Geofence Warrants – Part 1: The Sensorvault Architecture
Google built a database that tracks the movements of hundreds of millions of people. Law enforcement figured out how to search it in reverse—identifying suspects not by who they are, but by where their phones happened to be.
—
In 2020, law enforcement agencies served Google with more than 11,554 warrants directing the company to conduct sweeping searches of its location database.[1] Each warrant instructed Google to scan location records from hundreds of millions of devices and return a list of everyone present within a defined geographic area during a specified time window. The people identified weren’t suspects yet. They became suspects because their phones were there.
This investigative tool—called a geofence warrant—represents a fundamental inversion of traditional Fourth Amendment process. Traditional warrants identify a suspect first, then seek evidence about that person. Geofence warrants search everyone first, then generate suspects from the data. The result is what the Fifth Circuit would later call “near perfect surveillance at society level”—a description that, depending on which circuit you’re in, is either a constitutional problem or perfectly acceptable.[2]
On April 27, 2026, the Supreme Court will hear arguments in Chatrie v. United States (Docket 25-112), the first case to place the constitutionality of geofence warrants before the nation’s highest court.[3] The decision, expected by early summer, will determine whether police can continue sweeping up location data on thousands of innocent people per warrant—or whether the Fourth Amendment’s particularity requirement means what it has always meant: that the government must have reason to suspect you before it searches your data.
This six-part series examines the constitutional architecture of geofence warrants: how the technology works, how courts have split on whether it’s constitutional, what the Supreme Court might do, and why the answer matters far beyond a single investigative technique. The place to start is the machine itself—the database Google built, the tracking it enabled, and the three-step process that turned location history into a dragnet.
—
The Sensorvault
Google’s location tracking system operates at a scale that would have been inconceivable to the Framers of the Fourth Amendment, and that challenges the assumptions underlying every Fourth Amendment precedent the Supreme Court has decided since.
The system’s centerpiece is Sensorvault, a central repository that stores detailed location histories from devices attached to an estimated 592 million individual Google accounts.[4] Sensorvault drew from multiple Google services: Google Maps navigation data, Android device GPS coordinates, Wi-Fi and cell tower triangulation, Google Photos embedded location metadata, and the user-enabled Location History feature.[5] An internal Google presentation described the database as logging location data on average every two minutes from Android devices and Google apps on non-Android phones.[6]
The data was nominally voluntary. Users had to “opt in” to Google’s Location History feature. But the voluntariness of that choice is precisely what two federal circuits disagree about. An internal Google email—disclosed in litigation and cited by the Fifth Circuit—explained that the opt-out interface “feels like it is designed to make it possible” but is “difficult enough that people won’t figure it out.”[7] And in an era when a smartphone with location capabilities is, as Chief Justice Roberts wrote in Riley v. California, “such a pervasive and insistent part of daily life” that it is practically “a feature of human anatomy,”[8] opting out of location tracking is less a choice than a concession.
Google Maps creator Brian McClendon later called geofence warrants “fishing expeditions,” explaining that Google employees had originally assumed law enforcement would only seek Location History data on specific people—not on everyone near a crime scene. That assumption, as the data would show, proved wildly optimistic.[9]
—
How Geofence Warrants Work: The Three-Step Process
The mechanics of a geofence warrant follow a three-step process that Google developed in response to law enforcement demand. Each step narrows the search, but the first step—the one that implicates the Fourth Amendment most directly—searches everyone.[10]
Step 1: Identify devices. Law enforcement provides Google with a geographic area (the “geofence”) and a time window. Google searches Sensorvault for all devices that entered that area during that window and returns a de-identified list—device IDs without names, ranked by how closely their movements match the behavioral profile described in the warrant (e.g., “devices that left the area quickly”). This step requires searching the entire database to produce the subset. There is no way to query only “suspects” because the whole premise of a geofence warrant is that the suspect is unknown.[11]
Step 2: Focus on specific devices. From the Step 1 list, law enforcement selects devices of interest—typically those whose movement patterns match the behavior they’re looking for. They return to Google with a narrowed request for more detailed location data on those specific device IDs. Still de-identified; still no names.[12]
Step 3: De-anonymize. For the final set of devices, Google identifies the actual users associated with those accounts—names, email addresses, and whatever other account information is available. At this point, the police have a suspect.[13]
The process is architecturally elegant and constitutionally fraught. Step 1 is the problem. It searches everyone in Sensorvault—every device, every account, every person who happened to be near the geofence—regardless of whether there is probable cause to believe any particular individual committed a crime. The Fifth Circuit, in United States v. Smith, would later describe this as “the quintessential problem with these warrants”: “they never include a specific user to be identified, only a temporal and geographic location where any given user may turn up post-search. That is constitutionally insufficient.”[14]
The Fourth Circuit, sitting en banc in United States v. Chatrie, disagreed. It held that because users “voluntarily” shared location data with Google, accessing that data through a geofence warrant did not constitute a Fourth Amendment search at all—at least not when the time window was limited to two hours.[15] The same technology. The same process. Opposite constitutional conclusions.
