On August 6, 2024, thousands of Southern Californians received an alert on their phones warning of an impending earthquake. They had up to 30 seconds to take cover before the magnitude 5.2 tremor struck near Bakersfield. The warning didn't come from expensive seismometers buried in the ground. It came from the phones themselves.
The Accidental Seismometer in Your Pocket
Every smartphone contains an accelerometer—a tiny sensor that knows when you've tilted your device. Originally designed for mundane tasks like rotating your screen or counting your steps, these chips turned out to have an unexpected talent: detecting earthquakes.
The insight seems obvious in hindsight, but it took years to realize. If an accelerometer can sense when you flip your phone from portrait to landscape, it can certainly sense the ground shaking beneath your feet. The challenge wasn't sensitivity. It was scale and intelligence.
A single phone can't distinguish between an earthquake and someone dropping their device on a table. But millions of phones, all detecting similar motion patterns at the same time? That's a different story.
Racing Light Against Earth
In 2016, UC Berkeley researchers Qingkai Kong, Richard Allen, and Louis Schreier launched MyShake, the first app designed to turn smartphones into a seismic network. Using a patented neural network, the app learned to recognize the specific motion patterns of earthquakes versus everyday phone jostling.
The concept worked better than expected. MyShake has since recorded over 1,100 earthquakes and been downloaded 1.6 million times across more than 80 countries.
But the real transformation came four years later when Google built earthquake detection directly into Android's operating system. No app download required—every Android phone became a potential sensor by default.
"We're essentially racing the speed of light against the speed of an earthquake," explained Marc Stogaitis, a software engineer at Android. Radio signals travel faster than seismic waves, meaning alerts can reach people before the shaking does, at least for those far enough from the epicenter.
The math is straightforward. Earthquakes generate two types of waves: Primary (P) waves that travel quickly but cause little damage, and Secondary (S) waves that arrive later and do the real harm. Traditional early warning systems detect the P waves and send radio alerts that outpace the S waves. Google's system does the same thing, just with millions more sensors.
From California to the World
When Google launched its Android Earthquake Alerts System in California on August 11, 2020, it piggybacked on existing infrastructure. The state already had ShakeAlert, a network of 700 traditional seismometers maintained by the USGS, Caltech, and UC Berkeley. The phones simply delivered those warnings to more people.
The real test came in April 2021, when Google expanded to Greece and New Zealand—countries with far less seismic infrastructure than California. Here, the phones weren't just delivery devices. They were the detection system.
By 2024, the network operated in more than 90 countries, reaching approximately 3 billion Android phones. A 2025 study in Science journal found the system was detecting an average of 312 earthquakes per month, ranging from magnitude 1.9 to 7.8. It delivered roughly 60 alerts and 18 million individual notifications monthly.
The scale dwarfs anything possible with traditional equipment. Building a network of 700 seismometers in California cost hundreds of millions of dollars and took decades. Google activated 3 billion sensors in regions that had none, essentially overnight.
The Warning Time Problem
The system works, but not perfectly. User feedback reveals that only 36% of people who received alerts got them before the shaking started. Another 28% received alerts during the shaking, and 23% after it ended.
Those numbers sound disappointing until you consider the physics. Warning time depends on distance from the epicenter. If you're standing directly above where an earthquake starts, no system can alert you faster than the waves themselves. But if you're 50 miles away, you might get 30 seconds—enough time to drop, cover, and hold on.
The October 25, 2022, magnitude 5.1 earthquake in California's Bay Area demonstrated this clearly. Many residents received alerts before feeling anything. Others felt the shaking first. Both experiences happened in the same earthquake, just at different distances.
Even a few seconds matter more than you'd think. That's time to step away from windows, shelter under a desk, or pull a car to the side of the road. It's time to prevent trains from entering tunnels or planes from landing. The 2011 Tohoku earthquake in Japan showed what's possible: automated systems stopped 27 bullet trains before the shaking reached them.
When Billions Beat Precision
The traditional view holds that earthquake detection requires expensive, precisely calibrated instruments installed in carefully chosen locations. Smartphone accelerometers are neither expensive nor precisely calibrated, and they sit in pockets and purses, not bedrock.
Yet the crowdsourced approach offers something traditional networks can't: ubiquity. California experiences up to 100 small earthquakes daily, but only about 15-20 per year exceed magnitude 4.0. Traditional seismometers excel at detecting and analyzing these events in detail. Smartphones excel at being everywhere at once.
In regions that can afford dense seismometer networks, phones supplement existing systems. In regions that can't—which is most of the world—phones provide the only early warning available.
The trade-off isn't precision versus coverage. It's having data versus having none.
The Network That Built Itself
Perhaps the strangest aspect of this story is how little deliberate infrastructure it required. No one needed to install sensors, run cables, or maintain equipment. The network assembled itself through the ordinary process of people buying phones.
MyShake still operates as a standalone app for users who want to contribute data actively. But Google's system works passively, using sensors that were already there for other purposes. The earthquake detection is almost an afterthought, a bonus feature extracted from hardware designed to do something else entirely.
With 16 billion mobile phones worldwide—more than 3 billion running Android—the network will only grow denser. Each new phone sale adds another sensor. Each software update can improve detection algorithms across billions of devices simultaneously.
Traditional seismic networks took a century to build and still leave most of the planet uncovered. The smartphone network covered most of the planet in four years. It's not a better system. It's a different one, built on abundance rather than precision, and it arrived without anyone quite planning for it.