Ericka Pinon

Did You Hear SZA Was Paid to Be Justin Bieber’s Coachella Backup? Let’s Break That Down.

This weekend I was scrolling through Instagram recently when I came across a post claiming that SZA was allegedly paid six figures to stay “Coachella-ready” just in case Justin Bieber backed out of his headlining performance. As someone who follows both artists, my first reaction was just confusion. It felt like one of those things that sounds just believable enough to spread, but also kind of off. That mix of “maybe?” and “wait, really?” is usually my sign to actually look into something before believing or sharing it.

Step 1: Look at Where the Claim is Coming From

The original post came from an Instagram account called Blind Items paired with a page called “Music Life.” Right away that raised a flag. “Blind Items” is literally built around unverified celebrity gossip, rumors posted without naming sources or providing proof. The headline also used the word “allegedly,” which signals that nobody actually confirmed anything. That word alone should always make you pause before sharing something.

Step 2: Check if Anyone Credible is Reporting It

I did a quick search to see if any well known outlets were covering this story. When something is actually true, especially something as specific as a six-figure celebrity deal, you would expect sources like Billboard or Rolling Stone to be reporting it. I found nothing from any verified news source backing the claim up. If a story is only living on gossip Instagram pages and nowhere else, that is a strong sign it has not been confirmed by anyone with actual accountability.

Step 3: Go to Primary Sources

This is the most important step and honestly the one that settled everything for me. Not only did SZA comment directly on the original Instagram post calling the rumor out, but she also personally stated she was in New York and had not been paid a single thing. That comment on the actual post where the rumor started is about as primary as a source gets. On top of that, E! News covered her direct response and quoted her full comment for anyone who missed it. When the actual person a rumor is about goes out of their way to deny it twice, in two different places, that is a pretty definitive conclusion.

Step 4: Think About Why It Spread

Even after knowing it was false, it is worth asking why this rumor gained traction. Coachella 2026 had a lot of buzz around Justin Bieber headlining after a long public absence due to health issues. People were already uncertain about whether he would follow through, which created the perfect environment for a rumor like this to feel believable. Misinformation spreads fastest when it taps into something people are already worried about or curious about.

Conclusion

After going through these steps, the claim was clearly unverified gossip that SZA herself shut down in multiple places. But beyond this specific rumor, the bigger takeaway is that pausing before you repost something is always worth it. Taking two minutes to check the source, search for other coverage, and look for a direct response can save you from spreading something completely false. If a headline uses words like “allegedly” and only appears on gossip pages, that is your sign to dig deeper before hitting share.

Fighting Fake News: How X and Discord Are (Trying to) Handle Misinformation

If you’ve spent any time online in the last few years, you’ve probably seen something that made you think, “wait, is that actually true?” Whether it’s a wild claim or video that looks real but totally isn’t, misinformation is everywhere. Two platforms I use pretty regularly are X and Discord, and honestly, their approaches to handling this problem couldn’t be more different. One is very public and controversial, the other is more behind the scenes and honestly kind of easy to forget about. Let’s break down what each of them is actually doing, and what they should be doing better.

X:

When Elon Musk took over Twitter in 2022 and rebranded it as X, one of the biggest changes was gutting the platform’s traditional content moderation teams. The company cut employees who worked on election integrity and countering misinformation, loosened rules around hate speech, and reinstated accounts of extremists who had previously been banned. On top of that, X scrapped a user-reporting feature that had allowed people to flag posts as “misleading.”

So what replaced all of that? Basically one thing: Community Notes, a crowdsourced fact-checking system where regular users add context or corrections under posts they think are misleading. Notes are applied through a bridging-based algorithm that requires agreement from users on different sides of the political spectrum, rather than just a majority vote.

In theory, this is actually a cool idea. And there’s real research suggesting it can work, posts with public correction notes were 32 percent more likely to be deleted by their own authors than those that only received private notes. University of Rochester That means the system can pressure people into correcting themselves.

But the notes often don’t show up fast enough. “Content spreads rapidly across X, and if a note comes too late, few users will get a chance to see it. Notes that take 48 hours or so to go up have almost no effect.” During high-stakes moments, the system falls apart entirely. After the attempted assassination of Donald Trump in 2024, only five Community Notes were published to counter the 100 most popular conspiratorial posts about the shooting. Around the 2024 election, 74% of misinformation in a sample of posts failed to receive any notes at all. 

