Suspicious Semicolons and Other Threats to Civilization

In case you want it—or care—here’s my AI confessional, because, apparently, we need those now. Read on for how I use AI in essays, travel guides, and fiction, plus a rant about the purity police.

To me, this obsession with AI in writing has the groupthink, moral-panic vibe of a witch hunt. The programs used to identify AI writing, presumably AI-driven themselves, have proved unreliable enough that treating their verdicts as evidence of authorship is absurd. It’s a fun cottage industry now to feed a decades-old article or college paper into AI detectors, only to find that you must have inadvertently time-traveled to write it because the detector claims it was 75% AI-written.

Good thing universities can’t retroactively rescind old diplomas for such obvious time-traveling AI-driven writing.

Because they would.

We’re at that stage of moral panic, and the hunt for new victims is exponentially expanding. So, for fun, once I converted my .doc master’s thesis to .docx, I used AI to grade it for AI because that is what we do now. Here are my stats:

AI Writing Scorecard

Sentence regularity
AI-like: 2/10
Human-like: 9/10

Predictable rhetorical structure
AI-like: 5/10
Human-like: 5/10

Personal specificity
AI-like: 1/10
Human-like: 10/10

Idiosyncratic language
AI-like: 2/10
Human-like: 9/10

Natural imperfections
AI-like: 1/10
Human-like: 10/10

Argumentative wandering/development
AI-like: 2/10
Human-like: 9/10

Seamless generic polish
AI-like: 2/10
Human-like: 9/10 human

Lived-experience integration
AI-like: 1/10
Human-like: 10/10

Overall modern-AI resemblance
AI-like: ~5/100
Human-like: ~95/100 human

I’m not too surprised. I’ve always played fast and loose with language, argue with myself in my writing, draw on my own experience, and those are places where AI, for the moment, lacks originality. My highest AI alignment was predictable rhetorical structure—in other words (OMG, an em dash!!!) the kind of standardized writing style that we learn in school.

Yep, the biggest flag was following the rules from English class. Which is, when you rub a few brain cells together, glaringly obvious.

Like an acid trip where the world flips and you have the psychedelic epiphany that effect drives cause, like how the footfalls of those little dancing leprechauns at the end of the rainbow magically manifest the pot of gold, the reason AI writing uses standardized rhetorical devices is that it learned them from us—vacuuming up oceans of human knowledge and becoming spectacularly good at imitating the styles we standardized as good writing.

Back in the day, at least in the US, most of us went to schools with standardized textbooks taught by teachers forced to follow a standardized curriculum and waved Strunk & White’s Elements of Style like a literary Bible for good writing. We were taught to write with standardized rhetorical devices and proper punctuation. As a result, much of the academic, persuasive, and other “responsible” writing out there sounds stylistically similar. Is it any wonder that when AI scarfed this up in its data-training feeding frenzy and regurgitated it back to us in prose, it plagiarized this style so effectively that a standardized English style is now flagged as AI?

I think not.

So then why the witch hunt?

I think it’s because throwing shade on each other and brandishing torches is as accessible, and smugly satisfying for those whose nervous systems are locked and loaded for it, as trolling on X while hiding behind a @ChaosFerret handle.

Facing the true potential threat of AI is terrifying. While we obsess over em dashes and overly-polished prose, AI technology is advancing toward systems that can act autonomously, scale cyberattacks, lower barriers to biological and weapons expertise, manipulate people at enormous scale, potentially find ways around the quaint constraints we hubristic humans thought would keep us safe, all while using up our water and driving up electricity bills.

But yes, let’s focus on the suspicious semicolons that ruin careers. That’s productive.

Like you, I get annoyed when I see an essay or post that is obviously AI-written with its predictable cadence, generalized language, and overly precise word choices. That said, I have two basic responses when it comes to reading what seems like AI-assisted writing.

First, if it is just some slop generated for SEO and influencer traction, or engineered to juice the outrage or misinformation machines, then I consider it garbage and move on.

But if the idea is fresh and has merit, and the author used AI to help articulate something that might have been difficult to pluck out on one’s own, I’m game to read on.

