Hello.

Last week, it was brought to the GM team’s attention that there was a rumor a player was using generative AI to produce their work.

As a GM Team, we have previously discussed implementing a hard rule regarding the use of AI. However, conversation frequently would result in questions of implementation, actual language of the rule, enforceability, and how to make a policy that didn’t require extensive detective work, trying to puzzle out whether a writer was using generative AI and whether it was or was not in a context that might or might not be acceptable.

We’ve discussed and debated using different AI checkers, different thresholds for acceptable AI usage or percentage scores. And that’s left us here, at this moment in time, without a clear AI policy.

And for that, I apologize.

Going forward, we’ve updated the Rules/FAQ section of our website.

Old Rule:

Quote:Plagiarism of any kind is strictly forbidden. Do not have someone else write your role play. Do not steal a role play or any piece of writing you found on the internet. And no, you cannot reuse role plays from your old fed. Doing this will result in you getting a hefty ban.

New Rule (with added sections in bold):

Quote:Plagiarism of any kind is strictly forbidden. Do not have someone else write your role play. Do not steal a role play or any piece of writing you found on the internet. Do not have AI generate the contents of your role play, either the entire role play or any scene(s) of it.  And no, you cannot reuse role plays from your old fed for competition purposes. Doing any of the above will result in you getting a hefty ban.

So, using AI to generate the contents of your RPs will result in disqualification for that cycle and potentially a ban from the site.

What determines whether someone is using AI to generate the contents of an RP?

So, we’re aware of certain writers using AI for planning or outlining purposes, at which point, they write the story entirely themselves.

But, certain users could also just copy-and-paste straight out of an AI textbox.

We’re currently tweaking and perfecting a model rule that doesn’t punish the first, while strongly disincentivizing the second. Our first pass at an implementation policy is…

Copy-and-paste your story into ZeroGPT (https://www.zerogpt.com/). If it returns a score of above 10% AI GPT, please rewrite before submitting. Submitted work higher than 10% will result in a DQ.

Why ZeroGPT?

It’s free, it doesn’t require anyone to make an account and, in the GM Team’s experiments, it seems consistent at labelling human-written work as below 10% and AI-generated work as above 10%.

Why 10%? Isn’t that too low?

We’ve tested a fair number of role plays, both from the XWF and other sites, and stories generated using ChatGPT, in our efforts to craft a functional policy. In our experience, most human-written stories, fall between 0 and 2%, rarely as high as 4 or 5%. Therefore, a story is unlikely to simultaneously (1) be human-written and (2) report a 10% AI GPT score or greater on ZeroGPT.

Why 10%? Isn’t that too high?

Among our concerns in implementing an AI policy is concerns over false positives. In our testing, we have seen a number of RPs, written by humans that can score up to 5%. Additionally, there are claims that repetitive or excessively “prosey” writers will experience inflated numbers from AI checkers.

In our experience, it’s rare to see a score above 5% from a human-written story. In our limited tests, we’ve determined 10% is both high enough to avoid alerting on false positives, *and* low enough to hopefully weed out copy-pasted work from ChatGPT.

The policy should be THIS way instead!

At this time, and going as far as foreseeably possible into the future, I’m open to discussion. Feel free to DM me at Peter Principle with thoughts and concerns. I will say, this is VERSION ONE of the site’s AI policy. We anticipate a number of reviews, conducted both by players and by the GM team, in an effort to make this policy as clear and easy to comply with as possible. Anyone who simply writes their story themselves, start-to-finish, should feel that their process is completely unaffected by this new rule.

As an aside, I’d like to offer a personal apology to the players on the site that it took this long to implement a policy regarding Generative AI. We are making every effort going forward in handling these issues proactively.

Thank you for your understanding and patience as we confront this new dimension to the game we all love.

I am now prepared for you to barrage me with insults and complaints and will respond to each one in kind.

Sincerely,
Wes (aka Peter Principle)

I hate technology.

It's a shame that this is needed, but I'm all for it if it weeds out cheaters. I guess my biggest concern would be false positives but it seems like you've already taken that into account. As someone who's always maintained the strict ethos that efedding is and should be a writing competition first and foremost, I support it wholeheartedly.

I can't understand the want to do something like that. What do you get from it?

So out of curiosity I ran my XWF story and my FWP promo.

My XWF story was 3.5% but if I remove the Edgar Allen Poe quote and the location at the top, it's 0%.

My FWP promo, however, got nearly 10%. Now first of all, keep in mind that this is very short. Like 1k words, so anything hit will be a larger percentage. Second, I quoted a song multiple times. However, the curios thing is it marked two paragraphs as "possibly AI generated" which absolutely weren't. Anyway, pointing out that it is possible for it to be wrong, but it required a very short promo of 1k words combined with obvious quotes from song lyrics to get that high.

