Rules Preview

01

Eligibility

The awards honor excellent fiction made with machines in the process. A work is eligible if it:

• Is fiction, in any genre, subject, or form.
• Used a machine in its making, disclosed honestly.
• Is original to the entrant, or the entrant holds the rights needed to enter it.
• Is available in English, in the original or in translation.
• Is unpublished, or was first published in any language between January 1, 2025 and the close of submissions. Self-publishing, including Amazon KDP, counts as publication.

There is no minimum or maximum amount of machine-generated text. Fully generated, partially generated, heavily edited, and lightly edited work are all eligible. At least one named human author must direct the work, review it, and accept responsibility for it.

There is no subject restriction. A work does not need to be about artificial intelligence, or to reference it at all.

Entry is open worldwide, except where U.S. sanctions law prohibits participation. Simultaneous submissions to other competitions are allowed.

Fan fiction and unauthorized derivative work are not eligible. Founders, staff, administrators, and anyone serving as a reader, reviewer, or judge in the cycle, and members of their households, may not enter.

02

Categories and length

Length is counted in the English text and includes all narrative text, including epigraphs and in-world documents. It excludes the title page, dedication, acknowledgments, author’s notes, and synopsis.

Each author may enter one work per category per cycle: at most one Short Fiction work and one Longform work. Submit your best work.

03

Required materials

• The work: the complete story or manuscript, as PDF, DOCX, EPUB, Markdown, or plain text, with no identifying information. Files must be static: no macros, scripts, password protection, or links to outside websites. Internal navigation, such as a table of contents or “turn to chapter 12,” is fine. Submit in the format that best honors the work. If layout, typography, or illustrations are part of it, use PDF. For prose alone, Markdown, plain text, or DOCX serve it well.
• Synopsis (Longform): up to 1,000 words covering the whole work.
• Author statement: 500–1,000 words on creative intent and your relationship to the process.
• Process dossier: a structured account of how the work was made. See the dossier guide.
• AI disclosure: the systems you used and what they did, as part of the process dossier.
• Attestations: rights to enter, accuracy of disclosure, human responsibility, eligibility.

Prompts, drafts, logs, repositories, and diagrams are welcome but never required. A clear written account is enough.

04

Translated works

A work written in another language may be entered in English translation. Translation by the author with machine assistance is permitted; professional or certified translation is not required. Readers judge the English text, so its quality matters.

• At entry: the English opening (Longform: the first 10,000 English words), an English synopsis, and the complete manuscript in the original language.
• If the work advances past the first reading: the complete English translation is due within 45 days.
• If the complete translation is not delivered on time, the entry is withdrawn and the fee is not refunded.

05

Authorship and credit

Co-authored work is welcome. Every entry has at least one named human author who takes responsibility for the work. Everyone credited on an entry is identified privately to the awards and confirms the entry. Public credit may use any name the author chooses, including a pen name; legal names are never published without permission.

An entry counts toward the one-per-category limit for every person credited on it.

06

Youth entrants

Authors under 18 may enter Short Fiction with additional consent and privacy safeguards.

• A parent or legal guardian must submit, consent, and sign all attestations.
• The awards will not publish a minor’s legal name, contact details, school, location, private notes, or process materials.
• Minors may use a public author name or initials.
• Guardian approval is required before publication of any finalist summary, excerpt, author bio, or process note.

Eligible youth entries are considered for the Future Scriptorium Citation, which is awarded only if eligible youth entries are received.

07

Judging standard

All judging is done by people. The first question is whether the work succeeds as fiction: did it move, provoke, unsettle, delight, or stay with the reader?

1. Administrative screen. Completeness, eligibility, and file safety. Not a judgment of quality.
2. First reading. Independent readers read Short Fiction in full, and Longform openings with the synopsis, and decide whether each work earns a full read. Readers do not see the author’s identity or process materials.
3. Full reading. Advancing works are read completely and scored as fiction.
4. Process review. Leading works are assessed on their process dossiers.
5. Shortlist. The panel selects finalists, with written reasons.
6. Final panel. Final judges read every finalist in full and select winners.

