When AI Grandmas Go Viral, What Even Is Reality Anymore?

So I'm about to watch Trailer, it's a Swedish film that has been dubbed in English with AI, and I have no idea what it's about. Here we go. Nobody knows what happened to my dad. Okay, her dad maybe was kidnapped by aliens. There isn't a ton of dialogue in this trailer, but if you look really closely at their mouths, you can see maybe that it's AI-generated lip movement, but it's really hard to tell. I'm Nicole Nguyen, personal tech columnist at The Wall Street Journal, and like you, I've seen more uses of artificial intelligence in movies, music, and social media. Okay, well, I feel like my heart rate is 150Bs per minute right now after watching that trailer, so only at the end when she's like, I'm gonna go find my dad. The mouth didn't match up exactly perfectly, but throughout this trailer, if I didn't already know that the dub was AI-generated, I don't think I would have been able to tell. What you just heard is the trailer for the Swedish film Watch the Skies. Their production team used AI to adapt the actor's mouths to match the English language dub for its American release. I found it believable, at times. The latest crop of large language models, pairing what we know today as "generative" AI, are making it faster, cheaper, and easier to produce movie magic, and slop. Now it seems like AI versions of real people are everywhere, and popping up in unexpected places. Last year, the BBC used Agatha Christie's likeness to teach an online writing course. I must confess that this is all rather new to me. An earlier this year, an AI Val Kilmer delivered a controversial posthumous performance for the film as deep as the grave. Don't fear the dead, and don't fear me. Gen AI is advancing at a rapid clip, looking more and more realistic and becoming more accessible with each new model. More and more, it can be hard to tell what's AI and what's not. With more AI-generated images and sounds filling our screens, is it changing our perception of what's real? Today, and for the next couple of Saturdays, we'll examine just that, from entertainment and marketing slop to propaganda and increasingly sophisticated deepfix. We'll bring you conversations with AI researchers, content creators, and AI character, and the very people they fool. It's Saturday, August 1st. This is episode one of our special series AI and the blurring of reality. Granny, tell me about yourself. How old are you and where do you live? I don't count years, honey. I count zeros. I live wherever my jetlands and the champagne's cold. Ask me again next divorce. That's Granny Spills, and she's not real, or at least not human. She's an AI-generated grandma that went viral on social media, has been interviewed by Time magazine, and also sat down to talk with me. Or not exactly. This was my first interview ever with an AI-generated person. You have been to so many places this year. Coachella, Rolling Loud, Meaconos, which is your favorite and where do you find the energy? Monaco. The yachts are big, the men are rich, and nobody asks me to split anything. The energy comes free when you don't lift a finger or a single bag. As you might imagine, the interview didn't follow the typical process. I pre-recorded my questions, and the creators generated the responses. For interviews with humans, I don't typically send questions in advance, but made an exception just this once. As advanced as AI tools have become, they have a harder time spinning up real-time responses, at least for the time being. Now, let me tell you a little more about Granny Spills. Decked out in pink and luxury brands, this grandma is a three-time divorcee who got all of her money from her ex-husbands. She's loud, a bit of a potty mouth, and unapologetic about her belief that women should use men for their money. Her brazen persona has gotten her two million followers on Instagram. As for how many of those two million realize Granny Spills isn't real, that's anybody's guess. Blurr Studios, the digital influencers agency that made Granny, says it doesn't hide the fact that she's AI-generated. But if you're scrolling quickly on your phone, you might be fooled. Take one of Granny's most viral posts from Coachella this April. In a series of photos, she's posing with a singer Justin Bieber and his wife Haley. It's got nearly a million likes. Here's a sampling of some of the comments. You are so fire emoji, fire emoji, fire emoji, fire emoji. Where is that floral matching set from? This is so wholesome, nice to see Justin smiling. Wow, the amount of people that don't realize she's AI, it's a little scary. When I saw her first video, I thought she was real, until I actually googled it. How do you feel about being AI-generated? Honey, I've been called fake by men with real faces and empty wallets. At least I'm honest about my filter. While the character is AI-generated, the inspiration behind her is very real. So the grandma actually pre-resumbles my real life grandma. Eric Sarrez is the CEO of Blurr Studios. She was always sassy and pretty savage, but honestly, it was just a character that we know being in social media too, that the internet really loves elderly people. And when you see this cussing, savage, grandma character, it's really like, did she just say that? So it was really a shock value-driven content that we knew the audience would really resonate with. Sarrez started his career as a content creator, doing street-style interviews. What is your GPA? It is a 3.5. Do you think other people could guess that you have that? Yeah. Then when GNI tools became widely available, he saw an opportunity to scale. We saw a future where this AI content is, you're not going to be able to tell the difference between what's AI and what's real life content as the technology improves. So we did exactly that. We just hopped on, learned by doing as an attempt to future-proof our street interview UGC agency. UGC stands for user-generated content. We wanted to adapt and make sure that we're ahead of the