The Pledge: AI, Authorship, and the Fight for Human Creativity

Why a single sentence on a submission form opens up the most uncomfortable question in creative writing today.

A statement worth deep-diving into

I came across a clause recently, buried in a submission form for an upcoming Horror Fiction Anthology. It read:

"I confirm that the story submitted with this form is my original work and has not been generated or substantially written by artificial intelligence. I understand that submissions may be screened for AI-generated content and may be rejected if this requirement is not met."

It looks innocuous. It is not. That single paragraph is a legal and ethical pledge that you, the human, are the sole author of your story, and it is now a useful guideline for every writer working in fiction in 2026. So I wanted to deep-dive into what a creative person is actually up against in this new age of AI hysteria, and what that pledge really demands when you sign your name to it.

In short, signing it means agreeing to three things:

Total ownership. The story must be your original work. You did not plagiarise it from another author, a book, or a website. You are the creator of the plot, the characters, and the prose.

No AI involvement. You did not use a chatbot like Gemini, ChatGPT, or Claude to generate the story. You did not have AI substantially rewrite your drafts. Most places allow simple spell-checkers like Grammarly or Microsoft Word's built-in editor, but using AI to change the voice, structure, or flow of the writing breaks the rule.

Acceptance of screening. You give the editor or publisher permission to run your story through AI-detection software. They are telling you up front that if their tools flag your work as machine-generated, they have the right to disqualify or reject your submission immediately, with no further discussion.

The bottom line is simple: if you wrote a story yourself using your own ideas and words, you have nothing to worry about. If you used AI to help you write it, submitting would be a violation. But the interesting work begins when you ask, where exactly is the line?

Where the line actually sits

Two phrases in that pledge do most of the heavy lifting: generating the story, and substantially rewriting the drafts. They sound similar. They are not.

Generating the story

Generation is the most serious violation, because the AI is acting as the author rather than as an assistant.

The most obvious form is the one-prompt story: a writer types "write a 500-word story about a lonely astronaut" and uses the output. But generation also covers the Lego method, where a writer asks AI to write individual scenes or paragraphs and then stitches them together. Even with later edits, the substance was generated by the machine.

Strict definitions go further. Plot and structure generation — asking the AI to produce a detailed chapter-by-chapter outline for you to follow — is considered generating the story's DNA, even if you write the actual sentences yourself. The rule of thumb: if the AI is thinking up the ideas, dialogue, or descriptions for you, it is generating the work.

Substantially rewriting the drafts

This is where most people get tripped up. It refers to taking a draft you wrote and asking AI to fix it or improve it to a point where your original voice is lost.

A style mimicry request — "rewrite this in the style of Hemingway," "make this sound more professional and descriptive" — makes the AI the primary stylist. A structural overhaul — "take this scene and make it more dramatic" — typically results in the AI adding actions, dialogue, and emotional beats that were not there before. It has substantially rewritten the work.

The accepted comparison is human polish versus AI overhaul. Correcting a typo or a misplaced comma through a standard spell-check is fine. Asking AI to "rewrite this paragraph to be more engaging" is not. And it matters not just ethically but technically: AI-detection tools look for patterns in the syntax and the burstiness of the rhythm of sentences. When an AI rewrites your work, it applies a mathematical pattern to your words that detectors can spot, even if the original idea was yours.

The simple boundary

• Likely allowed: using AI as a sounding board for brainstorming, while you write the text. Basic spell-checking that fixes errors without changing your creative voice.

• Definitely not allowed: using AI to make stylistic and creative choices, "rewrite this scene" prompts, or anything where the AI is acting as the creator.

The safest mental model is to treat AI like a research librarian for facts, not a ghostwriter for prose. If you find yourself clicking copy-paste from the chatbot into your story, you are likely crossing the line into substantial rewriting.

"But the AI is just doing the graft" — the productivity argument

Here is the question I keep seeing on writing forums:

Is using AI to generate copy text that the human edits and rewrites later considered using AI to create something — even though the plot, the roles, and the creative thinking all came from the human? The AI is just doing some of the graft, increasing the author's productivity and output.

In a corporate or professional setting, this logic is unimpeachable. In creative writing, publishers, contests, and readers see it very differently — and that specific process would still violate the original-work requirement.

