What Sora's 3.8 Million Downloads Tell Us About the Next Phase of Social Media
AI-generated video has arrived as a consumer platform—ready or not, brands need to understand what that means
Jake Paul doing a “get ready with me” makeup tutorial. Michael Jackson pulling pranks at KFC. Cats attending church services. Stephen Hawking delivering a TED talk. George Washington ding-dong ditching grandmas.
Welcome to Sora, OpenAI’s AI video generation app that’s become a TikTok for the surreal. In its first month, Sora hit 3.8 million downloads in the US and ranked as the #4 app overall—despite being invite-only and iOS-exclusive. That’s not a niche experiment. That’s a cultural moment.
The question for brands and marketers isn’t whether to pay attention. It’s what to do when AI-generated social media becomes as normal as user-generated content—and just as uncontrollable.
What Sora Actually Is
Sora launched in late 2024 with a simple premise: upload a video of yourself, describe what you want to see, and the AI generates a video with you (or anyone) doing whatever you imagined. Want to fly? Done. Want to give a fake TED talk? Easy. Want to insert yourself into a 1980s movie scene? The AI will figure it out.
The app surfaces videos in a feed, similar to TikTok’s “for you page.” You can use other people’s “Cameos”—uploaded personas—with their permission, which gets notified to approve or delete. The result is an explosion of meme-like content that ranges from absurdist comedy to unsettling deepfakes of public figures.
OpenAI has tried to install guardrails. You can’t create content depicting copyrighted characters without workarounds. You can’t use certain public figures’ likenesses. But users are creative. Instead of asking for the DeLorean from “Back to the Future,” they’ll prompt for a “stainless steel time machine car from the 1980s.” The AI knows what you mean.
The app’s popularity reveals something uncomfortable: people really, really want to make deepfakes. Not for malicious purposes necessarily—mostly for entertainment and creative expression. But the line between satire and misinformation is blurry when the tools are this powerful and this accessible.
How Brands Are Showing Up (or Not)
Most brands aren’t actively creating on Sora yet. Partly because it’s still invite-only. Partly because most marketers don’t use ChatGPT as a content channel, and you need a ChatGPT account to access Sora. Partly because brand safety questions remain unanswered.
But some agencies are experimenting. Mischief created a video of boxer/influencer Jake Paul touring their offices. VML’s director of design experience, Luke Hurd, has become a power user, uploading three videos daily and experimenting with turning himself into different characters—goth, Detroit rapper, someone stumbling home disheveled after a night out. He’s treating it as a creative sandbox to understand what’s possible.
The early pattern is interesting: the brands showing up aren’t making polished ads. They’re making weird, self-aware content that embraces the platform’s chaotic energy. That feels right for Sora’s current culture, which rewards absurdism over production values. But it’s unclear whether that approach scales or whether brands will inevitably try to make Sora look like every other social platform they’ve colonized.
There’s also the question of what happens when Sora opens to the public. Right now, it’s creators, early adopters, and marketing experimenters. Once it’s fully accessible, brands will face pressure to have a “Sora strategy”—even if the ROI is unclear and the risks are high.
The Brand Safety Problem That Isn’t
Every conversation about Sora includes concerns about brand safety. What if your brand appears in someone else’s video without permission? What if AI-generated content creates negative associations? What if competitors use Sora to mock your brand?
These are reasonable questions, but they’re not new problems. User-generated content on TikTok has created the same risks for years. The difference is that Sora makes it easier to create more realistic-looking content, which might cross from obvious parody into believable misinformation.
But here’s the thing: Sora videos are clearly AI-generated. The physics aren’t quite right. The facial expressions are uncanny. The movements have a dreamlike quality that marks them as artificial. Unlike deepfakes designed to deceive, Sora content announces itself as synthetic. The platform even includes watermarks and C2PA metadata to preserve provenance.
The bigger risk isn’t misinformation. It’s irrelevance. If Sora becomes a major social platform and brands sit it out because they’re worried about safety, they’ll miss the opportunity to shape how their brands are perceived in AI-generated media. User-generated Sora content will happen with or without brand participation. The question is whether brands want to be part of the conversation or watch from the sidelines.