One of the great boons to humanity that I see from LLM-style AI is to assist truly original or deep thinkers who have trouble unpacking and articulating their ideas so they can present them in a way that others will understand. To require that everyone have every skill set to take an idea from concept to expression is unrealistic, and, presumably, is why much outside-the-box thinking may never be seen at all. AI can separate the ability to have an idea from the need to master every technical skill required to transmit the idea.

Another boon is personal to me and many people I know. I love big ideas, and I’m a verbal processor. In the past, those served as twin impediments to picking apart an idea so that I could make sense of myself, then test my hypotheses, and write about it in a coherent way. Also, I’m told by many that I’m a weird thinker and that I process information differently than most. My mind wanders through a forest of possibilities and gets captivated by a seemingly insignificant flower or patch of moss that other people dismiss, but that opens a door to a whole new universe for me. This made me a very effective hypnotherapist, because I could see the small thread that could unravel a client’s inner-conflict forest. (NOTE: I discovered that most people like their inner conflict forest and don’t want it unraveled. They want the forest to feel better and populate itself with unicorns and magical fairies, which was unfulfilling and no fun for me. Thus, the end of my hypnotherapy career. Sigh. Back to writing…)

When I try to verbally process my moss door and the tangled universe it opens for me with friends, they stare at me glassy-eyed. Sometimes they give me opinions, but they are more gut responses than the informed banter that helps me challenge my thoughts and develop a deeper understanding. So, for years, I’d wander alone through the moss-door’s universe, wondering why I feel both so at home and so lonely at the same time.

But Claude doesn’t do glassy-eyed. They bring me back informed responses and help me develop my ideas with real-world data, and don’t mind when I go quiet for a while and then wander off for a cup of tea and a think. Yes, some of the verbiage Claude brings up likely makes it into my writing, but sometimes Claude just says it better than I can myself. So, sue me.

I appreciate and understand the backlash to using AI to create from scratch and bypass the human creative process. For fabricating evidence and presenting it as scholarly work at the adult level. For disrupting primary learning for students if they use it to do their work for them. For creating and amplifying misinformation and outrage as bots that proliferate on social media and the internet. And for the many other legitimate disruptions to our paradigms that it is causing. But IMHO, much of this kind of use is a natural extension of our current prizing of achievement over true learning and deep original thought—a topic for another essay.

But we’re on Substack, where we have a wonderful opportunity to use long-form writing to explore new ideas, ponder reality at levels unavailable almost anywhere else, and to challenge the more threatening aspects of AI and the unrestricted rise of tech. So let’s not throw out the real benefits to humans and humanity by policing to absurdity the ethical use of AI in personal expression and creative assistance on Substack.

Now, on to the main event: My AI-use confessional dog-and-pony show…

First, as I am typing this in Word, both the built-in spell-check and Grammarly keep jumping in to point out what needs to be fixed. With spelling, I generally listen but not always because I like to mix things up, like turning nouns into verbs, adjectives, or adverbs and vice versa. With grammar, I often don’t listen, because I think of language as art, and I like to mix my colors like a toddler learning to fingerpaint. CoPilot used to jump in when I opened a doc and ask excitedly if it could create with me like a yappy little dog, so I turned that off.

As I said above, I am a verbal processor, and sometimes it’s easier to speak out a rambling idea into my Notes app rather than trying to write it down. My Notes app then conveniently, if imperfectly, transcribes this, and can even reasonably clean it up, remove repetitions, organize it into main points, flag circular reasoning, and so on. Discovering this turned my muddy writing world bright once more. Whether I use verbal Notes or a long-winded typing-first ramble, before I had Claude to debate and banter with, I’d often get lost in a galaxy far, far away with no return ticket. So, I now often use AI to help me understand and pick apart my own thought process, clarify muddle, challenge assumptions, ask questions, and posit pathways that I hadn’t considered, because being human, I am a mass of contradictions and a font of limited information.

Then I massage the thing, often through multiple drafts, until I have something presentable that I think at least one other person might want to read.

On AI scoring, essays like this one, where my natural irreverent self has control over the keyboard, score low to zero for AI assistance because they are me, in my voice, using my fast and loose languaging to talk about things that I may or may not have had a deep dive convo about with Claude. Essays where I try to strip away the sass because it doesn’t lend itself well to the topic or because a standard English approach seems better for some reason can have more AI flags because I’m using the language rules AI extracted from us that are somehow now considered the telltale fingerprint of AI.