The pledge specifically forbids work that is substantially written by AI, even if you edit the text later. If the AI provided the initial copy, it provided three things that matter enormously:

1. The sentence structure — the rhythm and length of your sentences.

2. The word choice — the specific adjectives and verbs used to describe a scene.

3. The flow — the way one idea transitions to the next.

When you edit an AI's draft, you are transforming its work. But the AI still did the substantial writing. In a literary context, the graft is the actual act of putting words on the page. The graft is the art.

In most creative fields, the goal of the no-AI rule is not to help you be more productive. It is to ensure the work is pure human expression. Many contests want to reward the struggle of finding the right word; using AI to bypass that struggle is unfair to authors who did the work manually. And the slip of the thesis problem haunts every "I edited it heavily" defense: if an AI writes 100 words and you change 40 of them, is it really your work? Most editors would say no — it is an AI-human collaboration, which fails the original-work test.

There is also the detection risk. Even if you believe your edits have made the story yours, AI-detection software is designed to find the underlying patterns of the original AI draft. If the AI scaffold remains, the detectors will likely flag it. The rules say works may be rejected on the screen alone. Nobody will ask you how much you edited; they will simply see the high probability score and move on.

So is there a better way to use AI for productivity? Yes — by changing when you use it. Use it to research ("what flowers grow in Ireland in May?"). Use it to brainstorm ("give me five reasons a character might be late to a wedding"). The red line is the moment the AI starts writing prose that ends up in your story. If you have signed a no-AI pledge, you should write every sentence from a blank page. If you used AI to generate copy first, you are technically making a false statement when you sign.

The originality war: a moral, not legal, question

Take the legalities out of it. Look at the moral ethics of authorship.

In creative work, the graft is not the grunt work to be optimised away — it is the act of creation. Even if the plot and ideas are all yours, if AI generates the initial blocks of text, you are not editing — you are curating. There is a fundamental difference between a writer staring at a blank page and a writer reacting to a screen full of text.

The first draft is the DNA. Writing is often described as thinking on the page. When an AI generates the first draft, it makes thousands of micro-decisions for you: where to start a paragraph, which metaphors to use, the cadence of dialogue. Even if you rewrite every sentence later, your brain is anchored by the AI's choices. You are fixing its work rather than building your own. If the AI laid the bricks, it built the house — even if you painted the walls and moved the furniture in.

The definition of "author." The word author shares a root with authority. It means you are the source. If you tell a ghostwriter a story and they write it, and then you edit it, the ghostwriter is still the writer. AI is a digital ghostwriter. Using AI to generate text to increase productivity is exactly what people mean by "AI-generated."

Translation versus creation. If you have a plot and a scene in your head but use AI to put it into words, you are essentially asking the AI to translate your thoughts into English. In a creative or academic field, the entity doing the translation is a co-creator of that specific version of the text.

In a casual or business setting — writing an email or a report — nobody really cares; they just want the result. But in creative writing, the graft is the whole point. If you did not wrestle with the sentence until it was right, the consensus is that you did not write it. You directed it.

Why AI is different from a pen

A reasonable rebuttal: surely Microsoft Word also "co-creates" with the author? The same can be said about a pen and paper — instruments are tools that participate in the work. So why is AI considered different? Is it because AI thinks?

It is a fascinating philosophical debate, but the world and the law treat AI differently from a pen for one reason: agency versus utility.

A pen and a word processor are passive tools. The pen has no opinion; it does not suggest a better word; it leaks ink only where you steer it. Microsoft Word provides a digital canvas. Even with spell-check enabled, it is comparing your work against a static dictionary. It does not know what the story is. It just knows how to flag a typo.

AI is an active agent. When you use AI, you are not using a tool to record your thoughts. You are using an engine that makes probabilistic creative choices — choosing adjectives, sentence rhythm, and metaphors based on a mathematical model of human language.

The difference also lies in where the spark originates. With a pen or with Word, the specific sequence of words is generated in your brain and transmitted through your fingers; the tool is a medium. With AI, the spark is generated by the model's weights and biases. Even if you give it the wood, the AI builds the fire.

There is also a labour-versus-output distinction worth naming. In intellectual property and creative arts, the graft and the creation are inseparable. If I give a sculptor a block of marble and say "carve a man holding a spear" and the sculptor does the carving, the sculptor is the creator. I am merely a patron. Microsoft Word is a chisel — it does nothing without your force. AI is a hired sculptor — you give it the idea and it performs the labour that produces the final form.

It is not wrong to call AI a tool. It is a tool that operates at a higher level of abstraction. A pen helps you write. AI writes for you. In a world of original work, that shift from doing to delegating is what makes it fundamentally different.

