Let me tell you what I feel every time I hit publish. Anguish. Utter, debilitating anguish. Because I’m dead sure what I just sent into the world is garbage. That it’s not good enough and I’m not good enough and I’m never going to be good enough.
Because I’ve never written a novel and I can’t write poetry and I’m not one of those edgy girls who writes wicked clever snark so I’m not a “real” writer whatever the hell that means, just some human dumping emotions on the page and I can’t seem to stop because if I go too long without writing I start to feel crazy.
When Emily Dickenson was dying, her maid sat down at the edge of the bed and told her she’s been rescuing scraps of poetry from her pockets for years and she’d like to read all of them. Margaret Maher was her name and Emily said really? Then I shall bequeath them to you but you must promise to consign them to the fire when you are done because they’re no good. There were thousands. A steamer trunk full.
I don’t need to tell you Margaret didn’t honor her promise.
Kafka said the same. Told his friend Max Brod to burn all his writing when he’s dead, it’s no good anyway. My writing is like me, he said, no good. Never good enough. Waste of time. Burn it all. His friend didn’t listen either.
Not all writers hide their work. Some publish despite their misgivings. Steinbeck kept a journal while he was writing Grapes of Wrath and in the journal there’s an entry that says “I’m not a writer. I’ve been fooling myself and other people.”
After publishing her eleventh book Maya Angelou said “I have published eleven books now and each time I think — they’re going to find me out now.”
There’s name for that, coined in 1978. It was called Impostor Phenomenon although we get it wrong, call it impostor syndrome. It’s not a syndrome, it’s a strange phenomenon. Know where it comes from? Trauma.
That’s why it’s a strange phenomenon. Because not everyone with trauma feels that way. But some people do and we still don’t really know why.
Toni Morrison said when she’s writing she feels like it’s the best thing she’s ever written but when it’s done she knows how terrible it is. Virginia Woolf, Sylvia Plath, I could write an entire piece on writers doubting their own abilities, writers who feel like one day they’ll get called out and those are the ones the world grew to know.
Not the nobodies like me.
I'm Nobody! Who are you? Are you – Nobody – too?
And typical of humanity we see a pattern and misunderstand it. Which explains the trope of the tortured artist. Because that’s what we think. Poor bastards dumping their pain into words or on the canvas, like Van Gogh. But it’s not true. The best work came from the times they were okay. Van Gogh painted his best in the asylum, on meds.
The magic of the time before the internet is the letters writers left behind.
Like Kafka saying his father looked at him like a bug on the floor. See the look in his eyes, and feel like a bug waiting for the shoe to crush him. But when he wrote, that faded away. He wrote because he had to. Because a non-writing writer is a monster courting insanity.
Like Sylvia Plath writing that the worse enemy is the self doubt.
One day, years and decades later, a bunch of rich a-holes took all the work of dead writers and living ones, dead artists and living ones. And used it to make AI.
And acquired over a million users in five days flat.
Last week I wrote about Substack’s AI report. I told a fanciful story of Van Gogh, a fairy godmother and a magic canvas to illustrate the different ways people use AI.
I called it A Blunt Conversation About AI and that post has over 450 comments so far. So many it still leaves me stunned.
This post is a little less fanciful and far more factual.
Let’s start here:

Have a good look at that. Okay?
Now I want to show you a little experiment I did.
Here’s a writing prompt I gave ChatGPT:
Prompt: Write a story about a woman who gets lost in the forest and realizes the forest is a metaphor for her fears, for how much she’s afraid of. Write in first person, around 500 words, using human sounding text.
Here’s what it gave me:
I tightened the straps of my hiking pack and took a deep breath. The forest loomed ahead, the trees clawing the sky. This was supposed to be a simple hike — just a few hours to clear my mind. But as I stepped onto the trail, a strange unease prickled the back of my neck.
The day started beautifully. Sunlight dappled the forest floor, and birds sang like a well-rehearsed choir. I felt calm at first, letting the rhythm of my steps drown out the noise in my head. But somewhere along the way, I must have taken a wrong turn. Familiar markers vanished, and the trail seemed to dissolve into wilderness.
I kept walking, telling myself it would be fine, but my chest tightened with each step. The forest grew darker as the sun dipped behind the trees. Shadows stretched long and strange, and the cheerful chirps of birds were replaced by rustling in the underbrush. My throat felt dry, and I tried calling out, but my voice was swallowed by the trees.
Panic started to creep in…
—written by ChatGPT
That’s not the entire story. You can read the entire story here, if you’re curious.
If you read that on a writing site, would you point a finger and shriek that it’s AI? Because I don’t think you would. There are humans who can’t write that well. I think ChatGPT wrote the pants off a lot of humans.
I read a Forbes article by Jonathan Gillham, the founder of Originality.ai. He said most people think they can spot AI, but they’re wrong. I would agree with him.
I do think some people can. But I think they’re the minority.
Substack surveyed over 2,000 Substack publishers, including writers, artists, podcasters, and video creators, from around the world.
21% were bestsellers (with 100+ paid subscribers), and 70% run paid newsletters.
There were publishers at every revenue level, from free to seven figures.
Read that again. Some make seven figures.
Out of those 2,000+ surveyed publishers:
45.4% said they’re using AI
52.6% said they’re not
2% were unsure
Do I use AI for my writing? No.
Not one word of anything I publish is AI. Not one.
Let me tell you why.
Do you know how AI got so good at writing like a human?
