While everyone is trying to sniff out AI slop, Ann Handley looks at the bigger picture and poses these two questions to consider...
The post AI Slop, Accusations, And A Better Way appeared first on Ann Handley.
A version of this story appeared in Total Annarchy, my fortnightly newsletter that helps you be a better writer, storyteller, marketer. Get it in your inbox; you’ll love it.
***
I see a social media post by someone that calls out “AI Slop identifiers.” It includes:
– the best part is…
– “it’s not X it’s Y”
– the word “honestly”
– the word “quietly”
– “that’s just x and y dressed up as Z”
– a medium-length sentence followed by short sentences like “An X. A Y. A Z.”
– the em dash
– A short question? Followed by a rhetorical question that it answers.
– “As if it…”
– “Here’s a thing no one talks about…”
PEOPLE. Come on. None of these, by themselves, are “AI slop identifiers.” They may appear in AI writing. But that does not mean they should be entered as evidence in a public trial.
That’s because all that above is simply language. AI lifted it from us, not the other way round. It’s how we speak and how we write, using everyday words and constructions. They are not gotchas. Honestly, people!
And hey. I know I get touchy about the em dash. But it’s only because I’m bone-weary of the witchy false accusations. I’m reading Ann Patchett’s new book right now (love!); she uses em dashes. Is Ann Patchett using AI? Is The Ann Patchett—one of our most celebrated artists, the winner of the 2026 Library of Congress Prize for American Fiction—secretly using Claude? Oh for God’s sake. OF COURSE NOT. Honestly, people!
AI learned all it knows from hoovering up what we all have written for generations, in books both good and bad, in both beautiful prose and the clunky text riddled with cliches and crutches. AI sucked it all up without discernment, because AI cannot discern. It mashed The Odyssey with How to Talk to Your Cat About Gun Safety with a Reddit post about fixing your dishwasher and then let it gurgle in its gut a bit before it burped it back out as “models” we can use to “write.”
Except it’s not “writing.” It’s just data. That’s not pejorative. It just is.
And by the way, “hoovering” is actually a nice way of describing what actually happened. The process itself was at times violent. Anthropic spent millions of dollars buying physical books in bulk, then used industrial hydraulic cutting machines to slice off the bindings and spines, fed the loose pages into high-speed scanners, converted the images into machine-readable text files, and then threw the paper remains straight into the garbage. Some of those books stripped for parts were reportedly rare.
If that’s not a metaphor for what’s actually going on with writing today, I don’t know what is.
* * *
I’ve been sick all week with a respiratory something so maybe that’s why my patience is about to snap. But the post about the gotcha words (before I threw my phone down in pure exasperation) is only part of it.
>>> On Substack, a machine tells you if your writing seems human enough. In July, the newsletter-slash-social platform launched an AI detection feature, via an integration with AI-detection software Pangram. People are now able to scan posts, replies, and comments on Substack to see how much of it was written by a human or with AI assistance. Pangram is often wildly wrong, in case you’re thinking there’s any upside to this whatsoever.
>>> LinkedIn just introduced an AI Slop button. Humans can push the button to snitch on other humans they suspect are using bots to pretend to sound human. Who are they snitching to? AI, of course.
>>> New tools/software features like Quillbot AI Humanizer and Grammarly AI Humanizer make AI writing seem like human writing. A robot helping another robot fake human at the request of the actual human. What a tangled web we weave.
>>> Actual writers are dropping typos and offbeat punctuation into their prose in a desperate attempt to make their own work look and feel undeniably human-crafted. Some writers are also abandoning a well-known writing construct called the Rule of 3. The Rule of 3 is a writing pattern that our brains find pleasing, which is why writers use it. But it’s a pattern. So of course AI has adopted it; pattern-recognition/replication is what machines do best. (See more on reclaiming the Rule of 3 below.)
It sounds dramatic to say that writers are increasingly in a surveillance state. But I don’t know any other way to describe it. “We’re getting to a point where people who can’t write see a bit of writing and accuse it of being AI because they can’t fathom how else that writing could have come to be,” novelist Dani Price says.
Let’s zoom out…
Tools trained on writers’ stolen writing are now being used to police whether writers are “real” enough. AI tools are prompting us to snitch on each other. We scan text for “gotcha” words and phrases merely because someone in 2026 puts them on a page. We have tools that help disguise robot-writing to sound like people-writing. Writers are adding mistakes on purpose. Legitimate writing constructs are being called to the witness stand to defend themselves by a jury that has already decided they’re guilty. I have a fever of 100.1.
Everything is just so stupid.
* * *
So what do we do.
I think we are coming at this entire issue from the wrong angle. We are locating the problem in the wrong place, like taking the batteries out of a chirping smoke alarm instead of looking for the fire.
We are trying to surveil each other and sniff out or disguise the AI instead of asking more important questions of ourselves, and what our endgame actually is.
I’m an idealist with writing, I know I am. But I’m also a pragmatist.
I know people are going to use AI to write. For all kinds of reasons, in all kinds of situations.
If you’re tempted to hand the writing over to AI, pause and consider these two questions about what this moment right now is asking of you:
These two questions can reframe “but it’s faster” and “but it’s easier” into something that has a true cost attached to it.
It’s a simple way to invite clarity, right? But it can be so hard to do.
Right away you can see this approach isn’t just about using or not using AI. It applies to any choice you can make. It’s about protecting the parts that you love and shoring up the parts that help you grow as a human being. It’s about honing your judgment and abilities in ways that matter in a thousand ways to who you are and who you want to become. It’s about choosing when not to take the shortcut.
Writing is one of the clearest choices for many of us here. But instead of writing it might be design, or code, or music—any skill where the process matters as much as the product to the person doing it. Or maybe it’s an important conversation, a relationship, a moment asking a bit more of you. Or maybe it’s something you do for no reason other than you simply love it.
That’s what I mean when I say that if you skip the struggle, you miss the growth. Sometimes the struggle is all that really matters.
* * *
My fever breaks. Everything is still stupid… but for you and me, I don’t think it has to stay that way forever.
Most of us should spend less energy trying to detect AI or not. Develop the judgment to know what not to hand over to the robots.
Choose when not to take the shortcut. Protect the parts of yourself that deserve more of you. The parts where the struggle helps you…
That’s the resistance. It’s the only way through.
That… and for pete’s sake use em dashes. They are blameless.
| # | Наименование новости | Тональность | Информативность | Дата публикации |
|---|---|---|---|---|
| 1 | Why ‘Helpful’ AI Tools Make Us Busier Still | 0 | 7.87 | 09-03-2026 |
| 2 | What AI Would Delete From Great Writing | 0 | 7.19 | 06-04-2026 |
| 3 | Email Is Dead, Long Live Email | 0 | 7.92 | 20-04-2026 |
| 4 | Why Care About The Stupid Em Dash Anyway? | 0 | 9.55 | 29-06-2026 |
| 5 | Sloponomics: who wins and loses in the AI-content flood? | 2 | 7 | 16-10-2025 |
| 6 | Is the Age of Reading Actually Over? | 0 | 6.01 | 16-07-2026 |
| 7 | The Hank Green AI Controversy | 0 | 21.28 | 10-08-2026 |
| 8 | How To Secure More Speaking Gigs | 0 | 7.27 | 27-07-2026 |
| 9 | Sorry for Ghosting: Where I’ve Been (Pssst New Book!) | 0 | 9.18 | 01-06-2026 |
| 10 | Why Claude’s Watermarking Won’t Fix Anything | 0 | 24.16 | 13-08-2026 |