Do we need a term (probably German) for the anxiety that one's work might look like it was generated by machines?
-
I heard from someone near me today that to make your work appear less like it was machine-generated the emerging rule is that you should not use the 'em dash', nor write in paragraphs, rather one text block.
I have prior heard another say that text summaries at the end of an article are seen as indication of genAI use, as is text free of typos.
Has anyone heard of other references to behaviour-change born of such anxiety?
This article shared by @slackline has some strongly-related fight in it https://www.theringer.com/2025/08/20/pop-culture/em-dash-use-ai-artificial-intelligence-chatgpt-google-gemini
-
@catileptic What an insightful take, thank you. I had not considered this as a marker/indicator. This makes sense.
@JulianOliver if you'd like to read a bit about the difference between metonymy and metaphor (which are, at their core, concepts from within literature critique and theory), but applied to technologies (like search engines), i highly recommend Alfie Bown's book 'Dream Lovers', chapter 4 (The Match: Metaphor vs Metonymy). you don't need to read the rest of the chapters in order to get something from this one
(but it's also a genuinely good book!) -
@JulianOliver if you'd like to read a bit about the difference between metonymy and metaphor (which are, at their core, concepts from within literature critique and theory), but applied to technologies (like search engines), i highly recommend Alfie Bown's book 'Dream Lovers', chapter 4 (The Match: Metaphor vs Metonymy). you don't need to read the rest of the chapters in order to get something from this one
(but it's also a genuinely good book!)@JulianOliver and if i can push this forward one more time:
in literature, legibility isn't a necessity. neither is coherence. and neither is a complete match between form and content. this is what made James Joyce, Pynchon, Heamingway great (to name the ones that most people from different countries might have come across, because we surely have our own examples in local literature)
genAI algorithms has no concept of form versus content. and thus can not create a divergence between the two
-
@JulianOliver if you'd like to read a bit about the difference between metonymy and metaphor (which are, at their core, concepts from within literature critique and theory), but applied to technologies (like search engines), i highly recommend Alfie Bown's book 'Dream Lovers', chapter 4 (The Match: Metaphor vs Metonymy). you don't need to read the rest of the chapters in order to get something from this one
(but it's also a genuinely good book!)@catileptic Thanks a lot. I'll look for this!
-
@JulianOliver and if i can push this forward one more time:
in literature, legibility isn't a necessity. neither is coherence. and neither is a complete match between form and content. this is what made James Joyce, Pynchon, Heamingway great (to name the ones that most people from different countries might have come across, because we surely have our own examples in local literature)
genAI algorithms has no concept of form versus content. and thus can not create a divergence between the two
@catileptic This is also insightful, and I must say encouraging.
-
@njr In agreement. I am opposed to any behavioural modification as a function of this would-be anxiety, just interested in it as a socio-technical symptom of our times.
@JulianOliver Sure. Agree with that.
-
I heard from someone near me today that to make your work appear less like it was machine-generated the emerging rule is that you should not use the 'em dash', nor write in paragraphs, rather one text block.
I have prior heard another say that text summaries at the end of an article are seen as indication of genAI use, as is text free of typos.
Has anyone heard of other references to behaviour-change born of such anxiety?
@JulianOliver Oh I added em-dashes to my writing to mess with people who use poor heuristics to detect AI use
-
@JulianOliver the construction that goes "it isn't only X, it's also Y" is something i've come to suspect as a mark of ai-generated stuff, in both english and romanian
in a more abstract sense, semanticly homogenous language is how i explain ai content to myself. people can (and sometimes do) draw paraleles between semantic domains based on subjective interpretarion. this is metonymy. a LLM can't replicate metonymy, only simile (metaphor) ("x is like y")
@catileptic @JulianOliver Yeah, I feel you there. I do a double-take whenever I see that construct on pre-2022 media.
In general anything AI does it does because it picks up patterns from human text. So there's no One feature that makes it AI, more like a cluster of features that comes across as a tone or a style. It's hard to describe. I think the best way to put it is the use of writing devices for the sake of using them, rather than to convey information or illustrate a point. This can also give the text an exaggerated, almost zealous tone.
When it comes to em dashes, "it's not x it's y", etc, it's more about how it sits within the text rather than simply its presence. If it's being used sparingly to make a point, probably human. If it comes across as weird and forced and overdone, probably AI.
