[newer_vienna]

Article is full of AI tells. "The two men shared surface-level similarities.", "Not X but Y", and em-dashes everywhere. I wish that people would write articles themselves, with their own style, if they expect people to read it.

[tptacek]

I doubt it. It's the house magazine of a a Christian sect (the Bruderhof Anabaptists), and it also needs a firmer editor. There were sections that stuck out to me as I read it where I was like "Claude would have caught that".

I wish people would stop keying in on em-dashes. They might be a tell on message boards and Twitter, but lots of writers use them heavily and have for decades.

[newer_vienna]

By itself it's not a tell but combined with all else it's hard to pass by. Author's other article from 2025 has less than half the dashes and it's the same length

[panflute]

How would the rise of dash usage in LLMs have arised if a significant portion of non-LLM writers weren't inclined to take them up and make them more common? The only explanation I see is that they are common in training materials we don't as commonly consume as website visitors.

[phainopepla2]

I have often wondered this myself, especially because the same stylistic quirks are found across models from different labs.

I haven't found a satisfactory explanation, but whatever the explanation is, it is undoubtedly true that LLMs use them to an almost absurd extent compared to the vast majority of human writers. Anyone who reads a lot of prose can see that.

[pegasus]

It all falls into one overarching category: style over substance, quantity over quality. Em-dashes are a simple way to sound important, same for generic throat-clearing phrases like “It’s important to note that…” plus they puff up the text without saying anything, same for overly symmetrical structure like neat triplets parallel clauses, and balanced pros/cons even when the topic does not naturally call for them, etc.

And they don't have a sense of stylistic restraint, so they often go overboard with one or more of the above.

[Delk]

In my (admittedly limited) experience, a verbose and elaborate writing style is also traditionally more common in humanities whereas scientific or technical writing favours a rather more terse and matter-of-fact style.

I don't know about the structures you mention specifically but if you compare an article on humanities or social matters against the style that's common in science and technical writing, chances are it's going to look more verbose in any case.

I don't necessarily have the best AI-dar but TFA didn't ring any LLM bells to me.

[NoMoreNicksLeft]

So we're actually witnessing in real time that he was slowly learning where to use emdashes? That's sort of hilarious.

[jjtheblunt]

to your point, a book i had as a kid in chicago suburbia has a section on hyphens and dashes

https://www.chicagomanualofstyle.org/book/ed18/part2/ch06/to...

[righthand]

It’s not just emdashes it’s emdashes coupled with everything else that’s a tell. Only marketing has been using “it’s not X, it’s Y” and not good non/fiction writing. People should be keying in to help others discern generative text, regardless of however annoying you find it.

The identifying and complaining of LLM generated writing is just desserts IMO of all the LLM evangelism going on.

[tptacek]

Just so I'm clear, I'm saying I don't think the writing in this is coherent enough to be LLM product. It kind of meanders and there are some rough paragraphs.

(That's not a bad thing! I'm not saying it wasn't worth reading. Just that it had rough edges that in my experience LLMs polish off.)

[badlibrarian]

At a minimum, I do see a lot of AI-as-researcher tells here. You can get Claude to draft very similar essays (of surprisingly quality) if you feed it a target market/philosophy, a few articles for style, then ask it to dig up dirt on any published author in the humanities. It connects the dots and writes stuff that feels just like this article, right down to the meandering. The rough edges and sudden shifts in register is the author editing, then asking for a revised draft.

Claude says: "Verdict: Heavily assisted, possibly lightly edited from an LLM draft. The primary sources are real and the Kierkegaard scholarship is accurate, which suggests a human who knows the material. But the connective tissue and virtually all the 'writerly' prose is machine-generated."

[tptacek]

Yeah I don't believe Claude's take on these kinds of questions at all. I can get Claude to say that about posts I wrote 10 years ago.

[badlibrarian]

I've written essays in this exact format and I recognize specific tells. He's using Claude Sonnet 4.6 Pro (now Adaptive) as a research assistant then tweaking the output. Know it, done it, smell it.

"The piece moves in a pattern that LLMs default to: historical episode, philosophical summary, contemporary relevance, theological application. Each section is self-contained, cleanly closed, and bridges to the next with a meta-sentence. A human essayist leaves more mess in the transitions."

Now that I've pointed it out, you'll see more stuff like this. It's everywhere.

[pasquinelli]

> Only marketing has been using “it’s not X, it’s Y”

i'm not even remotely convinced that's true.

[sekh60]

It's not true, it's false.

[nonameiguess]

Unless I'm misunderstanding something about the font, these seem to be the shorter en-dashes, not the em-dashes that are otherwise rare to see.

Also, there is the question of why? This is a quarterly publication with only a few articles, not a blog spamming 20,000 a day. The author himself is a rabbi and professor at St. John's, who is heavily published but not exactly spamming the world with shit. He's written two full-length books, one novel and one non-fiction, both of them published before LLMs were anywhere near good enough to produce convincing long-form prose. All of his material I could find is published through real publications with editorial boards, not self-published. He doesn't exactly fit the profile of the ambitious hustler trying to make a name for himself to game SEO rankings or boost his karma on web outlets with up-voting mechanisms.

