Sensors distributed across a city or district continuously listen for sounds matching the acoustic signature of gunfire. When a likely gunshot is detected, the system estimates its location by comparing the timing of the sound's arrival at multiple sensors, and sends an alert to police, typically faster and more precisely located than relying on a member of the public calling it in. At least three sensors must detect a gunshot-like sound; the data and a short audio segment are then reviewed by trained human analysts before an alert reaches police.

Marketed by SoundThinking under the ShotSpotter name, it is one of the most widely deployed AI-enabled policing technologies in the United States, and remains commercially healthy, with multiple multi-year renewals announced during 2026 including Albuquerque, Worcester, Richland County, Macon-Bibb, Peoria and Massachusetts State Police. NYPD continues to use it across all five boroughs.

The efficacy evidence is genuinely contested rather than settled in either direction, and this page states it as such. A New York City Comptroller audit found only 13 percent of alerts resulted in a confirmed shooting and criticised the system's claimed response-time benefits.

The list of cities ending use is now substantial: Dayton did not renew, citing cost and no measurable violence reduction; Canton replaced it; Durham, Charlotte and Winston-Salem cancelled over cost and alerts producing no evidence of gunfire; Chicago cancelled and then sought proposals for a replacement gun-violence detection system; Little Rock voted to end its contract and then reinstated it. Cambridge, Massachusetts voted in May 2026 to end use and remove devices within 90 days, after the system failed to detect a shooting on Memorial Drive that left two people critically injured.