Meta getting an AI app to 5 million downloads in 22 days doesn’t impress me as much as it’s supposed to. It’s not nothing, but it’s also not the same kind of “people can’t live without this” signal that everyone wants to read into it.
Because downloads aren’t love. Downloads are distribution.
From what’s been shared publicly, Meta’s new AI agent, Muse, hit 5 million downloads 22 days after launch. The comparison that’s making the rounds is that this beats ChatGPT’s earlier pace, which reportedly took 56 days to reach the same number. Muse also launched on both iOS and Android at the same time, and it’s been backed by heavy advertising. It even held the number one spot on the U.S. App Store’s free iPhone chart for 12 straight days.
That’s the story. Here’s my read: Meta didn’t just build an app. Meta flexed the machine.
If you can put an icon on millions of home screens quickly, you can “win” a launch. And Meta is probably the single best company on earth at pushing people toward a behavior at scale. They’ve been doing it for years, and not always in ways anyone should be proud of. So yes, they can drive installs. The more interesting question is what happens on day 30, day 90, day 180—when the ads calm down and people decide what stays on their phone.
This is where the hype gets slippery. People are treating “surpassing ChatGPT” like Muse proved it’s a better AI. But this isn’t a science fair. This is a distribution war. If you show up on both major phone platforms on day one and you blanket people with promotion, you’re going to rack up downloads. That doesn’t mean users would miss you if you disappeared tomorrow.
Imagine two different people. One is curious and bored on a commute, sees Muse everywhere, downloads it, asks three questions, then forgets it exists. The other uses an AI tool every day: to write emails, study, code, plan lessons, or just think. Those two downloads look identical in the public scoreboard, but they’re not even the same sport.
The part that worries me is what Meta’s success will push the rest of the industry to copy. If “winning” becomes about installs and chart rankings, we’ll get more AI apps optimized for hooking you fast, not helping you deeply. We’ll get louder marketing, more notifications, more cheap tricks to create a streak or a habit, more pressure to share, more reasons to give it your contacts, your photos, your messages. Meta doesn’t need to do anything shady for this to be a problem. The incentives are already pointed in that direction.
And the stakes aren’t just “annoying apps.” If Muse becomes the default AI for a huge slice of normal people—students, parents, small business owners—then Meta gets to shape what “asking an AI” even means. What it suggests. What it refuses. How it frames sensitive topics. How it nudges your choices. If you think social media feeds influenced how people see the world, wait until the interface is a chat box that talks back like it knows you.
On the flip side, there’s a real argument for this being good. If Muse is easy, fast, and actually useful, then wide distribution is a gift. Most people don’t care about AI drama. They care about whether it helps them write a resume, understand a medical bill, plan a trip, or get through a hard conversation with their boss. If Meta made something that simple and accessible, great. Also, launching on both iOS and Android from the start is what users deserve. It’s annoying when new tools feel gated to one platform, one country, or one type of person.
But I’m not ready to clap because a giant company proved it can buy attention.
Here’s another scenario. Say you’re a teacher and half your class starts using Muse because it’s the one everyone downloaded. Now the default AI behavior in that classroom matters. Does it encourage real thinking or lazy answers? Does it cite sources clearly or speak with fake confidence? Does it protect kids by default or push them to share too much? None of that shows up in a download number. But it shows up in outcomes.
Or say you run a small shop. You start using Muse for product descriptions and customer replies. If it’s good, you save time and make more money. If it’s wrong, you look careless. If it “helpfully” nudges you toward certain tools or services, you might not even notice you’re being steered. Again, not a download story—a power story.
So sure, 5 million in 22 days is a loud headline. I just don’t think it settles the question people want it to settle. It doesn’t prove Muse is better than anything. It proves Meta can still move a crowd.
What I genuinely don’t know yet is whether people are downloading Muse because it’s the most useful AI agent they’ve tried, or because Meta is simply the best in the world at getting you to tap “Install”—and that difference is going to matter a lot when the hype wears off.
If Muse keeps growing, do we want the future of everyday AI to be won by the best product, or by the biggest distribution machine?