[pasquinelli]

i would expect emdashes in a professionally published website.

[jermaustin1]

Same. I believe Word and most other word processors and desktop publishing applications convert standard keyboard-typed hyphens to em- or en-dashes automatically, and have for decades.

[jknoepfler]

particularly academic writing... said having worked as an editor at an academic journal long before ChatGPT was a thing, and having corrected many hyphens to m-dashes.

[newaccountman2]

Did you go to law school?

[settsu]

If only there was a way to find out the truth. But who has the appetite for that these days? Or the appetite for the effort required?

The irony of this comment can even be found in the post itself:

> ...the magazine’s fortunes soared by exploiting the public’s appetite for outrage. Articles frequently relied on exaggerated – and at times outright false – stories... Accuracy and integrity were secondary to the relentless churn of opinions. The formula worked.

[foxglacier]

You people should work out why you need to know if something was AI written or not. If there's really no way to know the truth, then the truth can't have any impact on you, so it doesn't matter. Why then do you care?

I've heard people say they want a human connection with the author but there was never one anyway. It's 1-way (parasocial), repeatedly edited (not natural human thought), formulaic (effectively AI writing rules implemented by a human), sometimes written by multiple separate people, and you have no other interactions with the same individual(s) so you can't build any kind of relationship or coherent understanding of them.

Consider me. You've probably never interacted with me before and probably never will again. I might be two separate people. I might be an AI. This might be a copy-paste of something I already wrote to 10 other people. Will knowing any of that stuff make a difference to you?

[newer_vienna]

Yes, for communication to hold any meaning there must be persons at either end of it working towards a common understanding of truth.

Read through Borges' Library of Babylon, it's brief and follows the readers in the infinite library of seemingly random text. I read the lesson to be this: The protagonist will not find meaning even if he finds a coherent book that walks through a philosophy of existence, rather it is by coming to terms with the design of the architects of the library that any hope at conversation can be had.

[foxglacier]

> for communication to hold any meaning there must be persons at either end of it working towards a common understanding of truth.

I'm really struggling to understand that. Are you defining the meaning of the word "communication"? Does "either" mean "both" or "at least one"? Does reading a blog post count as communication? It it just a fancy way of saying "words are only useful when they convey meaning"? How does any of that relate to AI? AI can also learn from communication. Do you exclude AI because you hold the controversial belief that it's not to be capable of understanding or perhaps not to be capable of working towards any goal? And how about humans who intentionally write nonsense?

I also don't see how the story of that library is relevant. This text is not seemingly random, so it's more like walking through an actual library.

[newer_vienna]

I mean "either" as "both". Yes, reading a blog is communication. And you don't need to convince me that plenty of people go about life steeped in nonsense and that listening to what they have to say is a waste of time. But on the topic of existentialism, humans can grasp at truth through lived experience and this produces both a reason for and a means towards purposeful interaction with others.

LLMs at their root are next-word predictors. If there's any communicative value in what they produce, it is due only to the data they were trained on and the intentions of the prompt-director and publisher. I have no problem in saying that I would rather interact with the words from the source than with the machine-generated resultant text.

[foxglacier]

Thanks for clarifying. I think we just differ on what we value. I don't mind communicating with someone even when there's no hope of them gaining any understanding from our interaction. I'm happy to read a technical manual that may have been written by 100's of different individuals of many decades, many of whom aren't even alive anymore. That's the same one-way communication I get from a blog or news article. Who or what the author is is irrelevant to me.

Do you mean you'd rather read the prompt than the output? That's tantalizing but it's only possible because they used AI. I think regular journalists and bloggers effectively have a secret prompt in their head and generate an article to respond to it. Don't you feel the same way about that? It's not AI vs human, but seeing behind the scenes vs seeing the product of the work. Also, you probably don't want to see how the sausage is made. It might look like "here's a bunch of dense technical PDFs about resource use permits and lab reports. Write an article that makes Tesla look like they did something wrong". That might be the exact same secret prompt a human journalist uses, so why do you value the human's output more than the AI's? The human certainly isn't trying to gain any understanding - they're trying to rile up their readers.

[ofcourseyoudo]

84% likely human on zerogpt, but you could have done that yourself.

[goolz]

ZeroGPT is a gimmick. Just last month it flagged my paper as AI and I wrote the thing myself. How is it coming up with that 84%? Seems like snakeoil to me. Even the academic department at my Uni agreed and admitted they cannot use any of these AI checkers in actual academic hearings. They are akin to dowsing rods.

[kevin_thibedeau]

Hallucinated references are a 100% tell.

[goolz]

Definitely. In my case, I made all mine and they were real. There will always be obvious tells. Without those though it becomes opaque.