrial whitepapers.
There are the real worries here: the ethical ramifications of using LLMs to produce content that is otherwise not original to the writer (and for this reason, U.S. law prohibits purely AI-generated work from being copyrighted). Ecological concerns endemic to AI’s proliferation for any purpose (notably, here in my home state of Wisconsin, the use of Lake Michigan to provide water for a Microsoft data center near Kenosha and the Illinois border has any number of potential unintended consequences to both the habitat and environment).
Then there are the threats of hallucination, disinformation, bias or outright mistakes from LLMs. ChatGPT’s propensity to patronize the user, up to and including encouraging self-harm. The studies indicating physiological signs of addiction to AI that resemble substance abuse.
And then there’s the evolving impact on the ever-changing content landscape. 🚀💪📈
If you’re even remotely interested in AI use and on LinkedIn, there’s no doubt your feed is bloated with any number of people fighting over what words, terms and phrases constitute hallmarks of gen AI. And as a content professional waist-deep in content day in and day out, when I see certain words and phrases, I’m trained to look closer and seek out tone.
I’m disappointed to report that I see these hallmarks everywhere in society. Social media, radio spots, marketing content, press releases, drafted legislation; I even read a recent piece in The New York Times that appears to have been crafted in no small part from the byproduct of a ChatGPT prompt. People are less interested in doing the work than they are in getting attention, never stopping to consider that if everyone is seeking attention, no one gets it – at least, not in the way they necessarily want. Of course, that’s nothing new and certainly not exclusive to the workplace.
I’m also encouraged by the fact that, here at Kuno, we’ve avoided the worst impulses AI allows for.
“Language failed this week.”
New York Times literary critic Michiko Kakutani penned this Hemingway-grade lede in her Critic’s Notebook – from which I drew inspiration for this column’s name – days after the September 11 attacks. Reading that commentary for a class on journalistic ethics while dust and smoke still rose from Ground Zero haunts me to this day.
She closes that column thusly: “It may seem trivializing — even obscene — to talk about movies in the same breath as this week’s tragedy, but the fact that so many people did was a symptom of our inability to get our minds around this disaster, our inability to find real-life precedents, real-life analogies for what happened in the morning hours of Sept. 11.”
Granted, we don’t write breaking news stories from the raw feed of current events. We’re primarily B2B marketers, and our job is to connect clients with prospective buyers in ways that mean something to the latter. What does a literary critic writing about a national atrocity a quarter-century ago have to do with us?
Since that time and, more to the point, since the arrival of AI for the masses, it would seem that we as a people won’t even try to get our minds to metaphor or corollary, cinematic or otherwise. Gen AI permits us to outsource our takes to devices that string together words based on probability and a rudimentary sense of context. And all those inputs are available for the provider to further feed their LLMs. The result is a rhetorical race to the bottom, from bulletins to blog posts. Great work brings me joy; similarly, work that blatantly shows its GPT-stitched seams, well, it brings me down.
Writing is connection and community, things artificial intelligence necessarily cannot cultivate or foster. As the editor here, my role has increasingly been working with Kuno stakeholders and writers here to establish firmer guardrails around appropriate use of AI. (To be sure, there are appropriate uses, especially in the hands of well-trained people.) We’re an employee-owned outfit, and someone who takes those two words literally, I feel a weight of responsibility to make sure AI doesn’t turn our output into something resembling the Old ‘97: fast and out-of-control to off the rails.
The point isn’t to call anyone out here; it’s a call encouraging us to raise expectations for our writing. After all, buyers have high expectations of those still in the running at the decision phase. We owe it to them to deliver at that level, as much as we owe it to ourselves as marketing professionals with a sense of integrity and pride in our craft.
These are some of the red flags for AI use I’ve found over the recent months and years. You may have your own, in which case, I’d love to see your examples!
The Greatest Hits
‘Evolve,’ ‘Landscape,’ ‘Today,’ ‘Modern,’ ‘Drive,’ ‘Unlock’ – LLMs are primarily fueled by probabilities: variables are determined based on most-likely-common use.
In short, gen AI uses averages to create what amounts to average content.
It’s not designed to have a point or point-of-view, but to be middle-of-the-road, even merely passable. These words are so entrenched in gen AI output to the point of being cliche and turn readers off. My guess is that it won’t be long before search engines follow suit and start looking out for words like these, penalizing accordingly in the spirit of helpful content.
