“No AI” Is an Ego Policy, Not an Ethics Policy

Originally published: 02/12/26
Publication: Generative AI   

A purity rule you can’t enforce just punishes the honest.

Let’s stop pretending this is about ethics.

Publishing’s “No AI” rule is an ego policy: a purity ritual designed to protect status in a world where status is getting cheaper. It isn’t enforceable, it isn’t coherent, and it doesn’t do what it claims to do.

It punishes the honest and rewards the liars. That’s the design, whether they admit it or not.

That’s the whole game. If process can’t be verified, then process rules become theater. And in an industry that no longer controls distribution the way it used to, theater isn’t harmless. It’s self-sabotage. You don’t preserve trust by demanding an unprovable pledge. You preserve trust by demanding responsibility for the work.

I’ve lived through this fight already. Photography had its own version when digital arrived, and the arguments were the same: authenticity, purity, protection of the craft. But underneath the moral language was the real fear: if this tool is legitimate, what does that say about the suffering I endured to get here? The medium didn’t collapse. The prestige story did.

Publishing is replaying that exact drama, just with worse logic and a worse economic position, because writing never had the thing photography could lean on.

Writing has never had provenance

Photography had negatives. Later, it had RAW files. You could argue about manipulation, but you could always point to an originating artifact and say: this is the capture; everything else is transformation. That’s why Photoshop was never banned. What was banned was deception. If you composited, you disclosed. If you altered reality, you didn’t pretend otherwise.

Writing never had that.

There has never been a “RAW file for prose.” No artifact that proves where an idea came from, how a sentence formed, or what portion of the work emerged from observation, memory, conversation, or subconscious synthesis. A writer can sit in a café, watch a couple argue, go home, and write a scene shaped by that moment. There is no receipt. There never was.

And yet publishing functioned just fine for centuries under that ambiguity, governed by one simple norm: don’t copy expression; take responsibility for your work.

AI didn’t break that rule. It exposed how much we were pretending otherwise.

We already accept “authorship” without keystrokes

For most of modern business history, senior leaders did not “write” their own letters.

A CEO or company president would call in an assistant and talk through the points. Sometimes it was stream-of-consciousness at full speed while someone took shorthand. Sometimes it was a stenography machine. Sometimes it was a microphone and a cassette tape that got transcribed later. The raw material landed in the assistant’s lap, and the assistant did the assembly work: turning fragments, tangents, and half-finished sentences into a coherent letter that said what the executive meant.

No one looked at that finished letter and said, “This isn’t from the CEO because someone else assembled it.” Everyone understood authorship as intent, responsibility, and ownership. The assistant was the conduit. The executive stood behind the message.

Now ask the question that collapses the whole “No AI” posture.

What’s the difference if one were to hire a person to do that assembly work, or use AI to help do it, if the inputs are the same either way? If a human being were given exactly what the AI is given, what changes?

The honest answer is: nothing that matters.

And notice what nobody is asking for. There aren’t disclaimers demanding, “Did you dictate this to someone else?” There aren’t purity pledges that ask, “Did an assistant assemble your draft?” There is no moral panic about outsourcing structure to a human. That process is not being policed.

Only AI is targeted. That is where the logic collapses.

Publishing has lived with the same reality for decades, even when it involves famous names. William Shatner’s TekWar is a clean example. Shatner supplied concepts, plot direction, characters, and development beats. A hired collaborator structured it into a readable book. The public still calls it a Shatner book because Shatner is the one standing behind what it is.

If a human collaborator can translate an author’s vision into finished prose and it’s still considered legitimate authorship, then the software version isn’t being rejected on ethical grounds. It’s being rejected because it threatens a status story.

This isn’t an ethics policy. It’s a stigma policy.

The blanket ban is intellectually lazy

When publishers say “no AI,” they aren’t drawing a meaningful boundary. They’re swinging a blunt instrument because they don’t know how to articulate a precise one.

Spellcheck, grammar systems, copyediting, developmental edits, workshop feedback, agent polish: these are all tolerated, even though they materially alter prose. Why? Because they’re framed as refinement, not origin.

Generative systems scare people because they can originate text. So instead of saying, “We require human-authored narrative content, though tools may assist refinement,” publishers opt for the simpler fiction: no AI at all.

That rule is not ethical clarity. It’s conceptual surrender.

Worse, it punishes the wrong people. The writers most likely to use AI carefully as critique, analysis, stress-testing, and revision are professionals who already understand authorship and voice. Blanket bans select against them, while bad actors either lie or route around the institution entirely.

That’s not quality control. It’s inverse selection.

The economics are brutal: this stance drives talent away

Even if you don’t care about the philosophy, the economics are unforgiving. A purity rule you can’t enforce doesn’t just fail. It fails in a specific direction. It selects against the very writers you want.

This is classic adverse selection. The policy doesn’t filter for integrity. It filters for who is willing to play games. The people who treat the work seriously, who keep clean authorship boundaries, who can defend every sentence, who care about responsibility, are also the people most likely to look at an incoherent rule and say, fine, I’ll publish somewhere else.

And “somewhere else” now exists at scale.

Publishers used to have a moat: distribution, marketing, legitimacy, editing infrastructure, and advance capital. That moat has been shrinking for years. AI-assisted workflows shrink it faster by reducing iteration cost and production friction. If publishers add a process-policing toll on top of that, they hand writers a simple decision:

Why am I giving away rights and revenue to be treated like a suspect under a rule that can’t be verified?