trend. And we found instant success creating, I guess, these AI characters. Granny's account grew from 0 to 1 million followers on Instagram in the first 22 days. So we saw growth on social media far more than what we've ever experienced in our own social media careers. Blur Studios created Granny Spills using Google's VO3. It's a video generation model. Like other AI chatbots, you type a prompt in a text box and the model then spits out a video. You can even use the tool to create audio, like sound effects and dialogue. We had multi-paragraph-long prompts describing every single detail of Granny's outfit, the people in the background, the location, the tone of her voice, everything, which is still what we do today. We asked the company's co-founder and COO Adam Vassarstein to read one of their prompts. The elderly woman is wearing a pink Gucci monogram head scarf, oversized pink deor glasses, pink Louis Vuitton coat with the signature LV pattern, and carrying the pink Chanel credit bag with obvious chain straps. Spits with a rapid fire delivery and a mature, elderly grandmotherly voice. That is literally just two lines out of the probably two hundred lines. Granny Spills is just one of several AI characters created by Blur Studios. While they say they're making some money through brand sponsorships, they're also betting on a burgeoning market for creating AI influencers for other companies, like retired corporate boardroom Big Shot Nana Nessa. When your boss says we're like family here, here's what they really mean. And man on the street interviewer Joey on the street. What's it like living with your best friend? Even before the popularity of Gen AI, people have been raking in real dollars from these digital characters. Cameron James Wilson, for example, created Shudu, the world's first digital supermodel nearly a decade ago. Her portrait went viral across social media. It was reposted by so many different people, Tyra Banks. And they were all asking who is she. And they thought that maybe she was a real person. And this was the very, very first time that I'd ever thought that maybe technology had gone to a point where we could actually make somebody who wasn't real pass as a real person. Shudu is beautiful. Maybe too beautiful. She has a dark, poreless complexion and is often adorned with gold jewelry. She's based off a princess of South Africa Barbie Wilson had when they were young. And while she may not be real, Shudu, who has over 236,000 Instagram followers, is influencing consumer behavior, driving sales of partner products like luxury clothing and accessories. People are trusting these virtual characters not as much as human influences, but they are actually buying from these accounts. They're buying from our accounts that are buying from other accounts. Shudu's success inspired Wilson to launch the digitals, an agency that uses AI to create characters for brands. In the cases of both Shudu and Granny Spills, the creators behind them make it known that the influencers are AI generated. Here's Eric Sarrez again. At the early stages, people didn't even really know to decide for between AI content and real life content. Naturally, that audience tended to be anywhere between 20 and 60 years old. So some of the older viewers just had no media literacy or internet literacy to even decipher the difference. But as the tech grew and people became more adapted to AI content on the internet, it's becoming a lot more obvious. And it's also disclaimed in our bios and everything. We don't try to hide it the fact that she is AI and in fact are more so leaning into the fact that she is AI so that we can do crazy things that a grandma couldn't do in real life. Blur Studios isn't hiding the fact that Granny Spills is AI generated. When looking at the posts on the Instagram app, you can see the AI creator tag on the account and each post. Though as of this recording, it's not visible in the desktop version. Blur and other creators aren't exactly going out of their way to make it more obvious. Many posts are marked as a collaboration with BlurStudios.AI. And many include hashtags like hashtag Granny Spills, hashtag relationship advice, hashtag personal finance, hashtag world cup even on a photo of Granny with Leo Messy. They don't typically say hashtag AI. The posts are designed to make you slow your scrolling to take a second look. Some are more obvious AI creations than others. For any influencer, the value isn't getting you to pay attention. I was nearly duped by one picture of Granny clinking a glass of champagne with a Kardashian. I did a double take and thought, who is this woman? It all made me wonder how the rise of AI influencers is impacting our ability to discern what's real and fake online. Can synthetic images be benign? What does it do to us as it fills our feats? I think it puts people in a perpetually skeptical mindset or a feeling that they can't know if it's true and that, you know, creates mental tax. We'll dive into that after the break. The explosion of AI tools like the one's Blur Studios used to create Granny Spills, and how quickly the tech has been advancing, has rapidly increased the amount of AI generated content online. Do you have a sense of how much AI manipulated imagery and video is out there? Is that a knowable thing? I don't think it's knowable. Honey Fareed is a digital forensics expert and professor of computer science at Dartmouth College. He specializes in detecting deepfakes and manipulated media. I don't think we know because first of all, it's a big internet with billions of pieces of content uploaded every day. One, being able to accurately determine what is real and what is fake in the best of situations is exceedingly difficult. And at the scale of the internet, it is essentially impossible. But what I can tell you is that anecdotally, you're seeing more and more of it. And it's becoming harder to tell what is AI slop. Fareed says while in the past there would have been some obvious tells. These days, the probability of recognizing