I do some technical editing, a service I do not vocally market here for various reasons. In the last couple of years, I’ve worked over a lot of AI-generated writing. As you might imagine, it’s the topic au courant in editors’ groups. Small wonder, when it touches much of what we do.

For some readers, the question is more basic: How do I even know whether or not it’s AI? (All right. Others are asking why they should even care, but they probably did not read past the title of a blog post covering what they don’t care about. Thus, they probably are not the audience here.)

AI is a clever machine with its uses, but like the so-called smartphone, it’s not that smart. There is a certain abstract connectivity among concepts known to humanity, an unconscious reality checking. This is absent with AI, allowing it to wend its way into strange areas with oddly irrelevant paragraphs or concepts. If the writing suddenly seems to have wandered into weirdness, kind of like when I tell a story for which only I can see the relevance, and thus baffle my listener.

There is a sort of fear-and-cowering around the em dash–in order to simulate it here, I must use two hyphens, like this–on the thinking that if you use a single one, everyone will think you wrote it with AI. Very common for those coming out of college, which I suspect holds the trenches of this long battle between plagiarism/cheating and original writing that demonstrates material learned. I have seen it with younger consultants’ tech writing and in discussions, as if Prof. Fussbudget might catch them out even in the private sector. I do not think you should cower over em dashes because if you use them sparingly, they’re a great tool. If you use them several times per page, they’re bad writing. And yes, that can also be an AI tell. AI won’t say “open you’re book to page x,” but it doesn’t know how to camouflage itself very well.

One hallmark I have seen of AI composition is prose too technically perfect and very sterile for the most part, lacking any sense of the person behind it (of course, there was none). It does use passive voice when it should not. Now and then the phrases are in the wrong order, because that’s problematic for AI to reality check. Our language is so difficult to master that it can/does trip up AI.

There are AI checking tools, but the consensus is they have a high error rate. My own checking tool is more basic, though it took six decades to acquire: Common sense paired with a long grounding in the language, associated with some critical thinking. I know that certain sources are very likely to use AI. If I think the author is under 30, to be honest, I begin with that assumption and seek to disprove it; I am seeing too many young workforce entrants who obviously never learned to think and express. (I do not fault them. This is the world we bequeathed them.) If the text comes from a corporation, I assume it’s AI but I don’t care because they always wrote that way before. If it’s marketing copy, same: likely AI, but I don’t care. I’m the one always trying to avoid letting my eyes apprehend marketing copy, because their objective is to imprint upon my brain and that’s not their province.

So first comes the question: Do I care enough to analyze it? Usually the answer is no, because I expected no better. If I do care, then I look for writing that is too technically perfect for the modern world. There’s a gas station near me with a big proud sign about their status as an eco-gas station, with text including the word ‘compatability.’ Yes. It got all the way through whatever marketing department without someone even spellchecking it; it’s been hanging there five years or more.

AI would not have written that. I never thought bad English would be a helpful tell, but it can be. Some years back, a new cheap cheep cheap grocery store opened up not far from me, and the marquee at opening said: YOUR TO GOOD LOOKING TO SHOP ANYWHERE ELSE. (How will I boycott you? Let me count the ways.) But say this for that lop-eared dumbness: It was fully original.

In a society of decreasing literacy, quality original composition English is not so common. Sterile quality composition sparks fundamental suspicion. And while I’m looking for the small errors that suggest I can take the thoughts seriously, I’m also reality checking. I’m waiting for the moments where AI seems to stand for Alzheimer’s Intelligence.

There’s a certain semi-incel movement among men who can’t quite adapt to what women want in a male partner, one that if it had its way would simply have a fembot. It (that’s not “she”; a she is a living female being) would be a pliant servant lacking mood swings, desires to buy new accessories, or stray body hair. It would prepare foods, provide sex, and refrain from challenging featherbrained perceptions. It would never fart, never want to repaint the living room. It would have stereotypically perfect looks that would never age. Some would even want it to appear as an age that would, in a living human, make sexual contact with them a felony. But it’s very difficult to imagine that a fembot would not get confused, or might bring the wrong thing, or be unable to throw that deadbolt because it’s touchy and has to be done in a certain way. What would not confuse you for a second could baffle this imaginary fembot. It could not reality check, any more than it could say “Hey, fuckhead, go mow the lawn.”

AI reminds me of a fembot. It’s a tool for those too lazy to write their own words, or those too busy to produce enough words…or those who never learned to begin with. You don’t have to be nice to it when you’re crabby. But at some point I think we will see a majority of the young workforce simply quit if they are told they cannot let AI write for them. They’ll have to. They’ll never have learned to write their own words. Take away the bot, and they’re done for.

That said, there’s opportunity there for my kind. I wonder what percentage of today’s self-published novels are written mostly with AI? I have good reason to believe it’s true of a great deal of business writing. It does carry a stigma, and it takes an authentic editor to de-AI writing. If that matters to people, they might end up engaging one of us to make it sound like the au did actual work.

I wonder.

Speaking of generational dumbdown and random segues, here’s one. This isn’t AI-trained kids; this is grown educated adults who have worked in intelligence, science, government, and the military. Some have doctorates.

When was the collective decision made that ‘nucular’ was an acceptable articulation for ‘nuclear’?

I do remember when a certain government executive, one dumber than most, started saying it that way. My theory is that people collectively made the wrong decision there.

Right reaction: “You goddamned cretin, you pollute the air with every ass-ignorant thing you blurt. Not only no but holy hell no.”

Wrong reaction: “Oh. Stupid isn’t so bad. Let’s embrace stupid!”

As usual throughout our history, we had the wrong reaction. Unfortunately, this one is word herpes. There’s no way to eradicate it. Free speech laws will not, for example, make it legal to junkpunch anyone pronouncing it ignorantly.

Within one more generation, no one will ever know it was ever considered a mark of ignorance.