So all in all, I think it isn't an issue unless someone is blatently using AI to write a lot. Having said that, I'm assuming discretion would be used that if someone clearly quotes an author or song lyric and it's clearly obvious that they are doing that openly, that if it that part is marked as AI generated it would be disregarded.

I'd like to weigh in here briefly to say that these new rules aren't necessarily because anyone cheated.  We can't confidently say that when the possibility of false positives exist even simply by using AI to translate from one language to another.

That's kind of what took us so long to come to any sort of agreement on how to police AI.  AI usage is not a new discussion among the staff.  This was at least the third instance of us trying to come up with a plan.

Having said that, we did make a decision to retroactively vacate a recent win because while we couldn't definitively say yes this person cheated, we also couldn't say that they definitively did not.  It was a tough situation that was many hours of investigation, experiment, and conversation.

(02-12-2025, 06:59 AM)gorgo Said: So out of curiosity I ran my XWF story and my FWP promo.

My XWF story was 3.5% but if I remove the Edgar Allen Poe quote and the location at the top, it's 0%.

My FWP promo, however, got nearly 10%. Now first of all, keep in mind that this is very short. Like 1k words, so anything hit will be a larger percentage. Second, I quoted a song multiple times. However, the curios thing is it marked two paragraphs as "possibly AI generated" which absolutely weren't. Anyway, pointing out that it is possible for it to be wrong, but it required a very short promo of 1k words combined with obvious quotes from song lyrics to get that high.

So all in all, I think it isn't an issue unless someone is blatently using AI to write a lot. Having said that, I'm assuming discretion would be used that if someone clearly quotes an author or song lyric and it's clearly obvious that they are doing that openly, that if it that part is marked as AI generated it would be disregarded.

There is always going to be a margin of error with AI detectors, hence what Peter said in the original post. If someone is quoting song lyrics or a poem or quoting anything else well-known/famous, if it is clearly meant to be a quote, we would disregard it immediatly so no worries there.

With shorter RPs, 1k for example like we have on Anarchy, yes there's a chance the percentage would be higher. This rule is in its infancy right now so it may change if this issue keeps popping up but to be on the safe side I would suggest just trying to re-write any highlighted sentence to see if the percentage drops as Peter said. If you change the sentence, even Google it to see where the sentence is from and find nothing and the AI detector is still showing a high percentage just make a GM aware.

Most likely if its just a sentence or 2 we'll brush it off; several sentences, a paragraph or more would be the bigger issue.

(02-12-2025, 01:59 PM)Atticus Gold Said:

(02-12-2025, 06:59 AM)gorgo Said: So out of curiosity I ran my XWF story and my FWP promo.

My XWF story was 3.5% but if I remove the Edgar Allen Poe quote and the location at the top, it's 0%.

My FWP promo, however, got nearly 10%. Now first of all, keep in mind that this is very short. Like 1k words, so anything hit will be a larger percentage. Second, I quoted a song multiple times. However, the curios thing is it marked two paragraphs as "possibly AI generated" which absolutely weren't. Anyway, pointing out that it is possible for it to be wrong, but it required a very short promo of 1k words combined with obvious quotes from song lyrics to get that high.

So all in all, I think it isn't an issue unless someone is blatently using AI to write a lot. Having said that, I'm assuming discretion would be used that if someone clearly quotes an author or song lyric and it's clearly obvious that they are doing that openly, that if it that part is marked as AI generated it would be disregarded.

There is always going to be a margin of error with AI detectors, hence what Peter said in the original post. If someone is quoting song lyrics or a poem or quoting anything else well-known/famous, if it is clearly meant to be a quote, we would disregard it immediatly so no worries there.

With shorter RPs, 1k for example like we have on Anarchy, yes there's a chance the percentage would be higher. This rule is in its infancy right now so it may change if this issue keeps popping up but to be on the safe side I would suggest just trying to re-write any highlighted sentence to see if the percentage drops as Peter said. If you change the sentence, even Google it to see where the sentence is from and find nothing and the AI detector is still showing a high percentage just make a GM aware.

Most likely if its a sentence we'll brush it off, a paragraph or more would be the bigger issue.

Oh I think you're all good. I was just posting what I saw when I gave it a whirl. I'm not worried about it myself. Just giving information/feedback when I filtered my own stuff. Personally I only use "AI" in the same way I use google search, which is "show me a list of (country of origin) first names" or "where is a creepy place in (insert city)". I think it is a good rule and the way you are doing it sounds good.