Any experience with AI critics?

[StillStunned]

That said, psychologically, if you're the type of person who is motivated by hearing you aren't good enough, but you don't have anyone around whom you would entrust with your creative endeavors, I suppose you might get something out of doing so.

I'm more than happy to tell anyone that they're not good enough.

 

[cw5523729]

Serious question - doesn't asking an AI to be harsh automatically guarantee that it's going to make up SOME reason to be judgemental, whether fair or not?

 

[cw5523729]

I'm more than happy to tell anyone that they're not good enough.

Me too. See below.

Ha! Shows what you know. I'm a complete a-hole. Like, legendary levels of a-holeness.

Other a-holes look at me and mutter "Wow, that guy is a real a-hole. We need to up our game."

 

[AwkwardMD]

@AwkwardMD may use hurtful words,

 

[EmilyMiller]

So my question is, have you ever used tools like "harsh critic" to check your stories, and was that useful?

Nope. I think people don’t understand how LLMs work.

What they do is to formulate what you are asking as a series of factors and then search for what on average people have said in reply (the closest statistical match on the most factors).

If the question is, “Does the Sun orbit the Earth?” Then many people have opined on this and you will most likely get an accurate reply.

If the question is, “What do you think of my story?” You will get a distillation of feedback that various humans have made to various stories, none of which is your story.

LLMs can’t critique your story, they don’t understand your story. All they can do is regurgitate, “This is a bland collation of story feedback from humans that vaguely matches a few key words or phrases in your story.”

Ask a human what they think.

 

[EmilyMiller]

Nope. I think people don’t understand how LLMs work.

What they do is to formulate what you are asking as a series of factors and then search for what on average people have said in reply (the closest statistical match on the most factors).

If the question is, “Does the Sun orbit the Earth?” Then many people have opined on this and you will most likely get an accurate reply.

If the question is, “What do you think of my story?” You will get a distillation of feedback that various humans have made to various stories, none of which is your story.

LLMs can’t critique your story, they don’t understand your story. All they can do is regurgitate, “This is a bland collation of story feedback from humans that vaguely matches a few key words or phrases in your story.”

Ask a human what they think.

TL-DNR - LLMs have no knowledge, they are statistical inference engines. For areas where there is strong agreement and many articles / books etc. saying the same thing, an LLM can regurgitate it fine. If it’s something that no one has written about (like your story) it will fake it by mashing together reviews of other stories.

 

[StillStunned]

Coming back specifically to the topic of the OP: a while ago someone ran one of my stories through AI for an analysis. No, I wasn't happy with it.

Anyway, the AI highlighted these issues as shortcomings:

* **Character Depth**: The cobbler is a good foil, but he’s a bit one-dimensional. He's a craftsman, he's humble, and he's easily seduced. What's his inner world like? What makes him more than just a tool for the lady's plot? We get a hint of his passion for his work, but his journey feels less like a transformation and more like a series of reactions. The lady, too, is a bit of a stereotype: the mysterious, powerful, cold "witch" figure. Giving her even a single, small vulnerability or a hint of a hidden motive would make her far more compelling. Right now, she’s a force of nature, but forces of nature are rarely as interesting as people.

* **Connecting with the Reader**: Because the characters lack complexity, it’s hard to truly identify with them. The cobbler's desire is palpable, but we don't understand *why* this specific encounter is so transformative for him beyond simple lust. The lady is an object of awe and fear, but not someone the reader can connect with emotionally. The story is a tableau, a series of events, but not a journey we take *with* the characters. The reader is an observer, not an participant.

* **Plot Complexity**: The plot is straightforward, which is fine, but it’s almost *too* simple. The second half, where she takes on the King, feels like a foregone conclusion. There's no real opposition, no challenge, no moment where her plan might fail. The King and his court are paper-thin obstacles. A good plot needs more friction. What if the King was less bored and more cunning? What if her boots didn't fit perfectly, creating a moment of panic? Adding a minor complication would elevate the stakes and make her triumph feel earned, not inevitable.

The story in question was Black Boots To Meet The King: a fairy tale, told in the style of a fairy tale, with archetypes and folktale types. The AI missed the entire objective of the story: the shortcomings that it lists are precisely the elements of what *makes* the story a fairy tale.

