Introducing: The Interface - What goes on in TikTok's Farlands?

10 Jun 2026 · 42 min · 16 chapters

Ask about this episode

Ask anything about it. ChatGPT or Claude reads this page and answers with the times it was said.

Connect VO and ask about every podcast you hear, including the moments you saved. Add to ChatGPT · Add to Claude

In short

The Interface episode “Introducing: The Interface - What goes on in TikTok's Farlands?” explores TikTok’s “Farlands,” a hidden edge of the algorithm where users can reach niche, uncanny horror/“brain rot” content.

Guest

Aiden Walker, a freelance reporter and self-described meme researcher (interviewed from Barcelona festival context).

Key claims

Farlands is accessed by “secret codes” (random letters/numbers) entered into TikTok search, acting like a “digital speakeasy,” though some users also reach it via the FYP. Content categories include niche brain-rot edits and creepypasta-style horror; examples include Smoorlater’s glitching car video with eerie music and brief evil clown imagery, and “iShow Eyes,” an AI-edited iShowSpeed with distorted neon/uncanny eyes. Walker argues it reflects human community-building and gatekeeping, while TikTok still benefits from longer watch time. The episode then shifts to AI detection: listener Grace describes students using Pangram to “humanize” essays, which backfires by evading detection while degrading writing; discussion cites Pangram false positives/negatives, “Shy Girl” accusations, and the arms race with AI “humanizers.”

Written by AI. May contain mistakes. Listen to the episode to check what was said.

Chapters

Tap a time to open that second in VO

Exploring TikTok's Farlands

1:58 to 6:15

Discussion about the TikTok Farlands and its implications on culture.

“Today on The Interface, we'll be diving into the horrifying world of the TikTok farlands.”

The Nature of Farlands Content

6:15 to 10:20

Examining the types of videos and community dynamics within the Farlands.

“And there's some like deep, crazy stuff out there.”

The Aesthetic of Farlands

10:20 to 14:01

Analyzing the visual style and thematic elements prominent in Farlands content.

“because people seem to have that year in it.”

Exploring the Aesthetic of Far Lands

14:01 to 18:40

Discover the unique and unsettling aesthetic of TikTok's Far Lands and its implications.

“They keep cutting back and forth with clips of this Twitch streamer, this guy named iShowSpeed, and there's this version of like an edited image of iShow Speed that's become like a staple of the Far Lands.”

Exploring the Aesthetic of Far Lands

18:41 to 19:29

Discover the unique and unsettling aesthetic of TikTok's Far Lands and its implications.

“You just need to start selling with The RealReal.”

The Evolution of Internet Culture

19:39 to 22:48

Examine how internet phenomena like creepypasta evolve and impact society.

“What would you lock away to explain your country to the future?”

Far Lands and Algorithmic Rebellion

22:49 to 25:36

Analyze the implications of Far Lands going mainstream and its impact on creators.

“But it was part of a new phenomenon from the internet of this kind of evolutionary collaborative art.”

AI Detection and Writing Anxiety

25:37 to 28:00

Discuss the effects of AI detection tools on student writing and creativity.

“Maybe you should make a TikTok account, Karen.”

AI Detection and Its Implications

28:00 to 29:10

Discussion on the rise of AI detection software and its impact on written works.

“So this gets to like a really interesting question, which has been coming up a lot recently because there's this new AI detection software called Pangram that has become really, really popular.”

The Accuracy of AI Detection Tools

29:10 to 30:50

Exploration of the accuracy rates of AI detection software and its consequences.

“So this is a horror novel that was supposed to come out and got pulled.”
Show all 16 chapters

Real-Life Impacts of AI Accusations

30:50 to 32:30

Personal stories illustrating the consequences students face when accused of using AI.

“Pangram, which there was one independent study, not yet peer-reviewed, but that did find that Pangram generally has a pretty high accuracy rate.”

The Nature of AI Humanizers

32:30 to 34:35

Discussion on tools designed to humanize AI-generated text and their effectiveness.

“like me, where they had a Google Doc where you could see in real time by looking through the history of the doc.”

The Influence of AI on Human Writing

34:35 to 36:58

Examination of how AI-generated text affects human writing styles and standards.

“But at the end of the day, when you think about it, first of all, these are probabilistic tools.”

Reevaluating the Purpose of AI Detection

36:58 to 39:48

Questioning the necessity and implications of AI detection tools in writing.

“Yeah, and that must be even, you know, we are in a position in our careers where we can, you know, demonstrate when it comes to it, we can like fight back.”

Future of Writing and Assessment

39:48 to 42:02

Discussion on evolving assessment methods in education amidst AI challenges.

“You know, we're being influenced both by some people using AI tools to generate the writing.”

Rethinking Education in the Age of AI

42:02 to 44:34

Explore how AI reveals structural issues in education and journalism.

