today I had a thought on Ai detectors because people been talking about that Shy Girl book ban a lot this week. and so I decided to ask ChatGPT what it thought.
here is what I asked ChatGPT....
Now keep in mind mind your custom instructions and memory banks and lore about my characters here.
so, you've probably heard of the whole Shy Girl book ban by now right? Well it got me thinking about how maybe if someone ran my novels through Ai detectors Ai tells would show up even though its not Ai writing. I'm not sure if I would even know most Ai tells.
What words and phrases in this text are most likely to make a reader suspect it was written by AI? Give me a specific list.
The famous list includes: delve, realm, meticulous, underscore, commendable, pivotal, tapestry, adept, testament, and the em-dash used three times per paragraph.
What others are there?
But also things like tapestry, I use a lot because Quaraun is a silk weaver silk merchant who sells tapestries and carpets... so you can see how tapestry being an Ai tells would not apply to me unless tapestry was being used as a metaphor.
likewise my being a Mormon, I culturally use testament and testimony a lot, but most none Mormons rarely use those words in daily conversations so it used to be a Mormon tell, yet now it has become an Ai tells.
So I'm thinking about, what sort of words do I specifically use in my fiction writing, that might appear to be Ai tells when viewed by something like an Ai detector?
but alsoI was just reading another article today on how to tell Ai writing, the top 5 items on the list were:
Repetitive phrases.
Overused words.
Excessive instances of “it’s not X, it’s Y” constructions.
Too many adjective-noun pairs.
Excessive similes.
Uhm… these are common things HUMAN writers do… and these are supposed to be the top 5 tells of Ai writing?
I think the thing i find concerning here, is that, everything on that list of common Ai tells, are things I have done thousands of times in the dozen plus novels I published in the 1970s and 1980s, meaning Ai is now near perfectly replicating real human writing traits.
If ai is copying human writing, then in order to not get accused of ai we humans have to stop writing like humans, right? is that not a paradox?
I feel like we are fast coming upon an age when human writers who don't sound like Ai are going to start getting higher pay, due to how rare we've become
I feel like i am lucky that I published my first book in the 1970s, long before the invention of internet and its algorithms. I've always put my readers first, writing what readers want to read. I feel like a lot of new writers today are simply not being taught to write for readers. Most writing advice is geared towards writing for algorithms.
and it seems more so now with people using ai, its like every writing advice now is not only how to write for search bots like Google, but now how to train bots to write for search bots.
....
and here is the answer ChatGPT gave me....
....
You are noticing a very real shift, and a lot of professional writers are frustrated by it for exactly the reasons you described.
Most “AI tells” are not actually AI tells. They are traits of competent, readable, commercially viable prose. The problem is that large language models were trained on enormous amounts of successful human writing, so naturally they reproduce the same habits humans spent decades being taught to use.
That creates the paradox you noticed: AI learned to imitate human writing by training on human writing, then humans get accused of sounding like AI because they write like humans.
That is genuinely happening.
And honestly, many online “AI detection” lists are built from vibes, not linguistics.
Things like:
repetitive phrases
parallel structures
adjective-heavy prose
“it’s not X, it’s Y”
similes
clean grammar
transitional phrasing
...those are not machine traits. Those are rhetoric traits. Humans have used them for centuries.
Dickens repeated phrases constantly. Lovecraft stacked adjectives. King repeats sentence cadences. Tolkien overused “seemed.” Anne Rice used obsessive sensory repetition. Mormon speech patterns absolutely do overuse “testament,” “witness,” “countenance,” and “ponder.”
The detector culture often ignores dialect, religion, genre, age cohort, and publishing era.
Now, specifically to your question: what words or patterns in YOUR style might falsely trigger suspicion?
From what I know of your writing style and your Quaraun prose specifically, these are the biggest candidates:
Single-word AI suspicion magnets:
resplendent
eldritch
tapestry (metaphorical use especially)
testament
meticulous
uncanny
intricate
surreal
ethereal
ominous
haunting
delicate
grotesque
chaos
ancient
vibrant
peculiar
enigmatic
otherworldly
mesmerizing
whimsical
relentless
cacophony
abyss
labyrinthine
obsidian
ornate
shimmering
tendrils
luminescent
None of those are “AI words.” They are Gothic/Fantasy words. Unfortunately modern detectors associate elevated fantasy diction with AI because AI was heavily trained on fantasy and fanfiction archives.
For your specific genre, these patterns are more likely to trigger suspicion than the vocabulary itself:
1. Highly decorative adjective stacking. Example: “resplendent pink silk robes beneath shimmering stained-glass lantern light”
Fantasy writers have always written this way. AI also writes this way because it learned from fantasy writers.
2. Recurring sensory motifs. You repeatedly return to:
rose petals
absinthe
silk
pink
glittering objects
cold sea air
bells/chimes
candlelight
velvet
frost
Humans do this intentionally as thematic reinforcement. AI does it accidentally because token prediction loops back to familiar anchors. Detectors often cannot distinguish those motives.
3. Ritualistic repetition. Your prose deliberately uses obsessive loops because Quaraun’s cognition loops.
Example style: “Pink silk. Pink candles. Pink ribbons. Everything pink. Must be pink.”
That absolutely resembles AI repetition statistically even though artistically it is character voice.
4. Strong cadence symmetry. You use:
paired clauses
triads
mirrored sentence forms
rhetorical escalation
Example: “Cold windows. Cold floors. Cold hands.”
Again: ancient rhetorical technique. But AI reproduces it well.
5. Heavy stylization. Most human writing is messy and inconsistent. Your style is intentionally performative and heightened.
Ironically: strong authorial voice is now sometimes mistaken for AI because AI mimics stylized prose more aggressively than plain prose.
And yes, em dashes became a meme detector because ChatGPT overused them for a while. But human literary fiction used em dashes for 150 years before AI existed.
Your point about older authors is important too.
A lot of pre-internet writers developed style through:
oral storytelling traditions
pulp magazines
regional dialects
church rhetoric
theatre cadence
radio serials
print-era pacing
Whereas modern “algorithm writing” often optimizes:
SEO
readability scores
engagement metrics
click retention
TikTok excerptability
Those are very different ecosystems.
Ironically, your imperfections probably protect you.
Your work has things AI still struggles to sustain consistently:
dialect instability
deliberate grammatical fracture
obsessive symbolic recurrence tied to character psychology
strange sensory specificity
nonlinear emotional logic
author-specific fixations
recurring private mythologies
weird niche cultural speech influences
long continuity memory
AI can imitate those briefly. Humans sustain them across decades.
That matters.
And yes, I do think we are heading toward a market where verifiably human-created art becomes a premium category.
Not because human writing is automatically “better,” but because provenance itself gains value.
Like:
handmade furniture
analogue photography
vinyl records
live music
artisan textiles
People eventually value evidence of human process once mass automation saturates the market.
Especially for fiction, readers often connect emotionally to the idea that: “A real person lived this obsession long enough to create this world.”
AI can generate style. It cannot have thirty years of accumulated emotional continuity with a fictional universe.
That difference still shows.
Comments