By contrast the readers rate the story 4.78, and many of the comments praise it for sounding exactly like a fairy tale.

 

[EmilyMiller]

Coming back specifically to the topic of the OP: a while ago someone ran one of my stories through AI for an analysis. No, I wasn't happy with it.

Anyway, the AI highlighted these issues as shortcomings:

* **Character Depth**: The cobbler is a good foil, but he’s a bit one-dimensional. He's a craftsman, he's humble, and he's easily seduced. What's his inner world like? What makes him more than just a tool for the lady's plot? We get a hint of his passion for his work, but his journey feels less like a transformation and more like a series of reactions. The lady, too, is a bit of a stereotype: the mysterious, powerful, cold "witch" figure. Giving her even a single, small vulnerability or a hint of a hidden motive would make her far more compelling. Right now, she’s a force of nature, but forces of nature are rarely as interesting as people.

* **Connecting with the Reader**: Because the characters lack complexity, it’s hard to truly identify with them. The cobbler's desire is palpable, but we don't understand *why* this specific encounter is so transformative for him beyond simple lust. The lady is an object of awe and fear, but not someone the reader can connect with emotionally. The story is a tableau, a series of events, but not a journey we take *with* the characters. The reader is an observer, not an participant.

* **Plot Complexity**: The plot is straightforward, which is fine, but it’s almost *too* simple. The second half, where she takes on the King, feels like a foregone conclusion. There's no real opposition, no challenge, no moment where her plan might fail. The King and his court are paper-thin obstacles. A good plot needs more friction. What if the King was less bored and more cunning? What if her boots didn't fit perfectly, creating a moment of panic? Adding a minor complication would elevate the stakes and make her triumph feel earned, not inevitable.

The story in question was Black Boots To Meet The King: a fairy tale, told in the style of a fairy tale, with archetypes and folktale types. The AI missed the entire objective of the story: the shortcomings that it lists are precisely the elements of what *makes* the story a fairy tale.

By contrast the readers rate the story 4.78, and many of the comments praise it for sounding exactly like a fairy tale.

And LLMs appear authoritative and informative to people who know little about the area the model is opining on. To an expert in the area, they are awful. This is smoke and mirrors. At some point the bubble will burst. Sadly as the whole stock market is being held up by AI companies, the burst will be very, very painful.

 

[StillStunned]

And LLMs appear authoritative and informative to people who know little about the area the model is opining on. To an expert in the area, they are awful.

The paradox is that everyone knows it's shit in their own area of expertise, but they're willing to use it for other things, even when the experts in those areas tell them it's shit.

 

[MrPixel]

Sadly as the whole stock market is being held up by AI companies, the burst will be very, very painful.

Already wavering. We have one particular computer hardware-related stock (NOT Nvidia!) in our portfolio that has grown enormously over the past year, more than likely due to AI. It's been an up-down sawtooth thing (mostly up) in the past three or four months, and I'm trying hard not to sell the dips. At what point do I let it go? I really don't know, and I don't know who to listen to any more.

Steve (the MMC in the Barstow series) doesn't know, either. He's a professional investor, too, and cusses the computer out every time he logs-on in the morning.

 

[alohadave]

The paradox is that everyone knows it's shit in their own area of expertise, but they're willing to use it for other things, even when the experts in those areas tell them it's shit.

It's funny how that parallels news reporting. You know when they are wrong about something you know about, but don't question it otherwise.

 

[joy_of_cooking]

Gel Mann paradox!

 

[joy_of_cooking]

Which Gemini informs me is actually the Gell-Mann amnesia effect.

 

[Bramblethorn]

I mean it's better than no opinion at all?

Theoretically, which is the reason for this thread. If you guys can confirm that it's just a bunch of nonsense with no relation to the text you feed into it, that would settle it.

It is definitely related to the text, but not necessarily in the way one would want it to be.

