I'm sorry for yet another chapter that isn't results, though this is extremely important and impacts you.

If you are not aware, one user in the K-pop community was recently exposed for using AI to write their stories, and this user submitted their work in various contests, mine included.

I will not name names as that is not the point of this chapter. The point is, this situation got a crap ton of attention because this happened in a popular contest with a popular account in the K-pop community, so people took notice. Left and right, contests have been posting new guidelines and cracking down on AI. I've seen literally 5+ contests making announcements and changing their guidelines.

So, where does that leave this contest, and why did I ask award hosts to read this chapter?

Let's answer that.

I work in AI. I have worked in it for over a year, so I have a slight advantage over most contest hosts in that regard. It's in my rules that AI content is not allowed, and if I discover you using AI, I will ask you about it. I personally have been on the receiving end of false AI accusations. Not on Wattpad—in my career. I will not automatically disqualify you and will instead ask you to provide proof that your work was written by you. If you fail to provide proof, you will be disqualified and blacklisted. If you admit you used AI, I will disqualify you but will not publicly blacklist you, but you will not be allowed back in any of my contests until you can provide solid proof your work isn't AI-written.

In this day and age, please always have a way to prove your work was written by you. For example, write on Google Docs, Quotev, Word, etc.—any site that has history so you can see all the changes down to deleting a comma. This isn't for contests anymore, this is for your own safety. If I didn't have proof I didn't use AI to write for the client that accused me of AI, there is a strong chance my career would have sank. I was just getting into the professional writing world at that time, so an AI accusation would have caused damage strong enough that I would have to sue the client for libel. That's how serious it was—I almost went to court over it.

Yes, it sucks we need to have proof, but it never hurts to have, and you might thank yourself for having it.

So, to summarize the answer to the first question:

- This contest already has anti-AI rules, but to clarify them, you will not automatically be disqualified if I suspect you of AI. You will be given a chance to provide proof your work is not AI-written.

- If you submitted AI-written work, please let me know now. You will not be publicly blacklisted as I would appreciate your honesty. Don't get me wrong, I would still be upset because you submitted work not written by you, thus taking chances away from other people; however, I would appreciate you for taking accountability. Please do not submit AI-written work. It's unfair, and it takes chances away from people genuinely working hard on their writing.

If you do not want to say this publicly, feel free to email me or reach out to me on Twitter. My Twitter is Jiminbiasnim, and my email is [email protected].

Now, the second question: why did I ask hosts to read this?

Hosts, you are making a grave mistake. Many of you are making a "percentage threshold" for AI-generated work, like if your work is more than 60% "AI-generated," you're booted from the contest. As an experienced AI worker, please listen to me.

That is a terrible idea.

The majority of people I talk to about this don't know what AI detectors actually do and think the percentages mean "Oh, 50% AI-generated? That means AI wrote at least 50% of this." That is not what it means.

AI detectors are only right about 60% of the time. Several studies have shown AI detectors being "right" as low as 49% of the time, but most show in the 56-70 range. This is because AI detectors lie to you about the "97% accuracy!" tag they put on their websites. How is this legally possible?

Loopholes, baby!

AI detectors do not (in many cases, though every detector is different) tell you how much of a work is AI-written. It gives you the chance. If you put your work into a detector and it comes back "40% AI-written," it does not mean 40% of your work is AI-generated, it means there is a 40% chance AI had influence on the work.

It's like plagiarism detectors that flag things like "He walked to the store" as plagiarism, or the titles of movies and video games you're referencing in your work getting flagged. That's not plagiarism, the detector is just wrong. The same applies to AI flagging random lines as AI-generated, but it'd be really weird if some random lines were AI-generated and not the whole thing.

What's the loophole? Glad you asked. If an AI detector says a human-written work is 99% AI-generated, then the detector is still technically right because of that 1%. That's why you'll rarely see AI detectors give such extremes, like 0 and 100: those are the only two numbers that make it so a detector is "wrong." It's a loophole that allows these companies to boast "97% accuracy!" when it's more like 60%.

AI is too new for detectors to confidently detect it. If you were to ask me, it's probably going to take at least a decade for detectors to be more than 80% "right." The tech is too new, making this situation a slippery slope.

So if you are an awards host and you're using this percent threshold, that means you're more likely than not going to disqualify an innocent person. Let's say your percent threshold is 50%. You would be kicking them out over a 50/50 chance from an unreliable detector.

That raises the question: how do you find AI-written work?

For one, never only use one detector. Never. I was falsely accused of using AI as I said before, and one of them even said there was a 100% chance my work was AI-written; however, if you took my text and ran it through literally any other AI detector, you would see my work came back human-written with low percentages of AI.

If you are an awards host, at minimum, run the text through 5 detectors. At minimum. If the majority of them come back with high AI results, then that's much stronger proof that the text is actually AI-written.

I'm not trying to say your suspicions are misplaced or the detectors are always wrong, but if you are going to throw someone out of a contest and seriously damage their reputation, you need to be sure. My reputation was almost thrown down the toilet over false claims because of a client who was too lazy to take these steps. Please, make sure you're using more than one detector. Please. I'm passionate about this because I have personal experience, so I hope I'm not coming off rude, it's just a really personal matter for me due to how serious of a situation.

Just to expand a bit more, I provided solid proof I didn't use AI and the client ghosted me, didn't pay me nearly $100, then left mediocre feedback on my profile (without giving a review, just a star rating).

Her company went out of business because of the false accusations, and Upwork support removed the review because it was left in bad faith. Karma.

(not by Jojo Siwa)

Also, there are other ways to find AI-written content without a detector. AI almost exclusively writes in purple prose, which means over-the-top language with more complex vocabulary than normal vocabulary. There are plenty of purple prose writers, so this isn't a 100% sign of AI, but it is a first initial tell.

Let me give a list of things that scream AI content:

- Descriptions of fingers, particularly the phrases "skillful" or "dancing" fingers; this is one of the most common signs of AI as AI adores using finger descriptions like those two

- Idealized characters - AI loves making idealized characters, like a princess living a glorious life

- Overuse of descriptions and long paragraphs

- "Core" human themes, AKA common themes that are extremely human (the AI is overcompensating for not actually being human). An example of this would be isolation.

So, long story short: hosts, please don't make it a percentage threshold. Go based on those traits and use multiple AI detectors.

Update on the copying situation: it has been handled. I will not go into details for privacy purposes, but the host has removed the copied content, therefore I will not be taking any actions against the individual, and I hope they have learned from this experience and will not copy again. Thank you to everyone who came forward after I posted my announcement revealing that you too thought that awards was copying me. I have now received over a dozen messages from people identifying those awards as the ones who copied me. I appreciated hearing your perspectives.

My birthday is tomorrow. I want to relax, have a good day, and not worry about Wattpad drama. I'm hoping this is the last time I have to update before I announce judging and results, but if another serious situation arises, I will let you know.

Thank you for reading.