Google sent the Taumarunui earthquake alert, not GeoNet or Civil Defence

A warning that came from Silicon Valley, not Civil Defence

When a 5.9 magnitude earthquake struck 5km south of Taumarunui at 4:51am on Saturday, thousands of Android users across the North Island got a jolt before the ground moved. Some Auckland residents reported the alert arriving seconds ahead of the shaking, with one calling it “crazy accurate.”

The striking part is who sent it. Not GeoNet. Not the National Emergency Management Agency. Not Civil Defence. The warning came from Google, running its Android Earthquake Alerts system as a background process on ordinary phones. The technology uses the accelerometers already built into Android handsets to detect the fast-moving P-waves that precede the destructive S-waves. When enough phones register matching signals, Google pushes an alert to everyone in the expected shaking zone.

For business owners and risk managers, this is the story: New Zealand’s most visible earthquake early warning capability is a free consumer feature owned by a foreign company, and it is filling a gap the government never plugged.

New Zealand never built its own

There is no official national earthquake early warning system in this country. GNS Science confirmed in 2023 that “while Aotearoa New Zealand experiences frequent damaging earthquakes, there is no official national earthquake early warning system,” and none has been announced since. New Zealand and Greece were chosen as the first two countries to trial Google’s system in 2021, precisely because neither had an official warning system in place.

The official Emergency Mobile Alert system, the one that blasts messages through cell towers, is structurally useless for quakes. Civil Defence Manawatu-Whanganui put it bluntly: “By the time we would have logged into the system, written a message and decided where to send the alert, the quake would have been and gone.” GeoNet, which recorded 23,080 earthquakes in 2024-2025, is a monitoring and reporting tool, not a warning service. Its app has over 264,000 installs but tells you what happened, not what is about to.

Google’s system is not supplementing an official warning network. It is substituting for one that does not exist.

Why you cannot build procedures on it

The Taumarunui event looked flawless on social media. The limitations underneath it are not.

Warning time depends entirely on distance from the epicentre. The Spinoff’s explainer noted that users far from the source might get 10 seconds or more, while those a few kilometres away get one or two, and anyone directly above the epicentre gets nothing useful at all. For businesses in Wellington, Hawke’s Bay or the Marlborough fault zones, the warning that impressed Auckland may simply not arrive.

The system also works only on Android phones with alerts enabled. iPhone users get nothing. Any workplace with a mixed device environment, and that is most of them, cannot assume all staff are covered. Civil Defence Manawatu-Whanganui also warned it “does not have the accuracy of the equipment GeoNet uses,” performs poorly in sparsely populated rural areas where phone density is low, and cannot detect offshore quakes until waves reach land.

Then there is the human problem. Massey University research published in 2025, drawing on a survey of 3,000 New Zealanders, found most people who received the alerts did not perform the correct drop, cover and hold response. The warning is only half the equation, and the trained half is largely missing.

Nobody knows who to hold accountable

The same Massey research exposed an accountability gap. Only a quarter to a third of respondents correctly identified Google as the source. Roughly as many thought it was NEMA, and more credited GNS Science or GeoNet. Most found it useful. Most had no idea who ran it.

That matters commercially. The system carries no service-level agreement with New Zealand businesses, no regulatory accountability, and no obligation on Google to maintain it. GNS Science noted it does not even use GeoNet data. Google can change the algorithm, shift the thresholds, or pull the service entirely, and no New Zealand authority could stop it.

What businesses should actually do

Do not write emergency procedures that begin “when staff receive the phone alert.” That trigger excludes iPhone users, fails near epicentres, and most recipients do not act on it correctly. Drop, cover and hold drills remain the non-negotiable baseline, because the data shows people know the right action but do not take it automatically. It is worth remembering that NEMA’s 2024 preparedness survey found 53% of New Zealanders are not prepared for an emergency at all.

For a country that logs more than 23,000 earthquakes a year, leaning workplace safety on a foreign company’s free feature is a fragile foundation. The Taumarunui alert was a genuinely impressive moment of consumer technology. It was also a reminder that resilience infrastructure this important should not be something a business, or a country, gets by accident.

Sources

• 1News: Why phones woke thousands before 5.9 magnitude Taumarunui earthquake (2026-07-24)
• GeoNet Annual Report 2024-2025 (2025)
• Massey University: Research shows New Zealanders know very little about early earthquake warning systems (2025-03-25)
• NEMA: Understanding emergency preparedness in New Zealand – 2024 Annual Preparedness Survey (2024)
• GNS Science: Earthquake Early Warning in New Zealand (2023-08-28)
• The Spinoff: Your mobile phone could soon warn you of earthquakes (2021-05-05)