For simple summaries of quantifiable things, like "this story has a lot of run-on sentences", it's probably good enough. Where it's much less reliable is in telling you whether that's a good thing. Mostly run-ons are bad but sometimes they're good, and while AI can probably regurgitate some anonymous human's thoughts about when it's okay to use run-on sentences, it can't apply that to assessing your story to gauge whether this is one of those times. (But that won't stop it from offering an opinion on the matter.)

 

[TheLobster]

Coming back specifically to the topic of the OP: a while ago someone ran one of my stories through AI for an analysis. No, I wasn't happy with it.

Anyway, the AI highlighted these issues as shortcomings:

* **Character Depth**: The cobbler is a good foil, but he’s a bit one-dimensional. He's a craftsman, he's humble, and he's easily seduced. What's his inner world like? What makes him more than just a tool for the lady's plot? We get a hint of his passion for his work, but his journey feels less like a transformation and more like a series of reactions. The lady, too, is a bit of a stereotype: the mysterious, powerful, cold "witch" figure. Giving her even a single, small vulnerability or a hint of a hidden motive would make her far more compelling. Right now, she’s a force of nature, but forces of nature are rarely as interesting as people.

* **Connecting with the Reader**: Because the characters lack complexity, it’s hard to truly identify with them. The cobbler's desire is palpable, but we don't understand *why* this specific encounter is so transformative for him beyond simple lust. The lady is an object of awe and fear, but not someone the reader can connect with emotionally. The story is a tableau, a series of events, but not a journey we take *with* the characters. The reader is an observer, not an participant.

* **Plot Complexity**: The plot is straightforward, which is fine, but it’s almost *too* simple. The second half, where she takes on the King, feels like a foregone conclusion. There's no real opposition, no challenge, no moment where her plan might fail. The King and his court are paper-thin obstacles. A good plot needs more friction. What if the King was less bored and more cunning? What if her boots didn't fit perfectly, creating a moment of panic? Adding a minor complication would elevate the stakes and make her triumph feel earned, not inevitable.

The story in question was Black Boots To Meet The King: a fairy tale, told in the style of a fairy tale, with archetypes and folktale types. The AI missed the entire objective of the story: the shortcomings that it lists are precisely the elements of what *makes* the story a fairy tale.

By contrast the readers rate the story 4.78, and many of the comments praise it for sounding exactly like a fairy tale.

Hot take: the AI isn't the problem here. It has simply regurgitated the bulk of what's considered the best contemporary advice about writing.

What's actually the problem is that most of said advice is utter horseshit.

 

[WeShallUnclench]

Hot take: the AI isn't the problem here. It has simply regurgitated the bulk of what's considered the best contemporary advice about writing.

What's actually the problem is that most of said advice is utter horseshit.

I can't speak for most advice, but I always keep coming back to something I saw on tumblr a while back, a post that quoted an "expert" who's main angle was to push "show, don't tell" to the absolute maximum:

Describe every sensory experience in crystalline detail, more detail in fact than an actual person would notice irl unless they are on some mind-opening superdrug. Never ever reveal a character's thoughts directly.
Basically the pinnacle of prose is to put a VR headset on your reader.

Now, as with most advice, you can take some of it, leave some of it, let it guide you towards new perspectives on your writing.

But what he did, put to the suggested extreme, was absolutely unbearable to read. You could barely follow whatever scene he was trying to write because you were drowning in the pretentious craftsmanship of his sensory experiences (TM). The pinnacle of someone trying to "perfect writing technique" with no eye on whether or not the result is enjoyable or even tolerable.

The most important lesson I took away from THAT is that sometimes tell is actually better than show.

 

[Bramblethorn]

I can't speak for most advice, but I always keep coming back to something I saw on tumblr a while back, a post that quoted an "expert" who's main angle was to push "show, don't tell" to the absolute maximum:

Describe every sensory experience in crystalline detail, more detail in fact than an actual person would notice irl unless they are on some mind-opening superdrug. Never ever reveal a character's thoughts directly.
Basically the pinnacle of prose is to put a VR headset on your reader.

Now, as with most advice, you can take some of it, leave some of it, let it guide you towards new perspectives on your writing.

But what he did, put to the suggested extreme, was absolutely unbearable to read. You could barely follow whatever scene he was trying to write because you were drowning in the pretentious craftsmanship of his sensory experiences (TM). The pinnacle of someone trying to "perfect writing technique" with no eye on whether or not the result is enjoyable or even tolerable.