The painful paradox: reclaiming the labour

Here is where the conversation gets uncomfortable. Some writers take the stance: AI was built and learned how to simulate the creative process because AI companies illegally scraped millions of books, including some of mine. So if I use AI to do some of the graft, am I not just reclaiming my own labour?

This is probably the most painful paradox of the current AI era — the test-to-service loop. If your work was used to train these models without your consent, your perspective on using them for your graft is a recognised philosophical stance.

The reclaiming-your-labour argument. From the author's perspective, you are not cheating; you are reclaiming. If a company took your data, your style, your words, your graft to build a machine, then using that machine to speed up your work feels like taking back your stolen time. It is like a carpenter whose tools were stolen, only to find a thief has built a robot carpenter using his techniques. The original carpenter might feel he has every right to use that robot to finish his projects.

The dilution of the self. The "ripping yourself off" feeling comes from how AI processes data. AI does not store your book; it turns it into a mathematical average alongside millions of others. When you ask it to help with the graft, it is not giving you your specific genius back — it is giving you a homogenised version of your genius, mixed with the genius and the mediocrity of everyone else. You are using a filter that has averaged out your own soul.

The stock photo comparison. Picture a professional photographer. An AI company steals their photos, builds an AI image generator, and then the photographer uses that AI to generate quick backgrounds for new work. Technically, yes, they are being more productive. Ethically, they are participating in a system that devalues their own craft. Legally, if they submit that work to a contest with a "no AI" rule, they will still be disqualified — regardless of the fact that their data helped build the machine.

Moral injury. There is a term in psychology — moral injury — that many creators feel here. It is the sense that you were forced to use a tool built on the corpse of your own industry just to keep up with the new speed of production. If the AI helps you write faster but the resulting prose feels slightly less than what you would have done by hand, you are indeed ripping yourself off — not just of legal originality, but of the pride of authorship.

The brutal reality: AI is different from a pen because the pen never ate your books to learn how to leak ink. But the gatekeepers setting the rules do not care about the fairness of how the AI was built. They care about the purity of the output. Even if you are taking back what was yours, they will see a machine doing the work of a human. It is a no-win scenario for the creator whose work was used to build the very thing they are now competing with.

How the industry is fighting back in 2026

The protection of human creativity has become one of the most intense legal and technical battles of the year. Three fronts are worth watching.

1. Proof-of-human infrastructure

Since AI can simulate a voice or a style convincingly, the industry is shifting from detection toward provenance — tracking where a work came from rather than guessing what made it.

• C2PA standards. Many publishers and camera manufacturers have adopted the Coalition for Content Provenance and Authenticity standards. This is essentially a digital paper trail proving a file was created by a specific human on a specific device at a specific time.

• The human-made seal. Similar to organic labels on food, 2026 has seen the rise of verified human-only platforms. Organisations like the Human Artistry Campaign advocate for labels stating no AI was used in the graft of the work.

• Watermarking requirements. Under the EU AI Act, fully rolling out this year, AI companies are legally required to embed invisible watermarks into their output. This makes it harder for someone to simulate creativity and pass it off as human-made without being caught by screening tools.

2. Legislative shields

Governments are finally moving past the wild-west phase of AI scraping.

• Rebuttable presumption of infringement. A major 2026 European Parliament resolution proposes that when an AI can mimic an author's style, it is presumed to have used that author's work illegally — unless the AI company can prove otherwise. The burden of proof shifts from the writer-victim to the AI company.

• Opt-out lists. Systems managed by the EUIPO (European Union Intellectual Property Office) now allow authors to officially opt out of AI training. If a model is caught using your work after you have opted out, the penalties are becoming massive.

• Non-copyrightability of AI output. Courts have largely held the line that AI-generated content cannot be copyrighted. This is a huge protection for authors: a company cannot use AI to churn out 10,000 books and own them. Only a human can hold the rights, which keeps the economic value tied to human labour.

3. Poisoning the well — guerrilla warfare

Authors and artists are using technology to fight technology.

• Glaze and Nightshade. These tools mask digital files. If an AI tries to scrape a book or image protected by them, it breaks the AI's understanding of the style — effectively teaching the model that your writing style is garbage, making it impossible to simulate you.

• Data unions. Groups of creators are banding together to negotiate collective licensing. Instead of one author fighting a tech giant, thousands act as a block, demanding payment before a single word is used for training.

How a single author can realistically protect their work

If you are submitting work today, the best protection is to keep your receipts.