“Obviously, that's, you know, more difficult in education systems from lower economic backgrounds where you tend to have bigger classes and it's much harder to teach in that kind of way.”
Hear the part that matters, and keep it.Open this episode in VO. Double tap your headphones to save a moment as you listen.
Get VO free

Transcript

Automatic transcript. May contain errors.

0:00This BBC podcast is supported by ads outside the UK.

0:30Make your travel dollars work harder. Follow Smart Travel on your favorite podcast app. Transform your home in one weekend and save big money at Menards. Protect and enjoy your outdoor living space with Thompson's Semi-Transparent Wood Sealer. It maintains wood's natural color and provides durable, year-round protection from water and sun damage in just one coat. Save big money on Thompson's Water Seal products. Our knowledgeable team is ready to help bring your dreams to life one gallon at a time. Stop by your local Menards today.

1:07Hello and welcome to the documentary from the BBC World Service. I'm Nicky Wolfe, one of the co-hosts of The Interface, the BBC's fiercely informed, fast and funny take on how technology is changing everything. Each week, we unpack the tech news stories that matter, whether they shook a government, broke the internet or quietly tipped the balance of power. Find us wherever you find your BBC podcasts.

1:58I'm Karen Howe. And I'm Nikki Wolf. Today on The Interface, we'll be diving into the horrifying world of the TikTok farlands. We'll be looking at what you can find there, but also what it means for internet culture. And we'll also be looking at how AI detection tools are changing our writing.

2:21Tom, nice background. Where are you? I'm in Barcelona. I'm going to a music festival called Primavera. Nice. And then Karen, you're sick. Again. We're still doing the pod though. That's how loyal we are to our audience. That is dedication. Tech waits for no one. And especially this week because we've got, I think it's a particularly crazy story. So over the past couple of weeks, I have gotten completely obsessed with something called the TikTok Farlands. And I found out about this, like I reached out to Nikki, I had to get him involved right off the bat. This is, I think, simultaneously one of the most unsettling but also like hopeful and optimistic things that I've seen on the internet in a really long time.

3:13Yeah, it's the whole package. It's like kind of, you know, a perfect summation of what the internet is and what it's like. So here's the idea. There's a hidden part of the TikTok algorithm that, It won't show you normally these videos that aren't coming up in people's feeds organically. And what people are saying is, if you want to see this stuff, you have to find your way to it with more intentionality. You have to seek it out. And apparently, the best way to get there is people will comment like a string of random letters and numbers, like this secret code word. and you put that into the search bar and it brings up this hidden genre of videos that I think is simultaneously, like this stuff is kind of horrifying and scary and uncanny, but also the story and what's happening around it, I think is one of the most positive things I've seen on the internet in a while.

4:16And it's people kind of taking control of social media to use it for their own community building and self-expression. Yeah, in a way that goes beyond what the app is designed for. And I found out about this from a freelance reporter and self-described meme researcher. His name is Aiden Walker. And I got him on the phone to talk about what's going on here and what this says about the state of the internet. So my first question for him is simple. what is the TikTok Far Lands? You could say trend, you could say phenomenon, but it's this space at the edge of TikTok. The word comes from this space at the edge of Minecraft where you'd go out far enough into the world that the computer no longer knew how to render the land correctly.

5:03And it's kind of the same concept with like the TikTok algorithm that if you spoof it, if you trick it, if you scroll long enough, it's kind of like if you scroll for 14 hours, you're going to end up in this place where you are seeing things that you are maybe not supposed to see. So once you get there, you're going to find videos that the algorithm would probably never have recommended to you unless you're some kind of sicko or some kind of avant-garde freak. It's kind of like hipstery in a way that you've transcended what most normies see. And a lot of the content that is in the far lands in practice, it kind of breaks down into kind of two categories for me.

5:40It's like the very, very niche brain rot, which is like i'd say it's like conceptual art in a way it's like these video edits that don't make sense that maybe are funny maybe are not are just kind of like vivid and crazy and odd or it's kind of like creepypasta content uh which is this internet tradition of kind of like folkloric horror of kind of just jump scare content of monsters pretty much that you've reached the edge of the map and as my friend etymology nerd says here there be monsters You know, you're out past the known area and you're in the zone that only the initiated know. And there's some like deep, crazy stuff out there.

6:20Can you describe some of this stuff? Like you say, brain rot, like for people who aren't on TikTok, what are we looking at? And also like horror, like are there some examples you can think of? It's these participatory, collaborative horror stories that people wrote online 10, 20 years ago. there'd be kind of spooky scary images of different types of ghouls or like different just uncanny creatures one of my favorite farlands posts is it's like a bunch of like invisible diaphanous beings walking across the screen and there's like a spooky like owl sounding soundtrack on top of it sometimes it's like jump scare content and what's interesting to me like being terminally online is a lot of these are like little ghouls or monsters that were being talked about in like 2015, you know, on like text-based forums.