Theft.
Theft of copyrighted works.
I wish I was kidding, but I’m not.
Advocates of AI, for the most part, don’t actually understand how it works. They say things like “well, people get inspired by books too, it’s the same concept.”
No. It’s not.
Here’s an actual comment someone left on one of my posts.
If J.K. Rowling read dozens of (copyrighted) fantasy books and then created her own book, infused with ideas from other authors and made billions on her books, why has no one sued her for copyright violation? The scale is different but the concept is the same. (source)
The concept is not the same. Not even close.
You can only think that if you don’t know how AI actually works.
Sure, a human can get “inspired” by details from books. But we still need to connect them together. Create a plot. Write the scenes. Make sure it’s interesting and compelling, the pace is good and the characters feel real, not flat. It’s harder than you think when you actually try it.
And if I did that, when I’m done, I own the copyright to it the minute I put it into a tangible medium. Print it out on paper, or put it online, the copyright is mine.
Here’s what that means. No one is allowed to use my work without permission.
Most people don’t understand how AI and LLMs work. They think it’s doing what we do. But that’s not what’s happening at all.
Simon Willison is a programmer who’s written an explanation of how large language models (LLMs) like ChatGPT, Google Bard, and Llama 2 work.
Basically, LLMs “guess” the next word the same way an iPhone keyboard does, but on a much bigger scale. AI is basically probability generation. What word should come next based on all the words it’s saved. Not read. Saved.
Everything AI scraped is stored for future use.
Let me show you another thing.
Can a human do that? No. Humans don’t have instant recall like that. We get inspired by books. AI uses copyrighted work without permission. Which explains why Atwood is one of more than eight thousand authors demanding compensation from OpenAI.
They’re not going to get it.
President Trump has already said he does not believe writers and artists should be compensated for work used to “train” AI. He doesn’t understand how it works either.
Willison explains exactly how LLMs work. Technically. From the perspective of a programmer who has programmed AI.
When materials are fed to AI, it creates “tokens.”
“The” with a capital T is a token with a number assigned. Let’s say it’s token #464. Let’s say the word “the” with a lowercase ‘t’ is token #262. There are tokens for phrases, full sentences, entire chapters, entire pieces of writing.
What makes AI work is the scale. Terrabytes of data, millions of dollars of computing powers, giant racks of GPUs running to examine that data, identify patterns and crunch that down to billions of floating token numbers.
Then assembling “tokens” at will when a human types in a prompt.
Literally pulling from copyrighted work.
And that’s such a simple version it barely does it justice. But it will suffice and if you’re interested, his technical description is about a 20-30 minute read.
The way it assembles images is an entirely different process that would take another half hour to explain at it’s most basic.
Point being, no, AI doesn’t just get “inspired” by words the way a human does.
It takes copyrighted words. Without permission.
Now remember what copyright grants a writer.
The legal right to decide who makes a profit from your work.
Here’s another snippet from the Substack AI Survey

Yay, we now have health advice from ChatGPT. Philosophy by ChatGPT.
So I did a little test. I asked ChatGPT how to deal with impostor syndrome. It told me that a gratitude practice or cognitive behavior therapy are successful. No, they are not. But the article was pretty and sounded convincing.
I will remind you of two facts from the survey.
of the 2000+ publishers surveyed, 21% were bestsellers
of the 2000+ publishers surveyed, 70% run paid newsletters.
Here’s one more sample from ChatGPT.
Prompt: A man is sitting by his father’s deathbed so his father doesn’t die alone. He’s filled with memories and feelings. Write a paragraph about his thoughts and feelings, in first person, with human sounding text.
Here is the text ChatGPT wrote…
The room smelled faintly of antiseptic and something heavier, something final. I sat in the hard vinyl chair next to my father’s bed, hands gripping the armrests like they might keep him here a little longer. I watched his chest rise and fall, each breath a tether to a life that felt thinner by the second. Memories flickered through my mind unbidden — his laughter when I was a boy, the weight of his hand on my shoulder, the arguments we never resolved. I want more time, but not like this. Not with machines doing the work his body had given up on. I wanted to stay here forever and I wanted it to be over, all at once.
— written by ChatGPT
People who write with AI tell me many reasons they do. Because it helps them write faster or better or organize their thoughts. Because they have ADHD, just like Jules Verne and George Bernard Shaw, Simon Sinek, Mel Robbins and Seth Godin.
There are lots of reasons people use AI. There is one reason I do not. Because I don’t want anyone to violate my copyright so I won’t violate someone else’s.
And if that means I agonize over a piece for hours and hit publish thinking it’s trash, and that everything I write is trash and I’m never going to be good enough to sit at a table with real writers — well, so be it. That’s what craft writing is for some of us.
As always, love to know what you think.
AI is all so pervasive now, but I would like to ask those that use it, "Don't you feel like you are cheating the authors that unknowingly contributed their best words to it?"
As a teacher I get ai to spit out everything from lists to worksheets to essays. It saves me a lot of time, but I use it for mostly low level learners. The style is pretty obvious - I call it relentless obviousness. "Echoes of laughter," "ghosts of games," from your examples are not obvious in of themselves, but the repetition of those kinds of descriptions, and also collocations, is definitely a tell-tale sign. Also, I'd look out for what I call mechanically-correct dependent clauses! After a while, you'll see ai prose is rife with them. So, in my opinion, those who are using ai to "write" stories are just fooling themselves. Thanks for the article!