The most important lesson I took away from THAT is that sometimes tell is actually better than show.

100%. There are very few absolute rules to good writing, but people are addicted to that "one weird trick" mentality whether it's "show don't tell" or "adverbs bad" or whatever. Many of those rules have some truth to them, but one needs to take them with a large grain of salt and ask: why does this "rule" exist? Do the reasons for it apply to this story?

There's a Simpsons episode where Homer is asked to design a car: "that leprechaun on the antenna that helps you find your car in a parking lot? Every car should have one!" Show-don't-tell is like that leprechaun: used judiciously it can help focus attention on key scenes but used everywhere, like you say, it just overwhelms.

 

[Zootonius]

@AwkwardMD we're out of rum, are you okay if we move directly to absinthe?

@WeShallUnclench why on earth are you at all interested in the "opinion" of a statistical engine? You might as well ask a waterfall its opinion on Kafka.

Although, I would be interested in a cockroach's opinion on Kafka.

 

[XerXesXu]

Case Study (Copilot).

Prompt:

If I were to ask you to review a passage of writing either 'harshly' or 'according to the standards of a literary magazine' would you review them differently; if so, in what ways?

Response:

Absolutely — those two approaches would yield quite different reviews. Here's how they'd diverge:

"Harshly" — A Brutal Critique

This style is unapologetically blunt, often bordering on caustic. It’s useful when the goal is to expose weaknesses with surgical precision or provoke growth through discomfort.

• Tone: Direct, unsparing, possibly sarcastic or scathing
• Focus: Flaws, clichés, structural weaknesses, overwrought language, lack of originality
• Examples of phrasing:
  • “This metaphor is so tired it should be retired.”
  • “The pacing drags like a lecture on molasses.”
  • “Reads like a first draft that forgot to become a second.”
• Purpose: To challenge the writer, strip away self-indulgence, and push for radical improvement

Literary Magazine Standards — Professional Editorial Review

This style is rigorous but respectful, modeled after how editors assess submissions for publication. It balances critique with appreciation and aims to elevate the work.

• Tone: Polished, constructive, discerning
• Focus: Craft, voice, originality, thematic depth, stylistic coherence
• Examples of phrasing:
  • “The imagery is evocative, though occasionally veers into abstraction without anchoring.”
  • “The narrative voice is promising, but the emotional arc lacks resolution.”
  • “Consider tightening the syntax to better serve the rhythm of the prose.”
• Purpose: To guide refinement, highlight strengths, and assess publishability

Why the Difference Matters

The “harsh” review is like bootcamp — it’s about breaking down to rebuild. The literary magazine review is more like mentorship — it’s about shaping and polishing. Both can be valuable, depending on your goals: shock therapy or professional development.

Would you like to test both styles on a sample passage? I’d be delighted to show you how they play out side by side.

 

[XerXesXu]

Prompt:

Yes, on this passage:

‘
It's a Grey, Grey World.

The hair of man turns grey with age, his soul, the vibrant colours of forest leaves before they fall ... the souls of most men do.

Trevor was lonely, desolate, and the dread of an empty life after retirement caused him palpitations. The palpitations were worst, shortly after waking, on cold, grey mornings in bleak winter.

Comforting cup of warming, sweet tea in hand, he stared through his kitchen window at the darkness of an early December morning. Already his navy-blue gabardine raincoat was pulled over his grey work-suit in readiness to leave. Morosely illuminated by the dull glow from street lamps, and revealed by their refracted light, flecks of snow driven by a wind from the north western Arctic gusted across the road. A car drove slowly by, its probing headlights picking out the tracks of a previous vehicle in the film of slurry formed by melting snowflakes. Lifting his foot, Trevor examined the sole of his scuffed, creased Oxford lace-ups. No holes. His feet may get cold, but at least they would not get wet.