From personal experience, Community Notes, when they appear, are actually pretty informative and usually link to credible sources. But I’ve also watched obviously false things go viral with zero notes attached, especially in those first viral hours. The system just can’t keep up.

Discord:

Discord is a totally different beast. While X is public, Discord is built around private or semi-private servers, basically invite-only group chats. This makes misinformation harder to see, but not harder to find.

Discord does have an official policy. The platform prohibits communities that share false or misleading health information likely to cause harm, including anti-vaccination claims and dangerous unsupported treatments, and relies on credible sources like the WHO and the CDC. There’s also a civic policy: Discord states it will remove harmful misinformation about civic processes when it becomes aware of it. 

Sounds good, except enforcement is the problem. Discord’s content moderation strategy for private groups appears to be reactive rather than proactive, depending primarily on reports from users inside those groups rather than automatic filtering. If everyone in a server believes the misinformation, nobody reports it.

As a regular Discord user, I’ve seen this play out. Some servers have active mods and clear rules. Others are basically the wild west. The platform offloads almost all responsibility onto individual server owners, which works great sometimes and completely fails in others.

Potential Change

For X, the biggest fix needed is speed. Community Notes is a solid concept but too slow for viral misinformation. X should use AI to flag potentially false posts faster and prioritize them for contributor review. Research found that rapid algorithmic fact-checking, though imperfect, was more effective than slower professional fact-checking suggesting a hybrid AI-plus-human approach could be powerful.

For Discord, the reactive-only model isn’t enough. The platform should invest in proactive scanning tools that can detect misinformation patterns before users have to report anything. Requiring larger servers to maintain baseline moderation standards would also help close the gap.

Both X and Discord have made some effort, but neither has truly solved the problem. X is betting on a crowdsourced system that’s too slow. Discord is hoping users police themselves, which is asking way too much. Misinformation thrives in exactly these kinds of gaps, and closing them should be a priority, not an afterthought.

Your FYP is Lying to You

Okay, I’ll just say it, I’ve been fooled by a deepfake. More than once, probably. A while back a video showed up on my feed of Taylor Swift promoting a Le Creuset cookware giveaway. “Hey y’all, it’s Taylor Swift here,” she says in the clip. “Due to a packaging error, we can’t sell 3,000 Le Creuset cookware sets. So I’m giving them away to my loyal fans for free.” It sounded a little off, her voice had this slightly robotic cadence to it, but I didn’t think twice. I kept scrolling. Later I found out it was a completely AI-generated deepfake designed to trick people into clicking a link and handing over personal information. Le Creuset had nothing to do with it. And I just sat with that for a second, because even knowing something felt a little off, I hadn’t actually stopped to question it.

I’m writing this because I think that experience is really common right now, especially for people my age. We’re on our phones constantly, which means we’re exposed to a massive volume of video content every single day. We’ve gotten really good at processing it fast. But “processing fast” and “actually evaluating what’s real” are two very different things, and deepfake technology is specifically designed to exploit that gap.

So here’s what I actually want you to walk away with, not anxiety, not paranoia, just a few concrete things to look for and a better mental model of what’s happening when you scroll.

What even is a deepfake?

The word gets thrown around a lot but I want to make sure we’re starting from the same place. A deepfake is a video or audio clip that has been manipulated, or in some cases, entirely generated, using artificial intelligence, specifically a type called a generative adversarial network, or GAN. Basically, two AI systems run against each other, one tries to create fake content, and the other tries to detect whether it’s fake. They keep improving each other until the fake is convincing enough to pass. The result can be a video of a real person saying something they never said, or doing something they never did.

The term “deepfake” comes from combining “deep learning” (a type of AI) and “fake.” It started getting mainstream attention around 2017, but the tools have gotten dramatically more accessible since then. You don’t need to be a programmer or have expensive hardware anymore. A deepfake video can now be created for about $5 in under ten minutes. That’s a pretty significant shift in what the average person is capable of creating, and in what your average viewer is up against.

According to a 2023 McAfee survey, 70% of people said they weren’t confident they could tell the difference between a real video and a deepfake. And that data is a couple years old now. The technology has only gotten better since then.