7:05And now they're in the TikTok farlands. And so it's a lot of that sort of off-putting shock horror type stuff that has always been there on the internet. And I think for people who grew up on it, it was like, here's this, you know, a little illicit thing that's going to spook my friends out, like the digital equivalent of the, you know, going in the mirror and saying Bloody Mary three times. So what's this whole thing where you need a secret code to get there? Yeah, so I was seeing that in the comments of a lot of the Far Lands videos that I was looking at. And people would just leave these codes of random letters and numbers, which to TikTok probably means nothing.

7:43I mean, nobody's like searching that. But you put it into the search bar of the platform, and you'll find a bunch of videos that are tagged with that code. So it's kind of like a little secret backdoor. Like I was calling it a digital speakeasy that you know this code, you put it into the totally normal seeming social media. media platform. And then you're going to access this weird layer of like the scary ghouls or the crazy brain rot that other people are getting. Not everybody is finding the far lands through the secret codes. A lot of people are getting it on the FYP algorithmically recommended.

8:15But the fact that like the codes, yeah, on the for you page, but the fact the codes exist, I think shows this desire people have to like gatekeep things on TikTok because like the point And gatekeeping then is to selectively open the gate for the worthy. And so people love to kind of broker the relationship with the platform on like a human to human basis. And so I think the far lands and the codes are kind of symptomatic of people wanting to use the platforms to do something differently. But then ultimately, TikTok is happy the longer people spend on TikTok. And TikTok doesn't care if it's a horrendous creepypasta character or a lie on the internet or an influencer's day in the life video.

8:54someone sitting down and watching that. Does it feel like people are pushing up against the walls of the machine here to make TikTok do their bidding in a way? I think that is what's happening to some extent because it's kind of an inevitable human thing. The second you tell a kid not to do something or you make a rule, people are going to try to find a way to game it or break it. Yeah. And I think it also shows the way that people want to feel like they own some of their experience on social media. I think there's like this stereotype of the passive scroll, which definitely is a big piece of a lot of people's, you know, consumptive life online.

9:32But folks aren't just, you know, going past every video, you know, soaking in the information they don't even remember an hour later what they saw. I think people want to like be in community online and they want to feel like they're building something with other people online. And so I think it's like people trying to reach out to each other across this medium that if you're using it totally the way it's supposed to be used, it's a slot machine built to sell team. It just wants you to stay there and then maybe click a shop link. But people find a way to use it to do all this other stuff, including scare each other with creepy pastas, but also create these little spaces outside of the platform.

10:05And I always wonder if those same kinds of tools or techniques or attitudes might someday be turned towards something more substantive than just like, let's scare each other or even do crazy cool video art. Let's maybe use it in some political way or some really way that would be subversive to the platform. because people seem to have that year in it. But I do think maybe the seed is there for something that's a little more full-flung because there's a lot of like Luddite stuff about like touch grass, like put down your phone, you know, disconnect to reconnect. And that current is like very strong, I think with all generations.

10:39But then I also think there's always been like, even in the terminally online spaces, this desire, like what if we used this to do something more interesting than just, you know, sit and passively consume. And my view is that like, any like effective resistance to maybe the evil project of tech if you see it that way is a combination of the two you know it's not just quitting it's also like reaching past it and making the internet into a thing that helps people rather than serves ads you know like it's i think i had a comparison where i said finding out that you know the ostensible purpose of the machine that connects everybody on earth's brain to everybody else's brain if the purpose of it is to sell ads, that's like finding out the purpose of the pyramids of Giza was to sell fishing equipment.

11:25Like there has to be like some higher purpose to the internet. And I think when people reach past what the platforms kind of serve you, they're kind of reaching for that higher purpose. I have so many questions after seeing this video, because the first thing that I want to know, I mean, Farlands is such a brilliant name. I want to know who even came up with that term. But also, what are the types of videos that are populating the Farlands and how do they get there? Like, are people producing content specifically to land in the Farlands? Or are they just producing bad content? And then it gets damned to the Farlands.

11:58But now there's an entire community regrouping and reclaiming the Farlands. Like, what's actually going on? So people are making videos specifically to go in this genre, to go in this place, the TikTok Farlands, because people are really excited about it. But then also there's this whole idea that there's like you have to go deeper and deeper, right? You can find the stuff on the surface, which is these popular TikTok Farlands accounts. But the real stuff is something you'd see from an account with only 10 followers, a video that no one else has seen before that is truly freaky and uncanny, which is not a totally new thing on TikTok.

