#ai detector

[quinloki]

LMAO alright so fucking get this bullshit. I run a fanfic written by AI through a detector. 0% detected. I run one I spent like two weeks writing. 26-45% detected. Let me off this cocky ball fart globe.

I have to start this with a PSA

Do NOT feed ANYTHING into an "AI Detector"

Never. Ever. EVER. Don't do it.

AI Detectors at BEST are useless, at WORST you are feeding your own work into a AI Collection/Scraper.

A robot cannot detect a robot's work because neither robot is actually intelligent. Neither robot can make a determination. Neither robot can come to a conclusion based on context because there is not yet any artificial anything that can think.

YOU have to learn how to spot AI content that isn't labeled on your own. I'm sorry, cause it sucks, and you're probably going to second guess yourself a lot, and you're just going to have to do your best and keep it to yourself because even if you are 100,000% CERTAIN that you are correct, opening up some kind of AI witch-hunt is absolutely NOT useful.

/deep breath/

Okay, harried lecture over.

Don't use AI detectors, they're functionally worthless and you're running the risk of consensually handing over your work to AI. I promise you any AI Detector site has some ToS policy about how you're consenting to letting them use what you feed into it.

So do not feed anything into it.

Ever.

EVER.

I do not care if you think [God] brought the most divine and pure of AI Detector sites into creation via infallible will.

DO NOT USE THEM.

[tippenfunkaport]

If you're not great at spotting AI art, there are a few free sites to help you figure it out...

(The first link, AI or Not is the fastest / easiest imho but selection is rarely a bad thing.)

Really helpful if you're getting a fanart commission or book cover or something and someone is trying to trick you into paying for an AI creation. AI technology gets better all the time so you can't always rely on the usual "tells." But if it looks off and you want to double check, here are some ways to do it!

[shamebats]

Beware: "AI detectors" are NOT reliable

Thread

[smokedvanillagumbo]

PSA that ai-detection tools are not yet as accurate as I wish they were. Here's hoping they improve QUICKLY, 🙏 but in the meantime, please be aware of both false positives and false negatives!

This reaffirms the importance of sharing WIPS, even though they are ugly and I don't wanna 😭😭

[pastelporcini]

AI scanners are such a fucking joke and it’s not funny at all.

I was writing my final English essay, and I put a lot of effort into it. It took a long time, like 40 something minutes, but I finished it and turned it in hoping to get a good grade.

It was flagged as AI by 5 different scanners. FIVE. DIFFERENT. SCANNERS.

I worked so hard on that stupid essay and all I got in return was having to desperately plead to my teacher that I DID in fact write it. I’ve had this teacher for 3 years now (I do in-person online classes and he oversees online work) so thankfully he didn’t penalize me or have me redo it because he knows I wouldn’t use AI. I got lucky, but what if it happened with a teacher who didn’t know me well enough to trust me on this??

I’m so tired of AI scanners. They don’t do their job correctly. I’ve heard so many classmates brag about using AI and getting good grades for it, but the moment I put in actual effort I get flagged FIVE TIMES????

It’s just such a fucking joke. AI scanners aren’t accurate, they just flag anything that has a formal/monotonous tone and uses big words. Which isn’t even accurate to the AI I have seen before. Why am I getting in trouble for putting in effort but the lazy jerk next to me gets an A for using chatgpt?? It’s just not fair and I’m so so incredibly frustrated I can’t even begin to explain it.

I don’t even think there’s ever going to be a way AI scanners can accurately tell what is AI and what isn’t, because it only flags certain writing styles. Not just AI uses these writing styles!! Where do you think it learned that from!! You cannot possibly tell what is AI and what isn’t just based off of that. It’s unfair, inaccurate, and screws over anyone who actually tries on their work. It’s bullshit.

[valy-gc]

Don't trust the "AI detector" things. I just wanted to see how accurate it could be, as it's a great way to know if writers are real one or not. Asked random text to AI.

Out of 4, the thing detected 3 as "probably written by a human".

Then I put my own text to see… same, I put 4 differents one and 2 have been noted as "probably written by an AI"… should I precise all the text I put in this were written waaay before AI even exist XD (thoses I wrote when in high school and some recent but still before this crazy AI thing) The reason? (because yes, there is details about it) "this sentence look similare to another book, it's problably copyrithed work copied by the AI" ah, yes "Once upon a time" and "they lived happily ever after" are SOOOO rare in books!!

Also, every time a difficult word is used, or a sentence is a little weird (yeah I'm sorry, english not my first language, of course there is mistakes) it immediately detect it as AI. Like… humans can't use difficult words apparently XD

And I put a little part of the book I wrote all in ryhmes… and so I'm forced to put weird sentence to have the ryhmes right XD BTW, the random text I asked to the AI were sh*t. I think we don't need a detector to know when we see one.

Remember the AI detectors are actually AI tools too...

Oh also… I tried with the french version of my texts. The exact same texts I tried in english and the percentages was waaaay lower, it only got crazy about the ryhmes book again… but then, like… every english peoples are AI??

To the teachers who use this stupid thing, don't thrust it! Don't give detention just because that thing said the text of your student have been written by AI. You'll totally know it if it's the case!

OMG I'm thinking of all of the poor students who got detention unfairly because of this thing...

[theladyhibiscus]

Hey y'all.

Check out Sightengine when y'all get a chance.

[mietxt]

just putting this out here in case someone got their art flagged or you found a drawing that wasitai.com or some other AI image detector says is AI generated- In my experience these detectors work much better for real life images compared to digital art and such.

[nonenosome2]

Stop Trusting AI Detection Tools

This is just a PSA, but if you don't trust AI to give you answers because they can lie, don't trust them to tell you when something else is AI generated.

[almach748-evil]

PLEASE I WROTE THIS MYSELF. I’M JUST AUTISTIC. YOU GOTTA BELIEVE ME, MAN

[bluntsam]

My evidence that AI could never replace me.

Read ‘em and weep boys. My bullshit is one-hundo-percent human.

[afiananvandot]

what’s the most reliable AI detector tool there is? bonus points if it works on video

[nonokoko13]

Found this web called Illuminarty that can help you identify whether an image is AI generated or not and thought some might wanna use it. It shows you a percentage of probabilities that image is AI and it can also be used for AI generated text. Here's an example:

Here's a logo I made and the proof of it. If I drop this image in Illuminarty it gives me a low percentage.

It doesn't need to be 0% to mean it isn't AI generated. This tool is probably analysing patterns it finds in AI art so it may need more versions to be better, but if it's an AI generated image the number usually is 50% or higher.

[wemlygust]

Also also: "AI detectors" CANNOT detect AI. Not even a little bit. It's a lie on every level. Their accuracy is equivalent to that of asking a Magic 8 Ball.

PSA

Just a heads up, there are bots going around on AO3 accusing people of using AI. Considering the timing, this is likely AI bros' retaliation for AO3 users calling them out for scraping their work. Examples of what you might be sent:

Screenshots from here.

If you get a comment like this, just report for spam and delete.

[ask-kirby]

Don't trust AI detectors

It's just source listing. How?

What in the world...

One second later...

That hit hard