Rules for the usage of AI in stories published on the sites
Version 2025-Nov-11
Preamble
The previously established AI rules, which required tagging stories with specific AI tags, aimed to protect the site, the admin team, and readers from an onslaught of AI-generated stories created quickly with simple prompts. Also, some readers expressed a desire to see beforehand whether a story had been created using AI.
This approach has worked well, but it also caused some trouble. Some authors were quite angry when the approver’s team detected any AI usage in their stories. In other stories, we missed the AI involvement, and readers got angry, too.
That’s why we decided to revise the rules.
Principles for our approach to AI involvement
-
We still, in a way, evaluate whether AI was used or not for the story, but just as a means of validating our feelings about the prose itself. In the end, we judge not based on AI involvement but the AI-likeness of the prose.
-
We don’t want to involve the site in anti-AI activism or spend resources to facilitate such things, but we do want to encourage authors not to frustrate readers. However, we don’t want to police terms that should remain between an author and a reader. Any sense of entitlement by readers to AI-use disclosure by authors makes no sense when the stories are offered for free. It is up to the author to choose to disclose or not. Still, we will strongly encourage authors to tag. If an author “tricks” a reader into reading something they don’t want to, then the reader can just say “Rude!” and never read another story from that author again.
-
We do not focus on how a story is created, which is impossible to truly determine anyway; we focus on the content of the story itself.
-
We do not focus on aspects of quality that have to do with the skill of a human writer. We want inexperienced authors to have their voices heard, too, and apart from cases where it makes their stories unreadable (which we already screen for), we don’t care about unskilled writing.
-
We focus instead on aspects of quality related to the soul and humanity in the story, things that make it feel too mechanical, uncanny valley, utter crap that feels like stock prose copied from other sources and recombined. This is what we want to avoid on the site.
Three types of stories and a pragmatic way to evaluate them
Stories that are clearly almost all unedited AI prose, effortlessly coming out of a few prompts.
It must be so clear that approvers in our team can see that the story has no place on the site because it seems like something coming out of a factory. If there is any hesitation, and only in that case, we will test them using two AI detectors, and if both indicate the story is more than 50% likely to be AI, the categorization will be confirmed. THOSE STORIES WILL BE BANNED OUTRIGHT.
Stories that have a substantial amount of AI prose but clearly were created and edited by an author, that retain a spark of humanity and soul, and show that some effort has been made to edit them, that there is a human creator behind.
It must be clear to a trained approver that the story contains a significant part of AI-generated prose. FOR THOSE STORIES: We don’t publish right away, but message the author saying we encourage them to EITHER use the ai prose tag OR add a disclaimer at the top of their story detailing in their own words their usage of AI, but that we will not enforce it. We give them reasons to do so, like the idea that it would protect them from nasty comments, and many readers would enjoy the option to opt out. Once the author responds, we accept their choice to tag or not and publish.
Stories that show no obvious AI prose or signs of AI generation.
AI could still have been used in many ways in their creation process, but it is impossible to tell with any certainty from the content itself, and thus should not bother the reader, and even if it did there’s nothing we can do anyway because nothing can be proven. THOSE STORIES ARE APPROVED AS NORMAL.
Tagging
The only type of AI use that will be tagged is the only one that can be detected easily by readers and approvers. It’s the one that bothers readers the most, so it’s the most useful one to filter. We will use the “contains ai patterns” tag to indicate that the story contains passages that read like AI generated text with its signature features.
Please note that this is not an actual change to the rules: the tag “ai assisted” was previously used only in those cases, despite what its name might have suggested, because that was the only use we could detect. For that reason, the tag “ai assisted” has been changed to “contains ai patterns” to make things clear.
Authors will be encouraged to use it when it fits. That is, when the story has a substantial amount of AI signature constructions, yet still retains a spark of humanity and soul, and shows that the author has made some effort to edit and reshape it.
If an author wishes to disclose any other use of AI in their story, they can do so by putting a disclaimer at the top of their story.
The tag “ai created” will no longer be used because we will no longer publish on our site any story that would have it.
The published stories that have any AI tag will get the tag “contains ai patterns”.
Neither authors nor readers can add or remove the tag “contains ai patterns” after a story has been published. If an author wants the tag added or removed, they have to ask the staff to do so.
Comment moderation
We consider any outright accusation of having used AI in a story as extremely hurtful to the author and unprovable with any certainty. Comments that make these accusations shall be removed because of the harm they would cause. Commenting that the prose sounds like AI, preferably accompanied by examples, can be useful, however, and will be accepted