—
The Scale: 11,500+ Warrants in a Single Year
Google received its first geofence warrant request in 2016.[16] By 2018, the number had reached 982. In 2019, it jumped to 8,396. And in 2020—four years after the first request—Google received 11,554 geofence warrants in a single year.[17]
To put that number in perspective: in 2020, geofence warrants accounted for approximately 25 percent of all data requests Google received from law enforcement.[18] A single investigative technique, targeting a single company, accounted for one quarter of all government demands for user data.
The warrants came primarily from state and local law enforcement, not federal agencies.[19] They were used to investigate robberies, burglaries, drug crimes, and homicides—the full range of criminal investigations where identifying who was near a crime scene could help identify a suspect.[20] And they were extraordinarily effective at generating suspects, which is precisely the problem.
Because Step 1 searches everyone, every person who carried a phone near the geofence became part of the investigation—regardless of whether they had any connection to the crime. Google’s transparency reports don’t disclose how many innocent people were identified and investigated as a result of geofence warrants. But court records and reporting document several cases where the consequences were real.
—
The Innocent Bystanders
A man riding his bike on a public path that passed near a burglarized home was identified through a geofence warrant, questioned by police, and required legal representation before he was cleared.[21] He had no connection to the crime. His phone was in the wrong place at the wrong time.
During the summer of 2020, law enforcement agencies across the country obtained location data for thousands of individuals involved in largely peaceful protests—using geofence warrants that covered the areas where demonstrations occurred.[22] The Knight First Amendment Institute, the Reporters Committee for Freedom of the Press, and FIRE documented these deployments in their March 2026 amicus brief in Chatrie, warning that “geofence warrants make it impossible to assemble, protest, newsgather, worship, and more away from the government’s watchful eye.”[23]
A restaurant employee working a shift near a robbery was identified because her phone placed her in the geofence during the crime window. She became a suspect despite having no connection to the robbery, and was only cleared when her employer provided her alibi.[24]
These are the documented cases. The undocumented ones—people identified, questioned, and cleared without generating a court record—almost certainly number far higher. As the Congressional Research Service noted in its analysis of reverse warrants, geofence warrants are “exploratory searches” that “invert” the traditional warrant process: rather than targeting a known suspect, they search first and identify suspects later.[25]
The absence of systematic data on innocent bystanders is itself a constitutional concern. When the government searches 592 million records to find one suspect, the ratio of innocent people searched to guilty people identified is enormous. But because there is no reporting requirement, no oversight mechanism, and no public record of how many innocent people are swept up, the constitutional cost of geofence warrants remains invisible. The Fourth Amendment’s particularity requirement was designed to prevent exactly this kind of invisible, unquantified mass search. The question is whether it still can.
—
The Constitutional Fault Line
The architecture of Sensorvault creates a constitutional problem that existing Fourth Amendment doctrine was not designed to solve. The Fourth Amendment’s particularity requirement—the principle that warrants must describe “the place to be searched, and the persons or things to be seized”—presupposes that the government knows who it’s looking for before it starts searching.[26] Geofence warrants operate on the opposite premise.