12:38Like years ago, people would talk about deep talk, which was kind of just weird, maybe actually bad videos that the algorithm wasn't serving up to people. But why don't I show you a couple of these as an example so we can get a flavor from what we're talking about here. So this one here, it's by a guy. I talked to his name is Smoorlater. Smoorlater. There's probably some really obvious way to pronounce this I'm not seeing. But he's sitting in a car, but the video keeps glitching. There's this eerie music if you go watch it. It's cutting to kind of distorted, you know, images and videos where it feels like it's not rendering properly or it's all like...

13:19Sort of an evil clown face keeps like very briefly appearing. Yeah. Like a lot of mouths and teeth go into a lot of this content. He's kind of subverting the genre of food review videos, right? It's a big thing on TikTok that there will be a guy who goes to a restaurant. He's sitting in his car eating it, talking into his camera about the food. So it's like... It's making it grotesque. Yeah, and then a lot of it is around like images of faces that just don't look right. That I think, you know, I'm showing us, we're looking at some of the tamer videos here, but some of these are, I think, actually really disturbing.

13:55Quite stubbing, yeah. Here's another one. A lot of this is kind of AI-generated images or AI-edited images. They keep cutting back and forth with clips of this Twitch streamer, this guy named iShowSpeed, and there's this version of like an edited image of iShow Speed that's become like a staple of the Far Lands. They call it iShow Eyes where it's this guy, iShow Speed, and his eyes are all like weird and messed up in this kind of uncanny way. Yeah, it's very kind of liquid. The colors are very neon and then it cuts to these black four figures almost so quickly that you can't see them. It's very like, it's like the design of a lot of these is just like a little shiver down the spine.

14:41You know, it's that kind of, that's the aim of it. I feel like all of these are kind of a similar aesthetic. Like, is that, is this the Far Lands aesthetic? Pretty much, yeah. Yeah, so there's a very distinct Far Lands aesthetic for sure. And a lot of it is using, this guy Lucas Wilm that I talked to said, he's using a technique that's called data moshing, where you take a video or an image and then you go in essentially using a computer program or like editing like the code of the file itself in order to make the JPEG or the video look kind of distorted, which is interesting, right? It's this whole thing where people are, you know, messing with the TikTok algorithm to make it do what they want.

15:31And then the content itself is like, messed up corrupted digital files. So I don't even know. I talked to some of these guys who were doing it. They weren't like necessarily doing this consciously, but even the content itself is like a comment on technology, which maybe we're getting a little too like academic art criticism here. I mean, this is so weird because the videos that you're showing right now, they are extremely popular, like 13 ,500 likes on this one. Some of these have millions of views, yeah. Yeah, so what is, I mean, I thought Far Lands was defined by people, very few people being able to access it.

16:12But is it actually just defined by the aesthetic, the non-mainstream aesthetic? Is that because it's now no longer far? Like people have discovered it and it's kind of already mainstream? What exactly is going on there is hard to say. It could be that this thing has broken containment and it has become mainstream. And it has. I talked to one of these guys. He's like 19 years old, right? I feel like I'm, you know, a retiree talking to this kid. And I'm like, you know, I work for the BBC. Like, is the fact that I am about to talk about this in the mainstream media, am I ruining it now? Like, is this like they used to talk about the hug of death on Reddit that you'd find some little beautiful thing and you put it on the front page of Reddit and then it would crash the server because too many people would look at it.

16:58And he was like, dude, it's already like this trend has like come and gone. You heard about it, by the time you heard about it, it's already over. But it's almost like, you know, people in, you know, people putting their graffiti on the subway car in New York in the 80s, right? Where it's like the subway car is not a vehicle for art, but people are co-opting the system and making it their own to use it for their own purposes. So it kind of is just this very human thing that's how the internet always has been. Yeah, I think graffiti is a really good example because it's about putting something in a public space that only a private group of people appreciate.

17:39The aesthetic of these videos reminds me of this other thing that's been happening with generated art, where there's this kind of creepy, goalish woman that will sometimes appear when people have spent too much time. What they do is they'll generate the art and then they'll pop it back into an AI generator and be like, make it more this or make it less that or whatever. And they'll just do it in this crazy cycle all the way down the rabbit hole until suddenly this particular artifact appears, this freaky, messed up face of a woman. I mean, that could absolutely be a Far Lands. Yes. So this is Loa.

18:18And yeah, she has a name. And so there's this whole genre of mythology, I guess, that has spun out of this artifact. But the moment I saw the Farland Studios, which I've literally never seen before, I was like, oh, it reminds me exactly of Loeb. You're listening to the documentary from the BBC World Service.