Version history. If you use Word or Google Docs, keep the version history on. It proves the story grew sentence by sentence over hours and days, rather than appearing in one copy-paste block.

Avoid the average. AI simulates the middle of human expression. The best protection is to lean into your most uniquely human quirks — your weirdest metaphors, your most specific personal memories, your non-linear logic. The machine is built to find the centre. Live at the edges.

The blank-page rule. If you want to be 100% safe from rejection, never use AI to generate the first draft. Use it for research. Use it to check whether something is factually correct. But never let it lay the first brick in the wall.

The industry is currently building a human-only moat. It is an exhausting time to be a creator, but the goal is to make sure that simulated creativity never carries the same legal or financial weight as the real thing.

And what about the publishers?

Here is where the power dynamic gets tricky. As an author, you are being asked to promise you did not use AI. Meanwhile, publishers — under immense pressure to cut costs — might be using AI behind the scenes to summarise, market, or even polish your work. To protect yourself in 2026 and beyond, you have to move from trust to contractual and technical enforcement.

The no-AI-training clause

The most critical protection lives in your contract. Major organisations like the Authors Guild and the Society of Authors have released model AI clauses every author should now insist upon.

• Prohibit ingestion. The contract should explicitly state that the publisher is not permitted to use your manuscript to train internal AI models or to upload it to a third-party platform like ChatGPT or Claude for any reason.

• Human-only editing. Request a clause specifying that all substantive editing must be performed by a human. This prevents the publisher from feeding your book into an AI to tighten its prose or fix its pacing without your consent.

• The Penguin Random House precedent. In late 2024 and 2025, major publishers including Penguin Random House began including AI-tooling language in their standard contracts. Read the General Rights and Operating Procedures section carefully — it is easy to accidentally hand over blanket permission to process your work through AI.

Guarding your subsidiary rights

AI has created a whole new catalogue of rights that did not exist five years ago. Treat them like film or foreign-language rights.

• Voice licensing. Explicitly forbid AI-generated voices for your audiobook unless you have right of approval and a fair share of royalties.

• AI translation. Ensure the publisher cannot use AI translation — which is often lower quality — to fulfil their obligation to publish in other territories without your express permission.

• Summary and RAG rights. Some publishers are now licensing books to tech companies for Retrieval Augmented Generation — letting an AI read your book to answer user questions. Reserve your rights to negotiate a high royalty split. The Authors Guild suggests between 50% and 80%.

Technical self-defence

If you are submitting to a small press or an agent you do not yet fully trust:

• Digital watermarking. Tools now exist to embed invisible noise or metadata into the text file. If that text is later used to train an AI, the watermark can be detected in the model's output, giving you proof of copyright violation.

• The poisoned manuscript. More common in art, but emerging in text — tools like Nightshade can scramble metadata to make it difficult for AI scrapers to derive meaning from a file, while leaving it perfectly readable for a human editor.

The transparency question

Before signing, ask your editor directly: what AI tools do you use in your workflow? Many editors use AI for boring graft — generating metadata, alt text for images, or catalogue copy. If they admit using it for editorial feedback, you have a right to know. An AI critique is based on patterns of successful books, which often strip away the very weirdness that makes your book original.

The bottom line: if they are making you sign a human-only pledge, it is entirely fair for you to demand a reciprocal human-only pledge from them. The relationship between author and editor should remain soul-to-soul, not soul-to-software.

Are creative works safe in public cloud services?

This is the billion-dollar question for creators in 2026. The honest answer is mostly yes, for now — and stay vigilant.

Public cloud services like Google Drive, Dropbox, and Microsoft OneDrive operate under different rules than the public internet. The single most important step is to check your settings and turn off the intelligence.

For example, in Google Workspace, an AI feature called Workspace Intelligence can read your files so it can answer questions like "summarise my document." It is convenient. It can also be turned off. Find that setting. Turn it off. Repeat the exercise on every cloud provider you use.

Final thought

A single sentence on a submission form has, in 2026, become a window into the largest unresolved question in creative work: what does it mean to be the author of a thing?

The pledge demands that you, the human, are the source. The temptation to delegate the graft is enormous, the productivity case is real, and the moral irony — that the machine was built on the back of work just like yours — is genuinely painful. None of that changes the answer...

In creative writing, the graft is not the cost of the art. The graft is the art.

If you wrote it yourself, you have nothing to declare and everything to be proud of. That is, in the end, the whole point of signing your name.