This inversion is why two federal circuits, applying the same Fourth Amendment framework, reached opposite conclusions about the same technology. The Fourth Circuit focused on the output of a geofence search—a “brief glimpse” at a limited time window—and concluded it was no more intrusive than traditional surveillance.[27] The Fifth Circuit focused on the process—searching a database of 592 million accounts to identify unknown individuals—and concluded it was “the exact sort of ‘general, exploratory rummaging’ that the Fourth Amendment was designed to prevent.”[28]
The Harvard Law Review, in a February 2025 analysis titled “Much Ado About Geofence Warrants,” identified the structural problem: “Current Fourth Amendment doctrine provides no administrable way to mediate between allowing and banning all geofence warrants because probable cause—the only readily available standard to filter warrant applications—can nearly always be shown in the geofence context.”[29] When a database contains 592 million accounts, probable cause that someone in a given area has committed a crime is almost always available. But probable cause that you committed a crime is an entirely different standard—and it’s the one the Fourth Amendment has always required.
Professor Paul Ohm identified this structural problem in 2010, before geofence warrants existed: “In increasingly common situations, whenever police have any suspicion at all about a piece of evidence, they almost always have probable cause.”[30] Geofence warrants are the technological manifestation of Ohm’s warning: they make probable cause too easy to satisfy to function as a meaningful constitutional limit.
The Supreme Court will confront this problem on April 27. But to understand why the circuits split so dramatically, it’s necessary to understand the historical context that makes this question so constitutionally charged—and the specific case that created the split.
—
Writs of Assistance: The Framers’ Original Motivation
The Fourth Amendment did not emerge from abstract principle. It was a direct response to a specific colonial practice: the British use of writs of assistance—general warrants that allowed customs officials to search any place, seize any goods, and arrest any person without particularized suspicion.[31]
The writs were not targeted. They did not name a place or a person. They authorized government agents to search wherever they wanted, whenever they wanted. James Otis’s 1761 argument against the writs—”a man’s house is his castle”—is credited with inspiring the Fourth Amendment’s particularity requirement.[32] The Framers understood, from direct experience, what it meant for government agents to search without knowing what they were looking for or who they were looking for.
Geofence warrants are not writs of assistance in the literal sense—they require judicial approval and probable cause. But the structural parallel is what makes the constitutional question so difficult. A geofence warrant authorizes the government to search everyone in a defined area without knowing who it is looking for, and to generate suspects from the search results. Whether that structural similarity is constitutionally dispositive is the question that has split the circuits and that the Supreme Court must now answer.
To understand the specific case that created the split—and the 126 pages of fractured judicial reasoning that the Fourth Circuit produced in attempting to justify geofence warrants under existing doctrine—we turn to Part 2.
—
What This Series Covers
– Part 2: The Chatrie Case — A 2019 Virginia robbery, a district court that suppressed the geofence evidence, and a Fourth Circuit en banc decision that reversed in a one-sentence per curiam opinion followed by 126 pages of concurring and dissenting opinions – Part 3: The Grand Canyon Split — Why the Fourth and Fifth Circuits reached opposite conclusions, and what their disagreement reveals about the limits of existing Fourth Amendment doctrine – Part 4: The Scale Problem — 11,500+ warrants per year, innocent bystander cases, and the absence of statutory limits on geofence scope – Part 5: Google’s 2023 Policy Shift — Google said it would stop storing location data centrally. Did it work? And does it matter? – Part 6: The Coming Precedent — What the Supreme Court might rule, what each outcome would mean, and why the implications extend far beyond geofence warrants