18:41So you're running out of closet space. The good news? You don't need to stop shopping. You just need to start selling with The RealReal. The RealReal is the world's largest and most trusted resource for authenticated luxury resale. Whether it's that mini bag that can't even fit your phone, or those boots you never fully broke in, The RealReal handles everything, from photography and copywriting, to shipping and pricing. So you can just sit back, get paid, and make room for things that actually feel like you. And with 10 ,000-plus new arrivals every single day from top designers like Prada, Celine, Louis Vuitton, and Loewe, all for up to 90 % off retail, you're bound to find something perfectly on-brand to fill that extra closet space with.

19:22Plus, right now, you can get an extra$100 to shop when you sell for the first time. Make room for what feels like you. Go to therealreal.com to start selling and get your extra$100 to keep shopping at therealreal.com. That's therealreal.com. Terms apply. What would you lock away to explain your country to the future? In 1876, as America turned 100, a group of citizens sealed their answers inside a safe meant to be opened a century later. But when the Sentry Safe was finally unlocked, what the world saw didn't tell an obvious story. It raised a question. What do the objects we choose to remember reveal about who we really are?

20:03This is the Sentry Safe on a history of the United States and 100 objects. Find it in the 99 % Invisible Feed wherever you get your podcasts. The thing that excites me as much as the phenomenon of the content itself is this has all happened before. This is an evolution of a genre that's as old as the internet itself. And this concept of creepypasta goes back to places like 4chan and something awful. These are like old internet forums that had a certain series of criteria that made them really fertile breeding ground for trolling and a lot of kind of horrible stuff, but also this really new and collaborative, creative process that was, I found, fantastic.

20:57And the aesthetic is exactly the same. Like it's very, very, whether it's intentional or not, kind of pays homage to... The old internet forum. The old internet forums. So the point is that these kind of memes and these kind of stories and aesthetics don't necessarily just stay on the internets stay in the places where they're made uh you may have heard that last week there was a huge um horror movie that got released and it's been really really well received and and it's called the back rooms and it's based on a meme that started on 4chan and other of these kind of horror memes that's sort of this officey office space um kind of unsettling vibe and that evolved into this massive hollywood uh movie another of these spaces from back in the day was called something awful.

21:49It was a sort of comedy forum and a similar anarchic vibe. And one of the things that came out of it, you may have heard of, is this thing called Slenderman. Now, Slenderman, if you've heard of Slenderman, you've probably heard because there was a really horrible attempted murder at a school in Wisconsin where these girls who believed that this monster was giving them instructions attempted to stab another of their classmates really horrible massive news caused this mass panic about this monster you know was internet trolls were they trying to make people kill people uh in that way that kind of media does when it doesn't quite understand the culture but when that when the stabbing happened the the artists who had collaborated to make slender man thought they had summoned a real monster into being somehow they were really really shaken by it that they had put this thing out into the world that had become bigger than the sum of its parts that they had collaborated to to create something that had taken on a life of its own and was doing things you know out in the world that nobody and no individual was in kind of control of i i thought that was really really interesting when I was back and covering that.

23:11But it was part of a new phenomenon from the internet of this kind of evolutionary collaborative art. And that's what the Far Lands really reminded me of. So given that, what you were saying, Nikki, like, has there been impact, either negative impact or positive impact that has come out of Far Lands? So Far Lands, while it's kind of going mainstream now, as Tom was saying, it's still kind of early in that process. Like there are some recurring themes that I think we can probably expect to see go wider and maybe take on this kind of life. But it's in that kind of primordial soup phase of art kind of collaboration.

24:03But the thing that is different is that there's something in this that's also a rebellion against the concept of algorithmic generation itself. You mean like algorithmic curation? Algorithmic curation, exactly. The fact that you're to share and be in the know of some of these kind of codes and numbers. And now that it's kind of gone mainstream and that's no longer the case, and if that rebellion is over, what happens next? What happens to the people who have gone into it as creators with this kind of sensibility and aesthetic? where do they take their art creation next? And is there a way for it to stay as kind of energizing as it is?

24:44What it says is that the internet is still human, right? It's shifted where it used to be just this totally independent rebel thing, but now it's this big corporate machine. But even for all the power that the tech companies do have over our lives, It doesn't work if there aren't people going onto it to express themselves and build community. And even within this system, people are finding a way to make it work for themselves and make it show them what they want to see and make it show other people the art that they're making. Because I think a lot of this stuff is art. And this system, despite the efforts of, you know, money, powerful forces to control it is still at its core a very human thing.

25:33I love that. Makes me actually want to go on the internet again, guys. Yeah, seriously. Maybe you should make a TikTok account, Karen. I kind of want to, what's the name of the AI monster? Loab. Yeah, Loab. Can we call Loab into, can we summon? I don't know if I want to call her. She seems bad. Yeah, I mean. Loab is actually me. I'm at the end of the AI generation. It makes so much sense. At the end of the generator. Karen is a monster that was created by AI. And now she's escaped containment. To scare you after you spent way too much time. Let's talk about something else. Yeah. Speaking of AI and creepy, creepy stuff.