His life, he reflected, was like his shoes, secure against the worst, but worn, cold and joyless. As he contemplated the unpleasant journey to his office, the thought that his soul had turned an unnatural grey tortured him. A dank walk through chilly blasts to the tube station would be followed by an hour-long journey in underground tunnels, on a stuffy train, being careful to avoid eye contact with fellow commuters, before emerging once more above ground for the short walk through a dark, cold, grey, concrete and glass canyon, to Westminster Town Hall Annexe. A local government employee earned too little to buy a house in a central London borough, which is why he lived so far from his place of work. But his employment was secure, his pension generous and inflation-proofed.

In his air-conditioned office Trevor would be neither too hot nor too cold; his work-surface would be appropriately lit, and his ergonomically adjustable chair would support him in comfort for the next eight hours. But he would not see the dawn.

At some point he would walk to the coffee machine to refresh his cup and from there see, above the roofs of adjacent buildings, the dull, grey, morning sky. He might even stroll to the window to look down on the figures hurrying back and forth in the cold, damp gloom, and feel gratitude that he was ensconced in the warmth. That would be the highpoint of his day. Worse yet, as he contemplated his future, that would be the highpoint of his remaining life. With no exceptional fortune of which to boast to his colleagues, he would shamefully confess to himself that his only consolation came from comparing his comfort to the relative discomfort of others.

For eight hours he would receive paper, sort, collate, complete, then file, or, with appropriate recommendations forward it. His position with a local authority was responsible and indispensable, though it was not his own local authority. Of course he needed the income, but it was the opportunity for human contact, the excuse to talk, the occasional smile, the odd joke, even sometimes a shared intimacy, that drew him to the office on time, every day. That was what sustained his life, and soon this prop would be removed; what would follow was too horrible for him to contemplate.'

 

[XerXesXu]

Response:

Here’s a side-by-side critique of your passage, showing how a “harsh” review and a “literary magazine standards” review would differ in tone, focus, and intent:

Harsh Review: “Grey, Grey, Grey — and Dull”

Tone: Brutal, unsparing, designed to provoke

• The opening metaphor — “The hair of man turns grey…” — is overwrought and vague. It gestures toward profundity but lands in cliché.
• Trevor’s emotional state is hammered repeatedly with synonyms: “lonely, desolate, dread…” — we get it. The repetition feels like padding.
• The prose is bloated with adjectives: “comforting cup of warming, sweet tea,” “scuffed, creased Oxford lace-ups,” “dark, cold, grey, concrete and glass canyon.” It’s like the thesaurus exploded.
• The pacing is glacial. We spend a paragraph on snowflakes and shoe soles, yet the emotional stakes barely shift.
• The central metaphor — life as grey — is hammered so relentlessly it loses impact. Subtlety is sacrificed for monotony.
• The final paragraph gestures at poignancy but ends in melodrama: “too horrible for him to contemplate” is vague and lazy. What’s horrible? Death? Isolation? Be specific or risk sounding like a soap opera.

Verdict: This reads like a first draft that’s trying too hard to be literary. Strip it down. Find the pulse beneath the grey.

Literary Magazine Review: “A Study in Monochrome”

Tone: Professional, constructive, discerning

• The piece offers a meditative portrait of post-retirement dread, anchored in atmospheric detail. The motif of greyness is consistent, though occasionally overused.
• The opening metaphor sets a somber tone, but its abstraction (“the souls of most men do”) could benefit from more grounding in Trevor’s specific experience.
• The descriptive passages — snow, shoes, office — are vivid, but risk overwhelming the narrative arc. Consider tightening to maintain momentum.
• Trevor’s internal reflections are poignant, especially the moment he compares his comfort to others’ discomfort. This line hints at a deeper moral reckoning that could be explored further.
• The final paragraph introduces a compelling tension: the contrast between bureaucratic routine and the human need for connection. This is where the piece finds its emotional core.
• Stylistically, the prose leans toward the ornate. A literary magazine might suggest trimming modifiers and clarifying syntax to sharpen the voice.

Verdict: With refinement, this could become a resonant character study. Focus on emotional specificity and structural clarity to elevate the piece.

Would you like me to help revise the passage in either style — stripped down and raw, or polished for literary submission?

 

[XerXesXu]

Prompt:

Yes, polished for literary submission.