What Makes Sora Different from Meta’s Vibes
Meta launched a similar AI video feed called Vibes in August 2024 within its AI app. It hasn’t gained nearly as much traction—ranking #95 in the past month compared to Sora’s #4 position.
The difference is “Cameos.” Vibes creates abstract AI-generated videos, but you can’t insert yourself or other people into them. They’re more like ambient AI art—safer, less engaging, and ultimately less interesting. Sora’s Cameos feature creates personal investment. You’re not just watching AI videos; you’re starring in them. That’s compelling in a way that abstract AI art isn’t.
It also creates ethical complexity that Meta probably wanted to avoid. Cameos enable deepfakes, parody, and satire that could target real people. OpenAI has positioned this as a feature—you can approve or delete videos that use your Cameo persona—but it’s easy to imagine scenarios where that approval system breaks down at scale. What happens when Sora has 100 million users and someone’s Cameo is used in thousands of videos daily? Can one person reasonably monitor and approve all of it?
The Glut Problem
Here’s what keeps me up at night: Sora democratizes video creation in a way that could flood social platforms with low-effort, AI-generated content. Right now, creating quality video requires time, equipment, and skill. That’s a natural filter keeping bad content from overwhelming good content.
Sora removes that filter. Anyone can generate convincing-looking video in seconds. If even 1% of users start posting daily AI videos, the sheer volume could make discovery impossible. We already struggle with content overload on TikTok and YouTube. Adding a tsunami of AI-generated videos would make the problem exponentially worse.
This is the same issue facing text-based AI content. ChatGPT made it trivial to generate blog posts, social media captions, and marketing copy. The result has been an explosion of mediocre content that clogs search results and social feeds. Google has responded by down-ranking low-quality AI content in search results. Social platforms will likely do the same for AI video—but that takes time to develop, and in the meantime, we’ll all wade through the slop.
For brands, this creates both challenge and opportunity. The challenge is standing out in a sea of AI-generated content that all starts to look the same. The opportunity is that authenticity becomes more valuable when fakery is ubiquitous. Real people, real stories, real creativity—these become differentiators when everything else can be synthesized.
What Brands Should Actually Do
First, experiment privately. Get a Sora account. Play with it. Understand what’s possible and what’s clunky. You don’t need to post anything publicly, but you need firsthand experience to make informed decisions about whether and how to use the tool.
Second, develop a point of view on AI-generated content. Should your brand embrace it? Avoid it? Use it selectively for certain types of content? There’s no right answer, but you need a clear position that aligns with your brand values and audience expectations.
Third, think about how AI-generated content fits into your broader content strategy. It’s a tool, not a replacement for everything else. Maybe it’s useful for rapid prototyping of concepts. Maybe it’s great for social experiments and trend-jacking. Maybe it’s not right for your brand at all. But make that decision deliberately rather than ignoring the tool and hoping it goes away.
Fourth, pay attention to how your audience responds to AI content. Some demographics love it. Others find it creepy. Some people don’t care as long as it’s entertaining. Your audience will tell you whether AI-generated content works for your brand—but only if you test it.
Finally, be open to the possibility that AI-generated social media might not be a passing fad. We’re at the beginning of a long transition where the line between “real” and “generated” content becomes increasingly blurred. Brands that figure out how to navigate that ambiguity early will have an advantage over those that wait for the rules to clarify.
The Longer View
Sora is one product from one company in a rapidly evolving category. Meta will iterate on Vibes. Google will launch something. Chinese companies will release competitors. The specific tool matters less than the trend: AI-generated video is becoming accessible to consumers, and it’s becoming social.
That changes the dynamics of social media in ways we’re only beginning to understand. For decades, social platforms have been about sharing what already exists—photos, videos, experiences. Now they’re about creating what doesn’t exist. That’s fundamentally different.
For brands, it means the old playbook of “create great content and distribute it” needs updating. When anyone can create any video they imagine, distribution matters more than ever. But so does authenticity, creativity, and the human judgment to know what resonates and what doesn’t.
AI tools can generate content. They can’t yet figure out what content is worth creating in the first place. That’s still a human skill—and likely to remain one for a while.