—
Notes
-
Google Transparency Report, geofence warrant statistics for 2020, reported 11,554 geofence warrants served. See Google, “Supplemental Information on Geofence Warrants in the United States,” https://services.google.com/fh/files/misc/supplemental_information_geofence_warrants_united_states.pdf; see also Jennifer Valentino-DeVries, “Tracking Phones, Google Is a Dragnet for the Police,” *New York Times* (Apr. 13, 2019), https://www.nytimes.com/2019/04/13/technology/google-sensorvault-location-tracking.html. ↩
-
*United States v. Smith*, 110 F.4th 817 (5th Cir. 2024), describing geofence warrants as enabling “near perfect surveillance at society level.” ↩
-
*Chatrie v. United States*, No. 25-112 (U.S., cert. granted Jan. 16, 2026). See SCOTUSblog case file, https://www.scotusblog.com/cases/case-files/chatrie-v-united-states/; see also Oyez docket, https://www.oyez.org/cases/2025/25-112. ↩
-
Sensorvault estimated to contain data from 592 million individual accounts. See *United States v. Smith*, 110 F.4th 817, 825 (5th Cir. 2024); see also Clayton R. Rice, “U.S. Fifth Circuit Rules Geofence Warrants Are Unconstitutional,” https://www.claytonrice.com/u-s-fifth-circuit-rules-geofence-warrants-are-unconstituional/. ↩
-
See Electronic Frontier Foundation, “Google’s Sensorvault Can Tell Police Where You’ve Been” (Apr. 2019), https://www.eff.org/deeplinks/2019/04/googles-sensorvault-can-tell-police-where-youve-been; see also University of California, Berkeley School of Information, “The Sensorvault Is Watching” (Oct. 2019), https://blogs.ischool.berkeley.edu/w231/2019/10/07/the-sensorvault-is-watching/. ↩
-
See *United States v. Chatrie*, 107 F.4th 319, 326 (4th Cir. 2024) (describing average logging interval); see also Clayton R. Rice, “Geofence Warrants and Google’s Sensorvault,” https://www.claytonrice.com/geofence-warrants-and-googles-sensorvault/. ↩
-
Internal Google email cited in *United States v. Smith*, 110 F.4th 817 (5th Cir. 2024); see also Amster & Diehl, “Geofence Warrants and the Fourth Amendment,” 74 *Stan. L. Rev.* 385, 410 (2022), https://review.law.stanford.edu/wp-content/uploads/sites/3/2022/02/Amster-Diehl-74-Stan.-L.-Rev.-385.pdf. ↩
-
*Riley v. California*, 573 U.S. 373, 395 (2014) (“These days, it is the rare adult who does not carry a cell phone—a feature of human anatomy, practically speaking”). ↩
-
Jennifer Valentino-DeVries, “Tracking Phones, Google Is a Dragnet for the Police,” *New York Times* (Apr. 13, 2019), https://www.nytimes.com/2019/04/13/technology/google-sensorvault-location-tracking.html (quoting Brian McClendon describing geofence warrants as “fishing expeditions”). ↩
-
The three-step process is described in Google’s published supplemental information on geofence warrants and confirmed in both the Fourth and Fifth Circuit opinions. See Google, “Supplemental Information on Geofence Warrants in the United States,” https://services.google.com/fh/files/misc/supplemental_information_geofence_warrants_united_states.pdf; see also *United States v. Chatrie*, 107 F.4th 319, 326-28 (4th Cir. 2024); *United States v. Smith*, 110 F.4th 817, 822-24 (5th Cir. 2024). ↩
-
See *United States v. Smith*, 110 F.4th 817, 831 (5th Cir. 2024) (“law enforcement cannot obtain its requested location data unless Google searches through the entirety of its Sensorvault—all 592 million individual accounts”). ↩
-
See Google, “Supplemental Information on Geofence Warrants in the United States,” describing Step 2 process. ↩
-
Id., describing Step 3 de-anonymization process. ↩
-
*United States v. Smith*, 110 F.4th 817, 831 (5th Cir. 2024). ↩