Response:

Wonderful choice. Here's a refined version of your passage, shaped for literary magazine standards — preserving its emotional depth while enhancing clarity, rhythm, and stylistic precision:

It’s a Grey, Grey World (Polished Version)

The hair of man turns grey with age; his soul, too, fades — once vibrant as autumn leaves, now dulled before the fall. For most men, it does.

Trevor felt the weight of solitude. Retirement loomed, and with it, the dread of an empty life. The palpitations came strongest in the mornings — cold, grey, winter mornings.

Cradling a sweet, warming cup of tea, he stared through the kitchen window into the December dark. His navy gabardine raincoat already cloaked his grey work-suit, ready for departure. Outside, the street lamps cast a dull glow, revealing flecks of snow driven by a northwestern wind. A car crept past, its headlights tracing the slurry left by earlier traffic. Trevor lifted his foot and examined the sole of his scuffed Oxford lace-ups. No holes. Cold, perhaps — but not wet.

His life, he thought, resembled those shoes: intact, functional, yet worn and joyless. The thought of his soul turning an unnatural grey gnawed at him. Ahead lay a damp walk through biting gusts to the tube station, followed by an hour underground in stuffy silence, avoiding eye contact, before emerging into a canyon of concrete and glass. Westminster Town Hall Annexe awaited. A local government employee earned too little to live nearby; hence the long commute. Still, the job was secure, the pension generous, inflation-proofed.

In his air-conditioned office, Trevor would be neither too hot nor too cold. His desk would be well-lit, his chair ergonomically sound. But he would not see the dawn.

Later, fetching coffee, he might glimpse the grey sky above the rooftops. Perhaps he’d stroll to the window, watching figures hurry through the damp gloom, and feel a flicker of gratitude for his warmth. That would be the high point of his day. Worse — the high point of his remaining life. With no exceptional fortune to share, he admitted, shamefully, that his only consolation came from comparing his comfort to others’ discomfort.

For eight hours, he would receive paper, sort, collate, complete, file — or forward with recommendations. His role was responsible, indispensable, though not in his own borough. He needed the income, yes, but it was the human contact — the excuse to talk, the occasional smile, the odd joke, even a shared intimacy — that drew him to the office each day. That sustained him. And soon, that prop would be gone. What followed was too bleak to contemplate.'

Would you like help crafting a cover letter or submission strategy to accompany this piece? I can also suggest literary magazines that might resonate with its tone and themes.

 

[Britva415]

So my question is, have you ever used tools like "harsh critic" to check your stories, and was that useful?

Given that stories infamously tend to get very little human feedback, and people tend to be reluctant to give even constructive criticism for fear of being rude (but de facto that just means radio silence with no growth opportunities), are AI critics a viable alternative to get some feedback? Or do they just produce irrelevant junk that cannot inspire better writing?

It's trivially easy to get actual humans to intelligently critique your writing. Google for "creative writing peer groups" and be prepared to do the same for them.

AI is going to roast you just because you asked it to. And the content of the roast isn't going to be based on what you wrote, it's going to be based on the content of other roasts which the AI was trained on.

Question: "Hey, AI, can you critique my piece?"

Answer: "Sure! This is what harsh criticism sounds like... [criticizes harshly but generically, maybe picking random quotes out of your work to create an illusion of relevance]... Was that helpful?"

 

[Britva415]

If you guys can confirm that it's just a bunch of nonsense with no relation to the text you feed into it, that would settle it.

Confirmed

If the chain of words is based on statistical trends in opinions, it might still indirectly serve as an opinion

It won't be an opinion about your actual content, though.

 

[Bramblethorn]

I shared this before: https://amandaguinzburg.substack.com/p/diabolus-ex-machina

She asked GPT for feedback on stories she was considering including in a query letter to an agent. It "quoted" passages that weren't in those stories, and described "The Summer I Went Viral" as being about social media fame, when it was actually about COVID-19. After she pointed out several of these issues, GPT responded:



It's possible that this explanation is itself bullshit - generative AIs don't have insight into their own reasoning, so when you ask them to explain why they did something, they'll try to replicate the kind of answer a human would give to that question. But whether it's true or false, it doesn't seem like the kind of technology I would trust for feedback.