AI-Generated Fanfiction: How to Read Archive Rules and Disclosure Requirements

AO3's position begins with its existing fanwork policy

The Organization for Transformative Works has explicitly addressed AI-generated works in its guidance on AI and data scraping. Its position is that the Terms of Service do not prohibit fully or partly AI-generated fanworks, provided those works otherwise qualify for publication on Archive of Our Own. This applies to machine-assisted prose as well as other creative fanwork formats.

The distinction between permission and endorsement is important. AO3's policy establishes that AI-generated work can fall within the archive's scope. It does not mean that every automatically generated document qualifies as a fanwork, or that all posting practices are acceptable. The archive remains a noncommercial home for transformative fanworks, with its existing Content Policy governing submissions.

The OTW explained this position in its AI and Data Scraping update and subsequent policy communications. Its April 2025 board response also stated that there were no plans to change the rules on AI use. As of October 30, 2026, the published position remains that AI use alone is not a reason to prohibit an otherwise eligible fanwork.

Optional tags give readers information about the writing process

The OTW encourages creators to use descriptive tags that identify AI involvement. Its July 2024 board meeting explanation identifies three relevant tags: Created Using Generative AI, AI-Generated Images, and AI-Generated Text. The first can describe a work with broader generative assistance, while the other two identify specific kinds of content.

For a chapter written with substantial AI-generated prose and subsequently edited by a human author, AI-Generated Text may communicate the relevant contribution. If a story includes generated illustrations, AI-Generated Images may be appropriate as well. A work involving both generated text and imagery can use both tags, provided each accurately describes the published content.

These are additional tags, not mandatory Archive Warnings. AO3 does not classify AI generation as one of the required warnings because the archive says AI use cannot be reliably detected. Its policy explanation notes that a ban or mandatory warning that cannot be consistently enforced would create practical problems. Optional disclosure allows readers to make informed decisions without requiring moderators to determine authorship through unreliable detection.

• Use Created Using Generative AI for broad generative involvement.
• Use AI-Generated Text when generated prose forms part of the work.
• Use AI-Generated Images when generated illustrations are included.
• Explain substantial human editing or rewriting in the notes when useful.

The difference between a generated chapter and a non-fanwork post

An AI-generated chapter must still qualify as a fanwork. AO3's Content Policy describes the archive as a home for transformative, noncommercial works created by fans. The archive permits a wide range of creative formats, including original fiction, fan art, video, and fannish nonfiction. It also allows original stories when they fit its broader mission and do not violate another part of the Terms of Service.

A generated story about an original character can therefore be eligible even when it is not based on an established television series, book, or game. The archive's Terms of Service FAQ explains that original fiction is permitted, subject to its other rules and the requirement that it remain suitable for the archive's noncommercial, fannish purpose.

A collection of unconnected automated text fragments, advertisements, prompts, or placeholders raises different questions. The archive prohibits works that primarily consist of non-fanwork content. A genuine short scene with characterization and narrative development is different from a post containing only a request for others to write a story. The presence of AI does not remove this distinction.

Plagiarism and copyright rules still apply

Permission to use generative tools is not permission to reproduce another writer's work. AO3's 2024 Terms of Service Update Guide explicitly explains that AI-generated works can violate existing plagiarism and copyright policies when they reproduce another work in part or in whole. The guidance gives direct reproduction of entire sentences or paragraphs as an example of a problem.

This means a writer should pay attention to the source material incorporated into generated chapters. If a chapter closely reproduces distinctive passages from a published novel or another fanfiction work, changing a few names does not resolve the underlying issue. Attribution alone also does not automatically authorize reproduction of copyrighted text.

The archive's existing rules provide the relevant framework. A creator who believes another work has infringed their work can use AO3's reporting process. The policy guidance asks reporters to provide links to the original and allegedly infringing material so the relevant committee can assess the complaint.

• Review generated prose for passages reproduced from existing works.
• Check that borrowed material is used within applicable permissions.
• Do not present another writer's text as original writing.
• Use the archive's reporting procedures for suspected plagiarism.

Posting volume and automated comments create separate risks

The OTW has identified spam as a separate issue. Its AI policy explanation says that posting a significant number of generated works in a short period could fall under existing anti-spam rules. The concern is the behavior and volume of posting, not simply the use of a generative tool in an individual story.

The organization has also discussed AI-generated spam comments. Its 2024 Terms of Service Update Guide says such comments remain subject to the archive's ordinary spam rules and can be removed when reported. This distinction matters because a legitimate fanfiction chapter and an automated promotional comment are different kinds of activity.

Writers publishing several chapters should therefore consider the archive's general posting expectations. A series of substantial, distinct chapters is different from flooding the archive with repetitive material or posting large quantities of empty or nearly identical entries. The exact moderation decision remains with the Policy & Abuse committee.

A practical disclosure example for chapter authors

Consider a 3,000-word chapter in which an author develops the plot and character relationships, uses a generative tool to draft two dialogue exchanges, and then rewrites those passages. A useful additional tag is AI-Generated Text if the generated prose remains a meaningful part of the final chapter. A note can explain that the author revised the dialogue and wrote the surrounding narrative.

A different example would be a chapter entirely drafted by a generative tool and lightly edited before posting. Its disclosure should not imply that the prose was primarily written by a human. In both examples, the creator should check the final text for copied passages and make sure the work meets AO3's eligibility requirements.

Disclosure does not need to become a lengthy technical account. A concise statement about what was generated, what was substantially rewritten, and whether illustrations were generated gives readers useful context. Writers can place this explanation in the work notes alongside the relevant optional tags.

Questions readers ask