Why should you care?

The obvious harm is misinformation, fake videos of politicians saying alarming things, fabricated clips that spread before anyone can debunk them. However, it goes further than that. Deepfakes are being used to impersonate celebrities in financial scams, like the Taylor Swift example above, or fake investment videos featuring Elon Musk, Prince William, and the UK Prime Minister. They’ve been used to create non-consensual intimate images of real people. They’ve been used to harass and embarrass private individuals. And they show up in our feeds mixed in with genuine content, often from accounts that look totally normal.

The part that really gets me is how deepfakes work with the way our social media feeds are designed. Algorithms reward things that get engagement fast, likes, shares, comments, emotional reactions. A shocking, convincing fake video can rack up thousands of shares before anyone bothers to check whether it’s real. In one documented case, a deepfake ad campaign featuring a Financial Times journalist reached nearly a million users across the EU in just six weeks. By the time a correction or debunk is posted, most of the people who saw the original have already moved on. That’s not a coincidence. It’s a feature of the system, even if it wasn’t intentionally designed to spread lies.

Okay but how do you actually spot one?

Here are the real visual cues that researchers and journalists use, things you can actually look for in the moment, not after you’ve already shared something.

Watch the edges of the face. Hair and the outline where the face meets the background are notoriously hard to fake. Look for blurring, weird flickering, or a halo effect at the hairline.

Look at the eyes. Blinking is often unnatural in deepfakes, too infrequent, too fast, or slightly off-rhythm. Also look for lighting inconsistencies, such as if the eyes reflect light the same way the rest of the face does?

Check the skin texture. AI-generated faces often look a little too smooth, especially in motion. Real skin has texture and pores that are difficult to replicate consistently across frames.

Pay attention to teeth and the inside of the mouth. This is one of the hardest things for AI to get right. Look for blurring, weird shapes, or teeth that don’t quite look like teeth.

Notice if the audio matches the mouth movement. Even a slight delay or disconnect between speech and lip movement is a major red flag, and looking back, that robotic cadence in the Taylor Swift clip was exactly this. Researchers found that humans can only detect AI-generated speech about 73% of the time, which sounds okay until you realize that means roughly one in four deepfake audio clips slips right past us.

Consider the source. Where did this video come from originally? Who posted it? Is there any identifiable context? A surprising or emotionally charged claim with no clear sourcing is worth slowing down on. A free giveaway from a major celebrity with zero official announcement anywhere? That’s a red flag hiding in plain sight.

None of these are foolproof, newer deepfakes are getting better at all of these things. Although, the point isn’t to become a forensic video expert, it’s to develop the habit of pausing before you react or share, especially with content that’s designed to make you feel something strongly.

What to do if something feels off

If a video is making a specific claim about a real person, go look for it somewhere else. Not just a different social media account, an actual news outlet or primary source. If something shocking genuinely happened, it’ll be covered. If you can’t find any corroborating reporting, that’s information. Had I taken ten seconds to search “Taylor Swift Le Creuset giveaway,” I would have found nothing, because it never happened.

You can also do a reverse image search on stills from a video using Google Lens, which sometimes surfaces the original context or earlier versions of the footage. It’s not a magic solution but it takes about fifteen seconds and has saved me from sharing something embarrassing more than once.

And honestly? It’s okay to just not share something you’re not sure about. That’s not cowardice. It’s just being a thoughtful person in a world where sharing something false has actual consequences for real people.

The bigger picture

I don’t want to end this on a doom-and-gloom note, because I actually think there’s something kind of empowering about understanding how this works. A lot of misinformation thrives on the fact that people feel overwhelmed and don’t know where to start. Once you understand that deepfakes work by exploiting how fast we scroll and how much we trust visual information, you have something real to work with. You’re not just a passive viewer anymore.

The technology is going to keep getting more convincing. The average American now encounters about 2.6 deepfake videos per day, with people ages 18–24 seeing even more, roughly 3.5 daily. Our ability to think critically, slow down, and ask basic questions doesn’t get outdated. That’s still the best tool we have.

So the next time your FYP serves you something wild, take an extra two seconds. Check the edges. Watch the eyes. Look for where it came from. It’s not that deep, it’s just paying attention.