Google’s Android Earthquake Alerts system failed to issue critical warnings ahead of the 2023 Turkey earthquake that killed over 55,000 people, sparking global concern over its reliability. Experts now question whether tech-based systems are trustworthy enough to replace national alert infrastructures.
Google has acknowledged that its Android Earthquake Alerts (AEA) system failed to deliver crucial warnings during the catastrophic 2023 earthquake in Turkey, leaving millions without timely alerts that could have saved lives.
The powerful 7.8-magnitude quake struck southeastern Turkey in the early hours of February 6, killing more than 55,000 people and injuring over 100,000 as buildings collapsed while most were still asleep. Despite the earthquake’s severity, only 469 “Take Action” alerts the system’s highest-level warnings were issued by Google. This stands in stark contrast to the estimated 10 million people who were within 98 miles of the epicenter and could have benefited from immediate warning.
Google later told the BBC that its early warning system sent out standard alerts to about 500,000 Android users in Turkey. However, those notifications are designed for minor tremors and do not trigger loud emergency alarms capable of waking users or prompting urgent evacuation.
The revelation has sparked outrage, especially as the tech giant had previously claimed that the AEA system performed well in previous trials. Google’s system, intended to serve as a “global safety net” in areas lacking national alert infrastructures, runs on Android devices present in over 70% of smartphones in Turkey. The system uses accelerometers in these phones to detect seismic activity and instantly send alerts based on the severity of ground movement.
On the day of the disaster, the AEA system did activate but vastly underestimated the magnitude of the quake. According to Google researchers, the initial reading pegged the event at between 4.5 and 4.9 on the Richter scale far below the actual 7.8 magnitude. As a result, the system only issued lower-tier alerts, which were insufficient to wake residents or prompt immediate safety measures.
A similar error occurred later that same day when a second major earthquake struck. This time, the system managed to send out 8,158 “Take Action” alerts and fewer than 4 million “Be Aware” alerts, which are less urgent and do not override Do Not Disturb settings.
Following the incident, Google researchers conducted a post-event analysis published in the journal Science. The updated algorithm was able to retroactively send 10 million high-level warnings and over 67 million “Be Aware” notifications to affected zones. However, the delayed insight did little to comfort those impacted.
“Every earthquake early warning system faces the same problem: adjusting the algorithms to the times when large earthquakes occur,” a Google spokesperson said, defending the platform’s long-term goals. They emphasized that the AEA system is constantly evolving and is currently active in 98 countries.
Still, the delay in acknowledging the system’s failure has alarmed scientists and researchers. Elizabeth Reddy, an assistant professor at the Colorado School of Mines, criticized the two-year lag between the event and the public disclosure of what went wrong. “I’m really disappointed that it took so long,” she said.
The core issue lies in the system’s detection algorithm, which struggled to gauge the true scale of the quake. Google has since revised this algorithm, aiming to improve accuracy and response time in future disasters. However, scientists remain concerned about relying too heavily on tech-based systems that may not yet be fully battle-tested.
According to Harold Tobin, director of the Pacific Northwest Seismic Network, there is a growing misconception in some countries that Google’s system is a sufficient alternative to national warning infrastructures. “I think it’s absolutely critical to be very transparent about how well it works,” Tobin told the BBC.
The concern is that developing nations may increasingly depend on AEA due to its widespread availability, assuming it’s a foolproof solution. But the Turkey earthquake showed just how wrong those assumptions can be.
The BBC has since contacted Google to inquire about AEA’s performance during the 2025 Myanmar earthquake. As of now, no response has been received.
Google continues to stress that its system is meant to complement, not replace, government-run earthquake alerts. Yet as the tech world plays a bigger role in disaster response, the Turkey earthquake may serve as a sobering lesson in the risks of overconfidence in artificial intelligence during life-or-death emergencies.