26:20We received this really interesting comment from a listener to one of our episodes. This is from Grace, where she was talking about how she works with economically disadvantaged students from across the world, supporting them to access higher education internationally. And she started noticing this phenomenon where students would work on, they'd write an essay for an application to an institution that was wholly their own, you know, no AI assistance, no AI generation whatsoever. And then because they were worried that their own writing would somehow get picked up as AI through potential automated detection tools, they were pasting them into these AI detection software themselves to then tweak their language to make it seem less AI.

27:12But in the process, the changes were actually really bizarre. Like it wouldn't actually make the essay sound better or make it sound more human. It would just evade the AI detection software and introduce all of these new artifacts that were just actually not good writing. and she wrote to us saying, you know, like she's tried to explain to these students, you don't have to do this. It's not necessary, but they just can't get over this fear that because AI detection software is being used more and more and all around us and they have a higher chance of getting flagged as English as a second language speakers, they just insist on going through this process and actually degrading and devaluing their own writing even more.

28:02They're making their own writing worse. That's so depressing. Yeah. So this gets to like a really interesting question, which has been coming up a lot recently because there's this new AI detection software called Pangram that has become really, really popular. And recently there's been a spate of headlines about people being accused of having their writing being partly AI-generated or even all of it AI-generated because of people pasting that writing into Pangram and Pangram then assessing it and determining that, yes, there is some level of AI generation in it. So we talked about the Pope last week and his encyclical that I loved.

Read the full transcript

28:48There were accusations afterwards that some of his content, some of his paragraphs were between 40 % and 100 % AI generated because of Pangram. There was also a particularly famous example. Have you guys heard of the novel Shy Girl? No. So this is a horror novel that was supposed to come out and got pulled. Oh, yeah. Because of accusations. Yeah, because of accusations that it was AI generated. and the Pangram executive team also waded into this discourse and they were like, yes, our tool says, in fact, that a significant amount of it is AI-generated. And was it? Did it turn out that it was? I mean, this is the problem.

29:38So the writer, of course, absolutely denies it. She did mention at one point that she believes someone that she asked to help edit and review may have used some AI tools in the editing and reviewing process, but it was like not her doing that. But this is the challenge is that you can't really know. After that, there was a kind of a little wave of people putting classic literature into these AI detectors. And I think it found that like the first half of Moby Dick is AI generated. So yeah, so there's like all of these examples, but a lot of them end up pointing back specifically to Pangram because Pangram claims that it is the most accurate AI detection tool on the market and is becoming more and more popularly used, especially in academic contexts.

30:32And so the listener that wrote to us, Grace, her students were actually onto something. They are increasingly moving through a world that is being controlled by this kind of software. And what's interesting about this software is that even for Pangram, which there was one independent study, not yet peer-reviewed, but that did find that Pangram generally has a pretty high accuracy rate. Pangram says itself that it's false positive rate. So falsely identifying human generated text as AI generated text is one in 10 ,000. And it's false negative rate, Pangram says is one in 70. Just so I'm following the company itself says if we if our software says something is AI, we're very, very good at that.

31:30But if you put in text and it says it's human, we're not so sure we get that wrong all the time. Like one in 70 is not a good error. That's presumably because they want it to, like getting someone a false positive, getting someone accused of using AI is a worse outcome than if something slips through that's AI, right? Like they have it, I would guess, erring way on the side of when it's unsure. But even then, you know, one in 10 ,000, if that's marking kids' exam results, that's a really high number still, even if it is the number that they claim is their positive rate. This is going to affect people's lives.

32:07You know, that's... People are doing this on the scale of millions of times, not 10 ,000 times. Right. You know, academic careers can be ruined by this, even at one in 10 ,000. I did a story a couple years ago where I like talked to a bunch of families where high school students have been like suspended, gotten in serious trouble. This is a big thing with college students for a while. Yeah. Or getting accused of using AI. I talked to one kid whose teacher just clearly doesn't understand how the world works. like me, where they had a Google Doc where you could see in real time by looking through the history of the doc.

32:42They were writing the thing and editing it, but that wasn't enough proof because the teacher had put their essay into a piece of software that said this is AI and like nothing could change their minds. This has real consequences. This is one of the big challenges. There was this really great article in The Atlantic written by a former colleague of mine, Mateo Wong, who says that the fact that Pangram is significantly more accurate than other AI detection software potentially makes the situation worse because it garners more trust in the tool, even when the tool is still wrong. Like it can absolutely be wrong.

33:20And the errors go back to the way that this technology works in the first place. It's not like eventually the technology will improve so much that it will be correct 100 % of the time. No, it will always have errors. And the reason is because of the way that Pangram actually trains its detection software, it trains it just like actually companies train large language models. It gives this detection tool tons of examples of AI-generated text and then tons of examples of human-written text. And the reason why its accuracy, the company says, is higher than other types of detection tools is because of the ways that it really tries to give particularly tough examples to its AI software during training.