if something is AI or not is basically a coin toss. We do perceptual studies in my lab where we show people images, we haven't listened to audio, we haven't seen videos. They're basically a chance. They're slightly better than chance at distinguishing what's real and what's fake. And that is a remarkable advance in just a handful of years. Because there are less tells like the six finger tell or. Yeah, it's exactly why. It used to be the six fingers and it was the, you know, the temporal incoherence when Will Smith was eating spaghetti in that first video. The spaghetti was flying all over the place and his face was distorted and now it's perfectly seamless and beautiful. Videos of cars. The cars don't suddenly go backwards. They don't suddenly change color. They have temporal coherence and consistency and this is a one way street. Understand, it's not going to get worse. It's only going to get better and better and better. Fareed says the worst part of that advancement in technology is that we don't know how bad we are at distinguishing between real and fake. I don't mind being bad at something. I mind you not knowing you're bad at it because that makes you very, very dangerous. And part of the reason it's dangerous is because of the sheer quantity of content. It's an information overload, creating sort of a virtual version of the boy who cried wolf. To put it more plainly, if you're constantly seeing posts that aren't real, eventually you'll begin to question those that are real. We can spend the next half hour talking about all the fake content but where things get really complicated is when real things are happening, people are now doubting them. So we've hit this point where now I just say you have to be not looking for tells and mistakes so much as just being wary of the interaction that you're having and was the likelihood that this is a real interaction or a plausible scenario. That's Renee D'Aresta. She's an associate professor at Georgetown University. She researches how the influence of media works in the digital age. She says now it's easier to just assume everything we see online is AI but that skepticism isn't necessarily good. As you note is Coachella Granny likely but the problem is we do hit this point where a lot of the scenarios are actually likely and that is really the issue. I think it puts people in a perpetually skeptical mindset or a feeling that they can't know if it's true and that is you know creates a mental tax. So we're seeing AI in our ads in our influencers in propaganda, most aspects of media. What does this constant flow of AI slop and not slop do to our ability to figure out what is truth, what is not truth and maybe that question isn't as binary. What does an influx of AI generated imagery do to us? I think it does influence the confidence levels that we have in our ability to tell what's real and I see videos that come across my feed now and I think is this real and this is what I do professionally at this point. I see a person and I wonder is this person real is this influencer who all of a sudden came out of nowhere, has a massive following and it's just producing effective rage bait and happens to be very attractive. Are they actually real? I don't know because we see so many of these folks who've just recognized that rage bait is highly effective, easily monetizable. But the problem goes beyond rage bait or having another strong reaction to a post. The inundation of AI slop also risks having us not react at all. Here's honey for it again. Everybody's overwhelmed and we're just people are giving up and it's like I don't know what to do. I mean I just you know I'm going to just go into a fetal position under my bed and we'll just hope for the best. In the lack of certainty is already creating broader problems on the platforms themselves. We asked Aiden Walker about that. He's written extensively on internet culture. He used to work for a website called Know Your Meme, the internet's meme encyclopedia. What are some of the issues you've seen arise from AI-generated content circulating around the web? So I've grouped them into two categories. There's the issues like the six fingered one which kind of live at the level of the post. The post is deceptive. The post is sloppy. The post is a scam. And then I think there's the problems that lie more at the macro level. Like systemically, what does all of this AI-slop content do to social media platforms, which are social infrastructure at this point. People rely on them for their businesses. People rely on them for their information. People rely on them for community building. About half of Americans rely on social media for news, according to a 2025 survey by the Pew Research Center. And increasingly that news is coming from creators, whether they're a seasoned journalist or fake ones that can be generated by anyone who knows how to write a prompt. Reliability may vary. On top of that, the stakes can even hire when you consider who is behind the AI content and what their intentions are. anecdotally, we are seeing a lot of this content, both for perfectly innocent purposes but also quite nefarious purposes. And we've seen seen it across the boards, where people either for ideological purposes, state-sponsored actors trying to sow civil unrest, people simply trying to monetize, are flooding the social media airwaves, if you will, with fake content. But we've also seen it coming from official channels. We'll get into that and more in our episode next week, and that's it for the first installment of Tech News Briefings Special Series, AI, and The Blurring of Reality. Today's show was produced by Julie Chang, I'm your host, Nicole Nguyen. Michael LaValle wrote our theme music and mixed this episode. Our supervising producer is Katie Ferguson. Our development producer is Aisha Elmesleem. We had additional support from Wilson Rothman, and Chris Innsley is the deputy editor of audio, The Wall Street Journal. We'll be back Monday morning with a TNB Tech Minute. Thanks for listening.