Longform is first read by its opening. Readers start with the first 10,000 words and the synopsis. If the opening doesn’t earn a full read, the rest of the manuscript is not read. Readers won’t dig for a good book under an opening that isn’t enjoyable to read; that is part of the standard. Make your first 10,000 words count.

Every score band, gate, and decision rule is published in How we judge.

Made with AI is expected. Unedited isn’t. Readers judge whether the text has been worked: readable, controlled, edited. Unrevised machine habits (a pattern repeated until it’s the loudest thing on the page, stock phrasing, reflexive structures) are craft failures like any other, and heavy, unworked output is declined at the first reading. Deliberate use is different: a character who talks like a machine, or a machine-flavored style that is chosen and controlled, is judged on whether it works. Nobody is expected to run text through a humanizer, and we do not use AI-detection tools.

Process is not scored on how much a machine contributed or on technical complexity. A plain-language account can earn full marks. Two process requirements are pass/fail: the dossier must show that a machine took part in making the work, and sources, permissions, and the privacy of real people must be respected. A work needs at least 8 of 20 process points to become a finalist. Why we score process at all.

08

Awards and prizes

The In Silico Grand Award; a winner in each category; the Galatea Citation, for a genuinely inventive method, at the panel’s discretion; and the Future Scriptorium Citation, if eligible youth entries are received. The panel may combine, reduce, or withhold any award.

The awards are not run for profit. Entry fees are spent in this order:

1. Card processingStripe’s standard fee comes off first.
2. ReviewReaders, judges, and process reviewers, paid per assignment; plus secure file storage and screening.
3. PrizesEverything that remains goes to the prize pool.

Prize amounts are announced when submissions close, once the cost of review is known. Prizes are paid internationally; winners are responsible for taxes in their own country.

09

Submission integrity

Misrepresentation, not machine generation, is the offense. An entry is rejected without refund for:

• Any attempt to instruct, manipulate, or bypass reviewers or review tools: hidden text, instructions addressed to software, adversarial metadata, filenames, or comments.
• Malicious files or links.
• Undisclosed or misrepresented machine use, or a process account found materially false.
• Fan fiction or unauthorized derivative work.
• Plagiarism or knowingly infringing material.
• Direct imitation of a living author without permission.
• Knowingly false rights or eligibility attestations.
• Spam, bulk, or content-farm submissions; harassing material; false identity.

Attempts to manipulate review tools also bar the entrant from all future cycles. The awards reserve all legal remedies. We do not use AI-detection tools. Integrity concerns are reviewed by two people, and where a concern is factual and curable, the entrant may respond. Minor problems, such as a missing section or unreadable file, can be corrected within 5 days.

10

Rights, privacy, and recognition

Entrants keep copyright and all publication rights. Entering grants permission only to review the work privately, announce finalists and winners by title and public name, and publish excerpts or process summaries the entrant approves. Winning does not require any publishing agreement.

Manuscripts, dossiers, and evidence are confidential. For non-winning entries they are deleted 90 days after winners are announced; a review record of verdicts, scores, and reasons is kept for 3 years.

Recognition. Finalists and winners are announced by title and public author name. Where a finalist’s or winner’s work is published, including after the awards, we may link to it and identify it as an In Silico finalist or winner. Authors choose where we link, or can decline the link; we never use affiliate links. Finalists and winners may also use the finalist or winner designation on their work. Entries that do not become finalists are never identified.

11

Fees, refunds, and appeals

Fees are charged in U.S. dollars by card, with no international surcharge. They are non-refundable once first reading begins, except for duplicate payments, errors by the awards, or a cancelled category (refund or transfer). Limited fee waivers are available on request and reviewed by a person; supporters can sponsor another writer’s entry to fund them.

Appeals are accepted within 14 days of notice, only for procedural error or wrongful exclusion on eligibility grounds. Literary verdicts and scores are not appealable.

12

File security and AI

Every file is security-checked with standard software before anyone opens it, and any active content (scripts, macros, embedded programs) is removed. Readers see the work's text, formatting, and illustrations as submitted. We never run the work itself through an AI model. If a process dossier is technical enough that a reviewer needs help understanding it, we may use an AI tool to help read the dossier; that is the only use of AI. Machines never evaluate, score, rank, or reject an entry, or choose a finalist or winner; people make every decision. By submitting, entrants consent to this.