34:10So it will take human-generated text and ask ChatGBT, Gemini, Claude, like the most frequently used commercial AI chatbots to then generate a passage that is on the same exact topic so that it has a very, very close AI-generated pair to the human-written text. text. And that is part of the reason why it's been able to drive its accuracy higher and higher. But at the end of the day, when you think about it, first of all, these are probabilistic tools. So it's only basing the differences between human written and AI generated with statistical patterns that it's calculating. So it's literally never going to be deterministic.

34:54It's never going to be 100 % accurate. And second of all, the tools that give us AI generated text, were literally trained on human written writing. So the Venn diagram for human writing and AI-generated writing has a massive overlap in the middle. And no matter what you do to try and train an AI detection tool, it just will not be able to discern. You know, some humans will write like AI because AI is intended to write like humans. Right. And there's sort of an arms race, right? In the earlier days of LLMs, it became kind of well-known as a tell that they used the M-dash, which is the longest dash.

35:40And then it became almost like a running joke that if something has M-dashes in it, then it would be AI-generated. And there was a lot of writers I know who were like, we used to really like using the M-dash. Well, and the reason for that is that, at least what the reporting says, is the AI companies used the work of journalists in particular as examples of good writing. And journalists are famous for loving M dashes. We love M dashes. Yeah. Well, we used to. We can't anymore. Now we get accused of being AI if we use M dashes. There's a ton of examples of what looks like AI writing, which are, I think, tropes of my own personal writing, right?

36:23Like there's like, oh, AI always uses lists of three things, like first thing, third thing, second thing. which is, that's one of my favorite ways to write. I have, I'm sure you guys have experienced it. I mean, we all are people who get paid to write. I have found myself questioning my own writing because I'm like, I'd have to change this because I'm worried that people are going to think that I used AI to make this. Like, I am self-conscious about using MDashes, which are one of my favorite things now, because I'm like, are people going to think that ChatGPT wrote this? Yeah, yeah, yeah. Yeah, and that must be even, you know, we are in a position in our careers where we can, you know, demonstrate when it comes to it, we can like fight back.

37:07If that's our exam results and we're a school kid, there's not really an appeals process for that. Right. And there's a whole other side to this, which there's a new class of AI tools called AI humanizers, where you can take a piece of AI-generated writing or your own writing, and it will make it sound more human. Supposedly. At least that's the promise. Supposedly. It'll at least apparently make it sound less like a robot, which, you know, I've played around with these. The thing that I've noticed them do is they just make your writing a little bit less clear and less coherent, or they just make it a little more random, essentially.

37:48So there's this cat and mouse game between the AI detectors and the AI humanizers. But either way, I have talked to people You know, I've talked to college students in particular who say like, yeah, I write an essay. I do it by hand and then I put it in the AI humanizer to make sure that I don't get falsely accused of using AI. Yeah, I mean, this is the thing, right? So interestingly, Pangram actually trains, the company says it trains for the AI humanizers as well. So it is trying to also detect the AI humanizers. But the central problem is these humanizers aren't actually humanizing the writing, right?

38:27Like they're actually just distancing your writing from AI-generated writing, which actually just makes it worse. Using AI-generated writing, the stuff that the humanizer is coming up with is generated by an AI. there's also the fact have you guys heard this there's been a bunch of studies which have demonstrated that real human beings writing is starting to sound more like ai or at least that it's very clear that like our language on a societal level is being influenced and we're using tropes of ai writing in our own work because there's all this text the world is filled with this stuff and it is influencing us so there is this constant back and forth which i imagine makes probably AI detection a little more complicated.

39:13Yeah, it's such a fascinating phenomenon because I think the reason why AI detection tools, or supposedly the reason why AI detection tools exist is because it's meant to revalue human writing. It's meant to name and shame people that are not actually using their own minds to craft sentences, but it's actually just adding another layer onto AI influencing our writing overall. You know, we're being influenced both by some people using AI tools to generate the writing. We're being influenced by people actually just like, whether intentionally or unintentionally absorbing the language and rhythm of AI writing into our own writing.

40:03And then detection tools themselves are causing, you know, these poor students from this listener that wrote to us to literally edit their essays to sound actually less human, less proficient in English to try to avoid AI detection software. So in all of these ways, it's devaluing human writing and making it even harder for us to actually just have real meaningful control over the things that we want to say and express in the world. I think baked in here is another question, which is like, do we even need this stuff? Like, does it matter? And I think probably, you know, you might have a visceral reaction.