-
I heard from someone near me today that to make your work appear less like it was machine-generated the emerging rule is that you should not use the 'em dash', nor write in paragraphs, rather one text block.
I have prior heard another say that text summaries at the end of an article are seen as indication of genAI use, as is text free of typos.
Has anyone heard of other references to behaviour-change born of such anxiety?
@JulianOliver I have got Gemini or another AI solution to make grotesque errors. i ended a "chat" due to these hallucinations.
-
I heard from someone near me today that to make your work appear less like it was machine-generated the emerging rule is that you should not use the 'em dash', nor write in paragraphs, rather one text block.
I have prior heard another say that text summaries at the end of an article are seen as indication of genAI use, as is text free of typos.
Has anyone heard of other references to behaviour-change born of such anxiety?
@JulianOliver
If just an em-dash is enough to make your text look like AI-slop, then maybe that text wasn't really good anyway. -
Do we need a term (probably German) for the anxiety that one's work might look like it was generated by machines?
@JulianOliver LLMposter Syndrome
-
I heard from someone near me today that to make your work appear less like it was machine-generated the emerging rule is that you should not use the 'em dash', nor write in paragraphs, rather one text block.
I have prior heard another say that text summaries at the end of an article are seen as indication of genAI use, as is text free of typos.
Has anyone heard of other references to behaviour-change born of such anxiety?
@JulianOliver This is a great listen, specifically about the much-maligned em dash and the em dash backlash: https://99percentinvisible.org/episode/658-the-em-dash/
-
@catileptic This is also insightful, and I must say encouraging.
@JulianOliver @catileptic https://archive.org/stream/alfie-bown-dream-lovers/Alfie_Bown_-_Dream_Lovers_djvu.txt find on page: The Match: Metaphor vs Metonymy
+1 from me on all the tells other people are mentioning. The blandness and inability to put a point on it usually tips me off.
It’s not just boring–it’s also trite
-
@JulianOliver if you'd like to read a bit about the difference between metonymy and metaphor (which are, at their core, concepts from within literature critique and theory), but applied to technologies (like search engines), i highly recommend Alfie Bown's book 'Dream Lovers', chapter 4 (The Match: Metaphor vs Metonymy). you don't need to read the rest of the chapters in order to get something from this one
(but it's also a genuinely good book!)@catileptic @JulianOliver Thanks for the book pointer. It looks really interesting.
-
I heard from someone near me today that to make your work appear less like it was machine-generated the emerging rule is that you should not use the 'em dash', nor write in paragraphs, rather one text block.
I have prior heard another say that text summaries at the end of an article are seen as indication of genAI use, as is text free of typos.
Has anyone heard of other references to behaviour-change born of such anxiety?
@JulianOliver I have seen typos recently that I wondered about. It wouldn't surprise me if humans added more humanity as a signal of authenticity.
I wonder if something like "shoe on head" video chat authentication will also come back out of necessity.
-
@themadhatter True! We don't hear much about the parrot anymore. We should bring it back -- a useful metaphor.
@JulianOliver this one is not bad either:
-
This article shared by @slackline has some strongly-related fight in it https://www.theringer.com/2025/08/20/pop-culture/em-dash-use-ai-artificial-intelligence-chatgpt-google-gemini
@JulianOliver
> always used the em dashYeah – crowds blame humans for using what the #siliconiac had stolen.
-
@JulianOliver
I just finished a ~30-page white-paper, and afterwards ran it through Claude asking whether there were any constructions or anything that it would flag as AI-written. It caught a couple of things which were a little stilted... I re-wrote them in more direct language. So, yeah, I'd certainly rather nobody thought that my writing was AI-generated, and I'm willing to do a little extra work to try to make sure of that.@woody @JulianOliver So to avoid coming off as AI generated, you put your writing through a chatbot?
-
@JulianOliver This is a great listen, specifically about the much-maligned em dash and the em dash backlash: https://99percentinvisible.org/episode/658-the-em-dash/
@scott Thanks! I will listen to it today.
-
@woody @JulianOliver So to avoid coming off as AI generated, you put your writing through a chatbot?
@jairajdevadiga @woody Admittedly I also find the strategy perplexing.
Bill is your concern that readers might put your text through an 'AI detection' tool and deem it machine made, or that a human might read it and deem it so?
Both?
If more the latter, might not soliciting feedback from people be a more fruitful approach?
If the former, I worry Claude would be so poisoned with bias you may risk giving a genAI stink to the text where it otherwise may have none.
