Speaker 1: I tested the five best services to bypass AI detection and here are the results. They may surprise you because the first thing I did is I took my abstract. This is one that I wrote for my PhD and it definitely had no AI because it was written in 2011. So this is the first thing I did is I took that and I put it in, turn it in, I put it in originality and I put it in GPT 0. So all of these results are essentially how likely it is that it's AI content. So the first thing we need to do is make sure that this is 0, which it is. So all of them come back and said that this PhD abstract was not AI generated, a great start. Then I took that abstract and I said hey rewrite it and this was in chat GPT. I said hey rewrite it for a PhD thesis and so what it did is then it took that thesis and here are the results. It said that turn it in said it was 40% AI, originality said it was 100% AI and GPT 0 said 94% AI. So this is great. Unfortunately turn it in wasn't very good at detecting AI, which is kind of like the main one that everyone uses in universities. Come on turn it in, you can do better than this. But originality in my experience and you'll see throughout this test is it was the best one that actually detected AI and you'll have a look to see now because now it is a game between the AI detection and the bypass removal tools. I tested five of them. The first one I tested was Phrasely. This is Phrasely AI and so I paid for all of these services, I paid for all of the detection so you don't have to. It was many hundreds of dollars to test this but essentially they all follow this same sort of thing, which is you put in your content here and then you get content back. To make sure that it wasn't just detecting me copying and pasting from a website, I put every single one of them into a Word document and saved it there and then used the text from there and pasted it into the service. So that made sure that it wasn't just detecting that it was coming from a website. But this is what happened. So you put it in here, you get this and I ended up with a load of different abstracts. They all tended to be about 500 words long because that's what my PhD thesis abstract was. So that was sort of like testing roughly the same amount of text. This is what I got back. This PhD thesis presents innovative work in the fabrication of organic photovoltaic devices. Good. Okay. It's fine. It seems fine. But a lot of these, and stay to the end because I'll show you why I probably wouldn't just sort of like copy and paste this into something that I was submitting to a university. Because there's some really weird choices that it makes. But I won't keep you any longer. You can see that it passed, turn it in. Originality thought it was a 3% chance that it was AI generated and GPT-0 was also zero. So Phrasely was able to get around this AI detection. The next one I tested was Stealthwriter. This is this one, Stealthwriter. And you can see they all share the same sort of like process where you put in text that you want to humanize. I had this turned up all the way and then I just copied and pasted that into Word. I took that Word content and I put it into the different services. And Stealthwriter came back with zero from turn it in, 39% originality score and GPT-0 was also zero. So originality was the only one that was able to detect that this was actually not written by a human to any significant extent. Okay, the next one is Undetectable. Undetectable is, in my experience in the past, been one of the best ones to use if you want to get around AI content detection. And so this is what it looks like. Once again, you put in your text here. I wanted more human and I clicked humanize. I'll pull up that text that was this one. And so this is what it came up with. This doctoral dissertation delves into the world of photovoltaic devices I probably wouldn't use into the world. So it makes, like I said, some really sort of strange choices but let's have a look to see what happened because this one, in my experience in the past, has been the best. So Undetectable, turn it in score is zero. So well done. It was able to get past turn it in but bad on you turn it in. It seems to be the worst performing one so far. Originality, it was able to get 2%. So Originality thought that this was actually human. There was only a 2% chance that this was AI generated which is actually the best one so far. It beats Phrasely by 1%. So Phrasely and Undetectable are the best at the moment. And then GPT-0 detected it. Not at all. So turn it in and GPT-0. All of these tools know that they'll be used in turn it in and almost certainly GPT-0. So those are the ones I think they're trying to game the most. Originality, however, is still doing the best at detecting AI content even when it's been through an AI bypass tool like this. Next one, Bypass AI. So Bypass AI looks like this. I actually accidentally bought a whole year's worth of subscription to this one. So that's where mainly the hundreds of dollars went. But Bypass AI, I hope that you can do better than the rest of them. So once again, you put your text in here. You get this back. This was abstract number five. Here it is here. This PhD thesis is on the frontier of organic photovoltaic devices. Once again, weird, weird choices. And down here, even though it's kind of kept the same meaning, to a normal PhD student or to a professor who actually knows what they're talking about, this is not good enough. However, if I was to copy and paste this straight in to the AI detection tools, this is what happened. So Bypass AI, I'm going to reveal it all at once. Boom. It passed. Turn it in score. Well done. Turn it in. Step up your game, mate. Originality, 100% chance that this was AI. So originality, it did not pass. It was just like we had just generated it from ChatGPT. So Bypass AI has not done a good job if it was to be detected by originality. And then the last one, GPT zero, a big fat zero. So that means that this was able to get around, turn it in, and GPT zero. But originality, it knows you are AI. And the last really popular tool that I tried was Hicks Bypass. Hicks Bypass is here. So you bypass AI detection with an undetectable AI tool. Once again, put my text in here. I chose the latest model because I wanted it to be up to date and most powerful. And this is what happened with Hicks Bypass. I'll take these and I'll open it up. Boom. 0% chance that it was AI according to turn it in. Originality, once again, the best at detecting AI generated even when it's been through these tools at 56%. And then GPT zero, a big fat zero. So it's only originality that thought that it had a really good chance of being AI generated. So those are the results, which means that turn it in and GPT zero are rubbish. Turn it in is actually the least capable of detecting AI. Even when I give it purely generated AI content, it said only 40% of it was actually generated by AI, whereas originality saw it right away. And GPT zero thought it was a 94% chance of it being AI. So then after that, it was easy to get around chat GPT zero, it was easy to get around turn it in. But originality had its eyes open. It was only Phrasely and Undetectable that was able to get around originality detection tools. So that means that if you want to make sure that your writing will actually get through any checks, you could put it through Undetectable, you could put it through Phrasely. But here's the thing, I don't think you should be using this as they want, which is just copy and paste your stuff in and then put it into submission because check this stuff out. And I would not want to put my name to the stuff that it puts out, fade transition. So let's look at the output from Undetectable and see what it actually gave us. So you know, it says here, we talked about this, the world of photovoltaic devices, I wouldn't use that, it's not very natural. And then look right down the bottom, in conclusion, this research trailblazes by introducing semiconducting nanoparticles in LPV device fabrication, opening avenues for creating efficient and cost effective. So if I was to give this to a supervisor and another academic, they'd be like, this is wordy, the word choice is really weird. But you see, all of this, what it does is if you are using AI content, you should never ever, obviously, just pass it off as your own work. Because you know this content will not be detected by AI detectors, you can then build upon it with your own understanding and your own sort of paraphrasing and your own information. Make sure that this isn't something that you submit, but it's something that you use to scaffold your own understanding and your own work. That is how I would use this. If you like this video, go check out this one where I talk about the 12 best AI tools for academics and researchers that you should know about.