The rules just changed.
Here’s what every writer needs to know.
The publishing, film, and television industries have spent two years writing policies they never needed before. Literary agents are amending their submission guidelines. Script competitions are adding disclosure requirements. Publishers are demanding statements about how your manuscript was made. And from 2 August 2026, EU legislation requires AI-generated content to carry machine-readable markings.
Whether you write every word yourself, use AI as a research assistant, or collaborate with AI tools as part of your process — the industry is now asking you to prove it.
ScriptShield exists so you can.
You wrote it. Prove it.
For writers who haven’t touched an AI tool — whose process is entirely human, from first draft to final polish — this moment is actually an opportunity.
The problem the industry is wrestling with is that it can no longer tell. AI detectors are unreliable. The slush pile is flooded with submissions that are “superficially articulate” but hollow. Agents are exhausted and increasingly suspicious. In that environment, a writer who can produce verifiable proof of human authorship doesn’t just protect themselves — they stand out.
ScriptShield’s SHA-256 hash system timestamps your work at the moment of upload, creating a tamper-evident record tied to your account. When you generate a Human Authorship Certificate, you’re not just filling out a form — you’re producing a cryptographically anchored declaration that says: this work, in this form, was created by this person, at this time, with no AI involvement.
That’s a different class of proof than anything else in your query packet.
“A number of literary agencies are adding amendments to their submission guidelines urging authors not to use AI in their submissions, as they see a ‘change in the nature of submissions.’”
You use AI tools. That’s okay. Document it properly.
The majority of working writers in 2026 use AI in some part of their process. A grammar pass in Grammarly. Structural feedback from Claude or ChatGPT. Research synthesis. Brainstorming a plot that’s stuck. That’s not a confession — it’s craft in the context of the tools available.
But the industry has drawn a line, and it matters to know where it sits.
Assistive use— using AI to improve, refine, research, or respond to your own writing — is permitted by almost every major publisher, competition, and agency, provided it’s disclosed. SAGE Publishing, Taylor & Francis, and Springer Nature all use this exact distinction. The WGA draws the same line for covered screen projects.
Generative use — using AI to write scenes, dialogue, prose, or significant portions of a manuscript — requires fuller disclosure, and in some submissions is prohibited. The University of Texas Press defines the threshold as five percent or more of total word count. Some script competitions disqualify AI-generated work entirely.
Neither category is shameful. Both require documentation.
ScriptShield lets you record exactly which tools you used, what you used them for, and when — and generates a disclosure statement in the format now required by publishers. This is what Elsevier, Taylor & Francis, and the ICMJE require in submissions. It’s also what script competitions and agencies are increasingly asking for in query packets.
The safe author posture
If AI touched your manuscript, assume disclosure is required unless the journal or agent explicitly says otherwise. Document the tool, the exact job it performed, and the human verification step that followed. Every major publisher policy above rewards that habit; none penalises it.
The documentation landscape — what’s required, where
This is what we know about current requirements across the major submission contexts, with each policy linked to its source. This page is maintained and updated as policies change.
Book manuscripts — literary agents (fiction & memoir)
Most literary agencies do not yet have formal written AI policies, but the direction is clear. Agents are adding disclosure prompts to submission guidelines and reporting a notable change in the nature of submissions they’re receiving. The safest approach for any query submission in 2026 is to include a brief, honest statement about your process if AI tools were used in any capacity.
What to include: A one-paragraph AI usage statement in your query letter or cover note if AI tools were used. If no AI was used, a ScriptShield Human Authorship Certificate provides a level of verification no other tool offers.
Book manuscripts — traditional publishers
Disclose at proposal stage or during writing; statement required in preface/introduction.
Disclosure: Book preface
No AI authorship; document LLM use in the Methods section; AI-assisted copy editing exempt.
Disclosure: Methods section
Permit AI for language improvement with oversight; declaration required in a dedicated section.
Disclosure: Declaration section, before references
Assistive AI (refining your text): not required to disclose. Generative AI (new content): must be cited and referenced.
Disclosure: Acknowledgements or Methods
No AI authorship; disclosure required for any AI use beyond basic editing.
Disclosure: Author statement
Work substantially written by AI (5%+ of total word count) will not be accepted; disclosure required.
Disclosure: Editor communication at submission
ICMJE standard (academic presses, general)
Increasingly adopted across university presses — disclose any AI writing assistance at submission.
Disclosure: Acknowledgements
Screenplays — WGA covered projects
The Writers Guild of America’s 2023 Minimum Basic Agreement established the foundational framework for AI and screenwriting on covered projects:
- AI is not a writer. Material produced by generative AI cannot be considered literary material under the MBA.
- Writers may use AI tools as they see fit, subject to company policy — but cannot be required to use them.
- Any AI-generated material presented to a writer must be disclosed to that writer by the studio.
- AI cannot receive writing credit.
Practical implication for spec writers:If you’re submitting a spec script to WGA-covered agents or managers, your script needs to be your own literary material. If AI generated any scenes or significant prose, you need to be aware that this creates a complication in the covered project system.
Screenplay competitions
Policies vary significantly. Most major competitions have moved to one of three positions:
- Disclosure required (dominant approach): submit with an AI usage statement if AI was used in writing. The Austin Film Festival, Sundance Labs, and most major competitions fall here.