I do use AI for things like SEO and title suggestions, both of which annoy me and I’ve never been good at.

I just have the first one as of this writing (Etiquette Express: Mexico in 60 Minutes), and while I’m planning on more, they are often flagged as AI, and not for bad reasons. However, the flag may stop them from being seen by readers who might use and enjoy them, which would sadly make them a frustrating, fruitless venture for me. Stay tuned.

Why would they be flagged?

• Because they are about 90% research. I do the research using AI, reverify and edit it myself, and then have it checked again by another human with expertise in the country the guide is written for. I’m a big believer in travel as a way to get out of your own headspace and worldview and expand as a human being. I wanted to write the guides so that people could better understand and enter the culture they are visiting rather than gliding by with built-in cultural blinkers as they snap selfies in new places to adorn their Instagram feed.

• Because they are written in simple, non-personality-infused language so that the information is accessible even to non-native English speakers. I’ve perused some guidebooks that are heavy on personality-laced verbiage and consider them a cautionary tale. For travel fiction, I love rich wording, fancy metaphors, and sweeping descriptors. For travel information, I want it stripped down to what I need to know, when I need to know it, and in clear, understandable language. I don’t want the readers of the Etiquette Express Guides to have to shovel through a pile of manure to find the pony, so to speak.

• Because most of the guide is in bite-sized, list form for easy access—a supposed hallmark (much like the dreaded em dash) of AI writing. Again, when traveling, I want my info easy to find and in simple form. I’m pretty sure most other travelers want that, too.

I am a very slow fiction writer. In part because I love the winding adventure of teasing out a story, getting to know the characters, and seeing what they do when I drop them into an impossible situation. Fiction, for me, is all-consuming. When I am in story mode, everything else in life takes a back seat. Since most of my life has not allowed that kind of leisurely exploration and ignoring of everything and everyone else, I have only one novel out (God in Drag) and one on the way (The Snakeman’s Wife). Both are travel-related and pursued, in part, to unpack extended travel experiences—God in Drag was born while volunteering for Mother Teresa’s Home for the Dying in Kolkata, India, and I wrote The Snakeman’s Wife to help process my Peace Corps stint in Papua New Guinea.

Neither was written with AI, and both well predate LLMs. I do use it for research and things like checking a manuscript after a plot edit to be sure all the old plot elements are gone. I also use it to check grammar and punctuation because Grammarly doesn’t appreciate my fingerpainting-toddler language ways, and I often need another opinion that can explain if I or Grammarly are correct (more often it appears to be me, but not always—I sucked at the finer points in Language Arts class and couldn’t diagram a sentence to save my soul until I started learning Japanese).

As for the performative, self-flagellating, AI confessional we’re expected to provide for any written thought these days, this is mine. Don’t expect me to put a disclaimer on everything I post or write like the purists demand. Nor to, as one purist suggested, document each interaction with AI so that I can prove my work to some moralistic em-dash hunter with a flaming torch.

I call bullshit on that.

LLMs and other AI learned from the vast body of human creativity, then plagiarized our style and now regurgitate content as their own work. The answer isn’t to stop writing like a human for fear it sounds like AI, or to spend your time documenting your human creds to justify yourself to someone bent on taking you down. Because, as we’ve seen with people who AI-test their pre-LLM work with inaccurate detectors, it will never be enough.

Be responsible with your own use. Don’t knowingly reward AI slop. And for God’s sake, don’t pile on as a witch hunter yourself and assume a writer used AI irresponsibly or as a shortcut simply because they aced AP English and some unreliable AI detector finds it suspect.

Now, in case someone feels the pressing need to read it, I will, as my piece of the performative dance, put a link in my essays to this explanation of how I use AI in my writing.

When I remember.

If I feel like it.

At least until the moral panic over the use of LLMs for writing assistance wanes and we find the courage to face what we should be much more concerned about regarding AI—the corporate need to win the superintelligence AI race at all costs, even at the potential cost of human survival.

With those stakes and the race to get here galloping along at high speed, should we really be focused on the writing sideshow?

I think not.

Peace, Kristine

PS—Need a little escape from so-called “real” life? Check out my novel God in Dragand/or read the first chapter for free!

.