The result is predictable. Quality writers move to self-publishing, direct-to-audience models, subscriptions, patronage, and smaller presses with saner policies. Meanwhile, the first publishers who drop the theater and adopt a coherent standard, human authorship attestation, responsibility for originality, and no deception, will quietly absorb the best talent. In a low-margin business, talent concentration matters. A lot.

This is the part gatekeepers never want to hear: policing process is a luxury you can only afford when you still control distribution. If you don’t, you train the market to route around you. And once a writer builds an audience directly, the institution doesn’t regain leverage later. It becomes optional.

The darker motive: redefining “professional” to keep the club small

There’s a darker reading of the blanket ban, and it’s not a conspiracy. It’s a guild reflex.

If you can’t stop a tool from existing, the next best move is to define legitimacy in a way that excludes anyone who uses it. Create a “professional” identity that only counts writers who meet a purity test, then treat that pool as the only real pool.

Photography could have tried this. Imagine a professional league that only allows film shooters. Not because digital images can’t be good, but because controlling membership controls access: jobs, credibility, prestige, and rates. The argument would sound noble. The effect would be cartel-like.

It’s worth saying out loud because even if this is happening, many of the people doing it won’t think they are. They’ll experience it as “defending the craft,” not as “excluding competitors.” That’s how gatekeeping usually works. It feels like virtue from the inside.

And even without bad intent, the incentives push institutions toward exclusion, because exclusion is the only lever they can still reliably pull.

But the economics don’t care what story you tell yourself. A purity boundary in a world of unprovable process doesn’t filter for quality. It filters for compliance theater. It rewards people willing to lie and punishes people who won’t.

And writing makes this collapse faster than photography ever did. Digital photography had hardware friction. Cameras cost money. Gear and technique were specialized. The influx was real, but it had drag. Writing is pure text. AI is pure software. Distribution is instant. That means the “onslaught” isn’t slowed. It’s amplified. Any attempt to choke it out of the professional sphere doesn’t stop the tool. It just pushes serious writers into parallel channels until the gatekeepers are negotiating with the leftovers.

Detection can’t do the job they want it to do

There’s another quiet problem no one wants to admit: enforcement is functionally impossible at scale.

Modern language systems are not crude cut-and-paste engines. They don’t reproduce text. They reproduce register. And register has always been class-coded. Educated writers favor abstraction, nominalization, and conceptual shorthand. Editors recognize that voice and leave it alone. Most readers glide past it without noticing.

I can sometimes hear the tool, not because it copies, but because it leans into educated diction. Words like optics show up where a more colloquial writer might say how it looks or the show. Trained readers recognize the term and approve it. Editors keep it. The public doesn’t blink.

That isn’t plagiarism. It’s linguistic diffusion, the same process by which academic vocabulary has always leaked into journalism and then into culture. Freud did it. Management theory did it. Postmodernism did it. AI just compressed the timeline.

So if “AI detection” becomes “this sounds educated,” you’ve already lost the plot. At that point you’re not detecting a tool. You’re penalizing a register. Which means the rule can’t be enforced consistently, only selectively. And selective enforcement is how trust dies.

Inspiration was never compensable

There’s an irony here that makes the moral panic especially thin.

If a writer watches two strangers argue and uses that emotional pattern in a novel, no one demands payment for “inspiration.” We accept, correctly, that ideas, observations, and experiences are free. What’s protected is fixed expression, not the raw material of thought.

That bargain is the only reason culture functions. If inspiration required compensation, art would freeze under its own accounting.

AI doesn’t change this. It participates in the same space of non-compensable influence that humans always have. The difference is not ethical. It’s psychological.

This is about ego, not ethics

People who spent a lifetime mastering a craft have every right to feel unsettled when the terrain shifts. That discomfort is human. What isn’t defensible is turning that discomfort into policy and calling it morality.

Difficulty endured is not moral entitlement. Scarcity survived is not authorship.

Photography learned this the hard way. The purists didn’t win. They either adapted, redefined themselves, or quietly disappeared while the medium moved on. The ones who survived weren’t the loudest. They were the ones who understood that tools change, but responsibility doesn’t.

Publishing will follow the same arc. Not because AI is inevitable in some abstract sense, but because institutions that confuse ego preservation with ethical clarity always lose relevance.

The only standard that ever mattered

There is exactly one question that has ever mattered in authorship:

Is there a human willing to stand behind the work and defend it?

Not how the draft began.Not what tools were touched.Not whether inspiration was mediated.

But whether a responsible author owns the meaning, coherence, and consequences of the text.

Writing never had RAW files. It ran on trust, accountability, and norms. The sooner publishing admits that, and stops pretending that a blanket ban restores a purity that never existed, the less damage it will do to itself in the meantime.

You can fight this shift. Many will.

But the longer the fight lasts, the clearer it becomes that the resistance isn’t about protecting literature.

It’s about protecting identity.

And identity has never been a winning argument against a change in reality.

Hold the purity line long enough and you don’t preserve literature. You preserve your own irrelevance. When the most responsible writers route around you, what’s left isn’t a protected craft. It’s an empty club, clapping for itself in the dark.

— — — 

This isn’t a defense of deception or plagiarism. It’s an argument that responsibility for the work — not unverifiable process rules — is the only ethical standard that has ever scaled.

Steven Gardner writes essays on culture, technology, and modern incentives.

He has also authored short stories and currently has a novel in the works

More: StevenGardner.com

First published in Generative AI

©2026 Steven Gardner. All rights reserved.