- Human authorship required: the script must be the original work of the submitting writer. AI-generated prose typically disqualifies.
- No formal policy(still common at smaller competitions): the “original work” clause in the rules is the operative standard.
When in doubt: treat any competition as requiring human authorship unless the rules explicitly permit AI-generated content.
Film, TV & streaming — industry submissions
Hollywood agencies increasingly use AI tools internally for submissions triage, but for writers, the expectation remains that submitted work is yours. California AB 412 would require disclosure of copyrighted works used in AI training datasets, and SAG-AFTRA’s agreements extend consent and disclosure requirements to AI-generated performances.
The general industry posture: AI in the room is fine; AI as the room is not.
Your permissions — what you control downstream
Separate from how your work was made is the question of what others may do with it using AI. These are two distinct things, and ScriptShield treats them that way.
When you protect a work with ScriptShield, you can set explicit permissions governing AI use of your content:
Training data
Can your work be used to train AI models? The default answer for protected work is no. If you want to explicitly permit this (some writers do, particularly in open licensing contexts), you can enable it per work.
AI adaptation
Can AI systems generate adaptations, continuations, or derivative works from your writing? Again: no by default.
AI translation
Can your work be machine-translated by AI-powered translation services? This matters for writers considering foreign rights.
AI summary
Can AI systems generate commercial summaries of your work? This has implications for licensing and for how your work appears in AI-powered search and discovery systems.
These settings are recorded in your work’s metadata and included in any certificate you generate. They are not legally binding contracts — they are a clear, documented statement of your intentions, which is increasingly what agents, publishers, and rights lawyers are looking for as the first line of any IP discussion.
This direction is aligned with the broader provenance standards now being built across the industry. The C2PA content-provenance standard — the Coalition for Content Provenance and Authenticity, whose members include Adobe, Microsoft, the BBC, Google, and OpenAI — defines how tamper-evident “content credentials” travel with a file. ScriptShield’s hash-based certification is architecturally aligned with that same goal: a verifiable record of where a work came from.
Why this matters going forward
The regulatory direction is one-way. EU AI Act Article 50 enforcement begins 2 August 2026 — requiring machine-readable marking of AI-generated content across all sectors. California AB 412 is moving through the legislature, and New York’s synthetic-performer disclosure law took effect in June 2026.
This is not a publishing-specific issue any more. It is a legal and commercial infrastructure issue.
Writers who have documentation of their authorship, their AI use (or non-use), and their permissions are in a fundamentally different position than writers who don’t — when it comes to submissions, rights negotiations, copyright disputes, and AI training data litigation. The U.S. Copyright Office has confirmed that copyright protects works of human authorship, assessed case by case — which makes a documented human record more valuable, not less.
ScriptShield started as a tool to prove you wrote it first. This is the natural extension of that: proving howyou wrote it, and what you will and won’t permit to be done with it.
The record exists. The certificate is yours. Use it.
How ScriptShield handles this
Your authorship record
Every work you upload is hashed with SHA-256 and timestamped at the moment of upload. This hash is the anchor for everything else. It cannot be altered after the fact. If anyone disputes when you created your work, or what it contained, the hash is the evidence.
Your AI disclosure record
You set your authorship mode — Human-authored, AI-assisted, or AI-generated — per work. If you used AI tools, you log them: tool name, version, purpose, date. This creates the structured disclosure record that publishers and agents are now requiring.
Your permissions record
You set your downstream AI permissions per work. These are stored in your account and included in any certificate.
Your certificate
When you are ready to submit — to an agent, a competition, a publisher — you generate a certificate: a clean, professional PDF that includes your work’s title, your SHA-256 hash, your authorship declaration, your AI tool log (if applicable), and your permissions statement. Dated, versioned, and tied cryptographically to your uploaded file.
This is what you attach to your submission packet. This is what you keep on file. This is the paper trail that matters.
References & sources
Every policy, law, and standard cited above links to the issuing organisation’s own page. Where industry practice is summarised rather than quoted from a single source, we say so. If you find a reference that has moved or changed, let us know and we’ll update it.
- 01Taylor & Francis — AI policy for authors
- 02Springer Nature — AI editorial policy
- 03Elsevier — Generative AI policy for authors
- 04SAGE Publishing — AI author guidelines
- 05Wiley — AI guidelines for authors
- 06University of Texas Press — Generative AI policy for authors
- 07ICMJE — Recommendations: AI use by authors
- 08Writers Guild of America — Artificial Intelligence (2023 MBA)
- 09SAG-AFTRA — Artificial Intelligence
- 10EU AI Act — Article 50, transparency obligations
- 11California AB 412 — Generative AI: training data: copyrighted materials
- 12New York — Synthetic performer advertising disclosure law
- 13U.S. Copyright Office — Copyright and Artificial Intelligence
- 14C2PA — Coalition for Content Provenance and Authenticity
- 15The Bookseller — “Literary agents urge writers to avoid AI as they see ‘change in nature of submissions’”
Publishing resources
- BooksBy — independent publishing and book distribution (booksby.com).
Protect your work. Own your record.
This page reflects current industry policies as of June 2026. The landscape is changing rapidly. We update this page as major publishers, agents, guilds, and festivals update their requirements.
Last reviewed: June 2026
This page is educational and does not constitute legal advice. ScriptShield is not a law firm. Consult a qualified attorney in your jurisdiction for advice on your specific situation.