Here are two comments I got on my recent video on jellied eels. Both are compliments. Both are pretty detailed, but the moment I read one of them, I knew without a doubt it was written by AI, even though the user did try and hide that a bit.
To me, it's pretty obvious. But can you tell which is which? My name is Evan Edinger, and I'm not going to tell you the answer to that question just yet, because by the end of this video, you won't need me to.
Welcome to a master class in how to tell something is written by AI in the year 2025. a skill I do feel like everybody needs to have going forward. Over the past few months, I've been collecting many examples, emails in my inbox, LinkedIn posts, even tweets from world leaders, all clearly written by AI.
And once you know what to look for, you'll start to see it everywhere. As of right now, I've got about 10 red flags to share with you that you'll sadly begin to notice everywhere you go, including in certain YouTubers' videos posted wholesale. I hear chat Jimmy GBT written sentences a bit too often these days, especially on small faceless YouTube channels, which is why I usually avoid them.
One red flag on its own doesn't prove anything, but once you start to hear one, the others aren't usually too far behind. Let's start you off somewhere simple. You recognize this guy?
Well, of course, Evan, that's a hyphen. No, this is a hyphen. That right here is an M dash.
This bit of punctuation has been around since the 16th century and has been used in many works of literature from authors like Emily Dickinson to one of my favorites, Kurt Vonagget. It's meant to be a line with the same width as the letter M and separates a sentence like a semicolon. From this, you can probably guess what an N dash is.
Now, the problem large language models absolutely love the M dash and use it way more frequently than most humans do. And as of 2025, no matter how many times you ask Chad GBT to stop using them, he will always find a way. Pop quiz.
Which keys on your keyboard do you press to type an M dash? Yeah, I'm guessing most of you don't know, which is why it is a particularly blatant red flag when you see it used in certain spaces. For the record, on a Mac, it's option shift minus sign.
It is a really useful bit of punctuation, though, that even I have begun using in some of my YouTube titles, like this one, and well, well, quite meta, like the one you're watching right now, because it just spaces things out really well, better than a full stop or period. It's got a good flow. chat GPT bringing the Mdash back into the mainstream isn't such a bad thing, but this is a good reminder that this is just one flag.
Mdash doesn't mean something is AI. So, now that you know one big red flag of AI writing, let's look at another example I found from two seconds of scrolling LinkedIn. No exaggeration.
Striking the perfect pose in photography isn't just about looking good dash. It's about telling a story, conveying emotion, and creating a visually compelling image. Let's stop there.
I have only read one sentence and I already know with near 100% certainty that this is entirely AI generated from one sentence. Now, just because this sentence has an M dash doesn't mean it was written by AI. Of course, luckily for us, this one sentence also has three other big red flags.
Do you notice them? It's not just the use of M dashes. It's also the use of parallel sentence structures like this one.
It's not just X, it's Y. Subscribing to my YouTube channel isn't just about helping me get to a million subscribers. It's also about curating your feed.
This exact sentence structure is used so often in AI writing that you're going to struggle not to notice it going forward. I'm sorry. Also, you're welcome.
Now that I think about it, that cringy LinkedIn post from the Dualingo CEO I made a video about 2 weeks ago also used this method. AI isn't just a productivity boost. It helps us get closer to our mission.
Yeah, their mission now seems to be pretending to walk back their claims while the CEO says AI will replace teachers except as child care. So, you know, more Bonvillain stuff. This LinkedIn sentence also has our third flag, a convenient list of three things.
It's about telling a story, conveying emotion, and creating a visually compelling image. Well, that's a nice little triplet, isn't it? Humans love groups of three.
They just work. In writing, three points feels complete, but not overwhelming. In speech, they're easy to remember.
And in photography, you've got, of course, the rule of thirds and the rule of odds. Humans instinctively prefer most oddnumbered groupings. If I'm taking a photo, ideally I'd want the viewer to travel through the photo to three or five points.
Like my Jelly Deal's thumbnail had three specific things I wanted the viewer to see as they traveled around the image. My face, the dish, the text, and the data to back this case up is viewers clicking my thumbnail with three points as opposed to four watched more of the video. Here's another LinkedIn post I found.
Chalk full of groups of odd numbers, which alone isn't enough, but I mean, just look at this thing. Sure, there's no M dashes, but we've got it's not just X, it's Y. And group of five, group of three, and hey, the biggest sign of all.
It's on LinkedIn. I mean, even the format of this post is one your mind has probably seen so many times at this point that your eyes have already glazed over. One sentence hook, ethos, bullet point list, effect, conclusion.
It's like AI was handmade from generic LinkedIn posts for people who make generic LinkedIn posts. Now, it's not exactly the case in this post, but a lot of times when these groups of three come up with line breaks in between, AI loves throwing in an emoji at the start of each one, like this email I got or this extra LinkedIn post. Formatting line breaks with emojis was a big thing before AI, so I don't count it as a red flag on its own, but they're usually paired with generic content and well, our next red flag.