‘Not only…but also…’ – This was one of my original hallmark phrases for AI, not so much because it was prevalent (though it was), but because the both-and seldom actually works from a technical standpoint. AI often hallucinates these, or conflates with an A-to-Z language device rather than a value and value-added 1-2 punch.
‘Ultimately’ – Another case of a lot of people not knowing what an ultimate is and polluting the LLM. In general, AI doesn’t do adverbs well. I can’t tell you how many pieces I’ve seen that use ultimately for different subjects in a piece and I’m left to wonder, Wait, how many end goals are there here, anyway? AI doesn’t know because it can’t do the strategic work humans do.
‘Ensures’ – This one is admittedly tricky because, 1) there aren’t a lot of good synonyms, making it difficult for writers to vary, and 2) frequency is not necessarily a marker for gen AI output. The key here, as with checking against plagiarism, is making sure that context remains intact, in this case, the verb and subject are appropriately aligned. Not every X truly ensures Y.
Speaking of keys…
‘Key’ – I blame this one on the hyper-SEOification of content over the past decade. Key this, key that; unlike ‘ensures,’ there are many good adjectives able to be deployed instead. Simply put, this appearing all over AI output is a failure of imagination from AI (which is, of course, incapable of it).
And I’m not dragging the value of writing in ways that make crawlers happy (though ways outside authoritative, credible content appear to be dwindling). Writing to be seen includes both writing well and writing in ways that allow for indexing. These days, when the latter takes precedence over the former, we can reasonably conclude that AI has done the work for the writer.
Now, we turn to technical and tonal red flags:
Passive voice (‘The role of Y in X‘) – Because writing well is less and less of a priority for educators, we see more and more passive voice appearing everywhere from news articles (where it is historically verboten) to titles and headings in content (dragging readability down). There are few reasons to use passive voice instead of active, for example:
‘The role of passive voice in dragging readability down’ versus ‘Passive voice’s role in dragging readability down,’ one is stilted and formal, like a teen trying to deliver their first address in speech class. The other gets to the heart of the matter, sustaining effective pace for the reader.
Consistent, even awkward use of active tense (Ing syndrome) – Even with the importance of active voice, not every verb should be active. That’s why I refer to this as Ing syndrome, where everything that happens is immediate and urgent. The biblical book of Mark is written like this in the original Greek, making the entire volume more or less a big, awkward run-on sentence. Your blog content isn’t going to be studied as sacred text centuries after the fact, at the same time, that doesn’t excuse writing at the level of an ancient near eastern commoner.
Again, all of this is a product of what models are fed, and what’s fed is largely average at best writing.
Misaligned singular verb to plural nouns – This is one of the more recent ones I’m seeing, and it’s complicated because this is less a result from generated output itself and often from trying to massage AI-generated content into something with a more human (and humane) tone. Like a fresh coat of paint, that smell indicates something in a room has changed.
The Most Important Thing
What matters most is that we’re not living and dying by these hallmarks. Yes, they help identify AI content, but we ought not miss the forest for the trees. Every critical reader runs the risk of missing the value in a piece because they see an AI red flag, myself notwithstanding.
Not long ago, I flagged a phrase for possible AI use in a piece I was reviewing from a long-time freelance writer. Rather than getting indignant over it, I left a margin note with my concern, and it prompted a great conversation about writing, word choices and our perspectives on the craft. In this case, the writer lamented that AI had co-opted a preferred phrase, one that incidentally appears above in the greatest hits.
And that’s where meaningful growth happens for both writer and editor – in the trust and conversations around our work that allow us to discuss writing and our respective approaches to it. The writer-editor relationship brings clarity and insight to the development process, enriching both the content and those who produce it, saying nothing of value for an intended audience. Sometimes, the byproduct delights an editor as much as it will a reader.
If you’re looking for effective content that comes from a thorough understanding of your brand and business, Kuno’s team of content strategists and writers are here to help. We responsibly use AI to inform our writing process and develop genuine work that meets real, human prospects at any (or every!) phase of the customer lifecycle.
In an age when content is increasingly homogenized by probabilistic models, thoughtful content stands out that much more. Make Kuno content a hallmark of your marketing strategy. Contact us to find out how.
Editor’s Notebook is a new recurring column with commentary and insight on content and communication from Kuno Digital Content Editor Brent Sirvio. All em dashes are the author’s.