40:49I'd be like, of course it matters, right? We need to know whether something was written by human or people, by human being, or people are passing off, you know, robot slaw. But in certain contexts, maybe that's the wrong question to ask, right? Like this is, this always happens in academia, where some new problem comes up, and they go, oh, what we have to do is catch the cheaters, right? When the alternative is change the system, change the way you're evaluating students. Maybe you should be writing essays in class, maybe you should be adopting new research methods that fit within the world, which has AI in it, right?

41:26It feels like we need this because AI exists. There's no other way to move forward. But maybe there is. Maybe this isn't a question that is always so important for us to answer, given the potential consequences of the only tools we have at our disposal. Yeah, I've heard from teachers who are saying that they are going to start prioritizing methods like in-class debate or, you know, quizzes or, you know, conversation within the class as a better means than the essay of testing level of education. So maybe it's just simply the essay that's kind of had its time and, you know, teaching will go in other directions.

42:12Obviously, that's, you know, more difficult in education systems from lower economic backgrounds where you tend to have bigger classes and it's much harder to teach in that kind of way. But I think it's, yeah, the system has to change in some way because you can't just have these robots trying to detect these other robots because either way, the students are losing out. Yeah. I want to read you this quote that I really liked from an essay written by a researcher and artist named Eric Salvaggio. And he says at the end of his essay where he's reflecting on this Pangrim problem, if using AI to write is at its worst an industrialization of the mind, then AI detection at its worst becomes a surveillance system for thought.

43:02And I think this goes exactly to what both of you are saying, which is that We keep just trying in all of these ways to superficially fix or patch fundamental structural issues that AI is actually revealing in our society. It's revealing, you know, the inequities in education. It's revealing how the model of how we teach our students was not always actually designed around facilitating their learning. It was more just trying to mechanize the process of knowledge induction itself. And what I often say when I'm speaking to journalists or speaking to educators or, you know, any institution or sector that now feels like they're facing an existential crisis with AI is now is actually the moment to ask, why do you exist in the first place?

43:57And if AI is really beginning to steamroll some of the processes or procedures that you have developed over time, maybe that's actually an indictment of how you were doing things, like how you started to do things, because it might not actually be getting at the core or the essence of what you were meant to be doing in the first place. And so rather than using all of these different new AI, yet more AI tools to try to patch that over, use this as an opportunity to revert back to the fundamentals and reinvent how you should be doing journalism, doing teaching, doing education in and of itself.

44:41Thanks for listening. Remember, normally The Interface won't be here in the documentary. So if you'd like to hear more episodes from us, search The Interface wherever you get your BBC podcasts.

From the publisher

The Interface is your weekly guide to the tech rewiring your week and your world. Hosted by journalists Thomas Germain, Nicky Woolf, and Karen Hao, each episode unpacks, week by week, how technology is shaping all our futures. No guests. No jargon. Just three sharp voices debating the stories that matter - whether they shook a government, broke the internet, or quietly tipped the balance of power.

In this episode, Tom and Nicky head deep into the TikTok Farlands - the semi mythical place you supposedly reach if you scroll too far, too late, until your feed stops looking normal and starts serving up surreal, eerie and deeply unhinged videos. The name comes from Minecraft’s Far Lands, the glitched edge of the map where the world used to break apart, and TikTok users have borrowed it to describe the “end of the algorithm”: a strange zone of distorted edits, ominous warnings, weirdcore imagery and recurring figures like the now iconic fat bee playing the violin. TikTok’s Farlands have become a shorthand for what happens when doomscrolling tips into digital folklore.

But the Farlands aren’t just a joke. Tom and Nicky ask what this trend says about internet culture now. In a platform ecosystem dominated by polish, branding and optimisation, the Farlands feel like the return of an older internet: raw, surreal, handmade and proudly bizarre. At the same time, the meme also works as a critique of doomscrolling itself — turning algorithmic exhaustion into shared mythology, and making people newly conscious of how deep into the feed they’ve wandered.

So in this episode, we ask: is the TikTok Farlands a genuine return of weird, creative internet culture — or just another algorithmic genre?

Also in this episode: Karen looks at how AI detection tools may be changing the way we all write. As detectors spread through schools, publishing and professional life, students, teachers and writers are increasingly shaping their prose around what software might flag - dropping stylistic quirks, sanding off rhythm, and checking their own work in advance for fear of a false accusation. Researchers say the central problem is not just whether detectors catch AI, but how they balance false positives and false negatives in high stakes settings. And with a growing parallel market of “humanizer” tools promising to make AI text sound more human - and pass detection - the result may be an arms race that leaves everyone writing in a flatter, safer and more paranoid style.

To hear more, search The Interface wherever you get your BBC podcasts.

More from The Documentary Podcast

All 508 episodes
Introducing: The Interface - What goes on in TikTok's Farlands?The Documentary Podcast · 42 min
Listen in VO