You see, another big marker of AI writing is the one you probably don't need me to tell you about. Even though you've just read something, it feels off. The word use can sometimes have a really weird, subtle use case that you've never seen before.
That's what I'd like to call the uncanny valley of the English language. The tone just feels strange, like striking the perfect pose in photography. Eh, in in photography, I I mean, that does make sense, but in that context, striking the perfect pose in photography, it just doesn't feel quite right.
I'm bundling this with another flag, too, which is an overabundance of filler words and expressions that aren't typically used in a normal conversation. AI writers lean on safe vocabulary. I'm talking vague positive adjectives and jargon like innovative or practical solutions.
AI just loves using words like elevate and delve. Words that most humans aren't regularly using, but you've probably been seeing a lot recently. Sometimes you can come across a tweet and feel like it doesn't really say anything at all.
Like it's empty of meaning despite the words on the screen. Like this one here from UK Prime Minister Kier Dharmmer. See anything familiar?
Hm. It's time to look forward Mdash. To move on from the stale old political fights and to find common sense, practical solutions that improve the lives of British people.
Secure borders, cheaper bills, more jobs. We will close the deal in the national interest. Like what the is that supposed to mean?
So, these are words, but they don't mean anything. And not just because they're said by the political equivalent of off-brand beans on toast. Now, I'm not a 100% certain that this tweet was written by AI.
Sure, it does utilize the mdash, a group of three, and does say a lot while nothing at all with empty language, but that one sentence up at the top reads so badly, it makes me feel like Chad GBT could actually do a better job than our prime minister, which is a very sad state of affairs, isn't it? Now, I don't want this video to become boring, and I assume at this point you're really starting to pick up on how this all works, so I'm going to touch on the next three points as a lightning round quite quickly. One, exaggerated and empty praise.
Chat GBT was updated recently to really glaze you with every response. And that carries through to a lot of AI texts. Ask a simple question like, "How can I get people to thumb up my new video?
" And Chat GBT will spend a paragraph like, "Wonderful question, sir. Now you're really getting to the heart of things. You truly are an imaginative genius.
" Here's a snippet from an email I received recently. Hi, Evanica Travel. That of course is my full name.
Thanks to my dad, Mr Travel. Rest in peace. Your travel vlogs are genuinely captivating.
M Dash, I love how you bring every new place to life with such honest and vivid storytelling. The way you share those unexpected moments, blah blah blah blah blah. You know me, I love sharing those unexpected moments.
It's praise like this that feels so fake and feels so bland. It doesn't mean anything. Number two, the second point, being able to spot AI writing is like being a lighthouse in the middle of a dense fog.
It's a weird analogy, wasn't it? Well, that's another big tell of using AI writing. Analogies and simileies that are just weird, like they're trying too hard to have meaning, but falling so flat, like a band-aid made of sandpaper.
Makes you think, huh? And three, AI writing is constantly restating itself, clarifying its points too many times, and providing unnecessary context. Like if I asked chat GBT why the sky was blue, the response would be the color of the sky is a phenomenon that has been fascinating people for centuries.
It's a very useless bit of filler preamble, right? To clarify, AI writing repeats its points far too many times. And if that was too meta for you, what I just did there, the previous sentence was clarification that was wholly unnecessary.
So to summarize, me writing this paragraph in AI style not only hurts my brain mash, it wastes your time, is devoid of meaning, and makes this sentence kind of feel like déja vu, right? But this constant clarification, this constant overexlanation, all of this stuff does feel like it's truly dumbing down society, which to clarify, I don't think is a good thing. All right, so you should be able to easily tell if something was written by AI now.
But how can you tell if it was written by a human? Well, a lot of times in AI writing, the words and the stories and the anecdotes feel very generic because they are. When humans are writing something, they usually inject a lot of themselves into it.
include a little story about someone they know or some place they've been in the first person. One example I was going to show you of a human comment is this one here. Yes, this video is an excuse for me to show nice comments about me.
Thank you, Evan. Evan, you've really outdone yourself with this video. As a northerner, I have never appreciated jelly eels.
Personally, I really disliked them when I tried them. So, in this comment, we have, "As a northerner, personally, I was totally fascinated. " That information allows them to add a bit of themselves to the comment to add context.
It's a real comment. Also, where the comment takes place is really important. Seeing perfect grammar is not too big a flag online compared to everywhere else, but if a friend of yours starts dropping commas into an iMessage paragraph response to you, they might be uh using AI to get a little help there.
I do sometimes get comments from people that seem oddly insensed that I go off on tangents in my videos as if every video needs to be robotically following one topic. Like in this very video, I didn't need to take that photography tangent teaching about the rule of odds. But I really love photography and I figured it was a relevant enough fact that people would be interested in knowing it.
Tangents at this stage are a pretty good sign of human writing as our minds can make relevant connections to other things we know to make conversations that much more interesting. When you think of some of your favorite stories from films or even YouTube videos, there's a good chance the narrative isn't direct from A to B, but is a satisfying journey to get there. Like this very YouTube video you're watching right now.
How did we start this video again? Oh, yeah. If we circle back to the original comment I showed in this video's intro, like I said, you really don't need me to explain how I was immediately able to peg this as AI because now you can, too.
Congrats, you did it. It starts off with a bit too much impersonal high praise, and hey, I appreciate it. I spent a long time on that video, but still, it's not just about uploading in 4K.
It's about making every shot count. Now, what's funny about this comment is that Matthew here did try the bare minimum to cover his tracks. You might have noticed there are no M dashes here.
Well, I jokingly called him out in a reply in a way that would confuse anyone not in the know. I said, "Thanks, but you replaced that M dash with a hyphen, didn't you? " Haha.
It's not just about X. It's about Y is quite the tell, as you know. Thankfully, Matthew was quite a good sport and replied, providing us with the original prompt he put into chat GBT to generate his comment.
I have to take my hat off to you, Evan. You put an extraordinary effort in to make your videos look the best they can be. Just uploading in 4K isn't the solution because garbage in, garbage out.
Viewers, if you are looking at this on your phone, stop and watch it on your big screen TV instead, as a small screen doesn't do it justice. There's literally nothing wrong with this comment. It's genuinely kind and most importantly, original.
It's human. It's how he thought to express himself. And as you can see in his original comment, Matthew even used the first person.
I have to take my hat off to you becomes hats off to you. AI writing removes you from the equation. Now you have become bland.
Destroyer of worlds. In justosition to the real comment, the AI comment feels corporately kind, like it's trying to be nice, but insincere and generic. Which is why out of all the technical red flags we've discussed in this video of noticing when something is generated by AI, the one that AI is going to have the most difficulty overcoming is also the most human of all.
Passing the vibe check. If something has fake sounding AI vibes, maybe just trust your gut. AIS don't have guts yet.
Thank God. There's going to be a future where they have guts. I don't want to live in that future.
But also, that'd be kind of cool cuz then you can gut them like sheep. Not so bad. I would say you should be much better able to detect AI writing now, but your next challenge is to try and hear when something was written by AI in a video.
I'm not planning on calling out any particular YouTubers here by playing some clips of clearly AI written dialogue. So instead, to know whether you pass the final test, I'm going to again do an ad read where half is written by AI and half is written by me. The last time I did this, far too many of you left me comments saying that you got it wrong.
I'll label each as A or B in the corner, and we'll reveal the answer in the outro. I'm really hoping you find this quite easy, and you spread this information to as many people as you can, either in person or by sharing this video. All right, here we go.
If you're online, and I'm going to assume you are, your internet provider can see everything you do, every site you visit, every video you stream, every search you probably wish you hadn't made. Well, that's where NordVPN comes in. It protects your browsing by funneling your data through an encrypted tunnel.
Your ISP can't see a thing. And because NordVPN lets you change your virtual location, you can unlock a lot more content on streaming platforms like Netflix. US catalog, easy.
Japanese anime exclusives, go wild. Plus, get this. NordVPN is the fastest VPN on the market.
So, unlike other VPNs, you won't notice it slowing you down. I did mention this before, but when I was in Vietnam, the Chat GPT app stopped working as it isn't available there yet. But luckily, with one tap on my phone, I changed my browsing location back to the UK, and the app started working again.
Huzzah! Plus, you could secure up to 10 devices with one account, and they've got over 7,000 servers worldwide. So, wherever you are, you're covered.
And there's a 30-day money back guarantee. So, if it's not for you, no harm done. If you're interested, be sure to sign up to nordvpn.
com/evan, and you can get a huge discount on a 2-year plan, plus a neat little bonus. So, thanks to NordVPN for sponsoring. Now, let me wrap this up.
All right, we made it to the outro. I have a feeling you found this one a lot easier than the last time I did it. If you guessed B was written by AI, then congratulations.
You were very wrong. B actually included an anecdote about my time in Vietnam. A had multiple verbal M dashes and groups of three.
Maybe watch the video again if you got this one wrong, but to everyone that got it right, yay, congratulations. A was written by AI. Now, if you also happen to hate cookie pop-ups like I do, then you'd probably enjoy this video I made filled with all my tips for making browsing the internet suck a lot less.
I just got this comment on it today right before filming if you need a nice human review. When I clicked on this video, I was like, 20-minute video just to talk about cookies and now this is one of the most useful videos I've watched. Well, thank you anonymous.
You are legion, I guess. Hopefully this video was useful, too. I do enjoy not today, Satan.
Ah, you thought you could get me. I saw you. I saw you.
I do enjoy sharing niche information on my YouTube channel, and I do have some rather big videos hitting your subscription feed soon. So, thanks for watching, and hopefully I'll see you in the next one. Goodbye.