-
*United States v. Chatrie*, 107 F.4th 319 (4th Cir. 2024) (en banc) (holding that “the government did not conduct a Fourth Amendment search when it obtained two hours’ worth of location information from Google”). ↩
-
*United States v. Chatrie*, 107 F.4th 319, 323 (4th Cir. 2024) (noting Google received its first geofence warrant request in 2016). ↩
-
Google Transparency Report data: 982 warrants in 2018, 8,396 in 2019, 11,554 in 2020. See Google, “Supplemental Information on Geofence Warrants in the United States,” https://services.google.com/fh/files/misc/supplemental_information_geofence_warrants_united_states.pdf; see also *Geofence Warrant*, Wikipedia, https://en.wikipedia.org/wiki/Geofence_warrant (citing Google transparency data). ↩
-
Google 2021 transparency report showed geofence warrants constituted approximately 25% of all law enforcement data requests. See *Geofence Warrant*, Wikipedia, https://en.wikipedia.org/wiki/Geofence_warrant; see also CRS Report LSB11274, “Geofence and Keyword Searches: Reverse Warrants and the Fourth Amendment,” https://www.congress.gov/crs-product/LSB11274. ↩
-
See *United States v. Chatrie*, 107 F.4th 319, 325 (4th Cir. 2024) (noting majority of geofence warrants came from state and local law enforcement). ↩
-
Id.; see also CRS Report LSB11274, “Geofence and Keyword Searches: Reverse Warrants and the Fourth Amendment.” ↩
-
See *United States v. Chatrie*, Fourth Circuit en banc opinion (documenting innocent bystander cases from the record); see also NBC News, “Google Tracked His Bike Ride Past a Burglarized Home. That Made Him a Suspect” (Dec. 2019), https://www.nbcnews.com/news/us-news/google-tracked-his-bike-ride-past-burglarized-home-made-him-n1151761. ↩
-
Knight First Amendment Institute, Reporters Committee for Freedom of the Press, and FIRE, Amicus Brief in *Chatrie v. United States* (Mar. 2, 2026), https://knightcolumbia.org/documents/avogw49zgx; see also Press Statement, “Knight Institute, Reporters Committee, and FIRE Urge Supreme Court to Reject Dragnet Geofence Warrants” (Mar. 2, 2026), https://knightcolumbia.org/content/knight-institute-reporters-committee-and-fire-urge-supreme-court-to-reject-dragnet-geofence-warrants. ↩
-
Id. (quoting Will Creeley, FIRE legal director: “Geofence warrants make it impossible to assemble, protest, newsgather, worship, and more away from the government’s watchful eye.”) ↩
-
See *United States v. Chatrie*, Fourth Circuit en banc opinion (documenting innocent bystander cases); see also CRS Report LSB11274. ↩
-
Congressional Research Service, “Geofence and Keyword Searches: Reverse Warrants and the Fourth Amendment” (LSB11274), https://www.congress.gov/crs-product/LSB11274. ↩
-
U.S. Const. amend. IV (“no Warrants shall issue, but upon probable cause, supported by Oath or affirmation, and particularly describing the place to be searched, and the persons or things to be seized”). ↩
-
*United States v. Chatrie*, 107 F.4th 319 (4th Cir. 2024) (characterizing geofence output as a “brief glimpse” into whereabouts unlikely to “offer insight into habits, routines, and associations”). ↩
-
*United States v. Smith*, 110 F.4th 817, 831 (5th Cir. 2024) (quoting *Coolidge v. New Hampshire*, 403 U.S. 443, 467 (1971)). ↩
-
“Much Ado About Geofence Warrants,” *Harvard Law Review* (Feb. 2025), https://harvardlawreview.org/blog/2025/02/much-ado-about-geofence-warrants/. ↩
-
Paul Ohm, “The Probable Cause Mismatch,” 94 *Minn. L. Rev.* 521, 523 (2010), https://www.minnesotalawreview.org/wp-content/uploads/2012/03/Ohm_MLR.pdf. ↩
-
See generally William J. Cuddihy, “The Fourth Amendment: Origins and Original Meaning 602-1791” (1990) (comprehensive historical account of writs of assistance and their influence on the Fourth Amendment); see also *Payton v. New York*, 445 U.S. 573, 583 n.21 (1980) (“The Fourth Amendment was in large part a reaction to the general warrants and writs of assistance used by the British Crown.”). ↩
-
See *Entick v. Carrington*, 19 Howell’s State Trials 1029, 95 Eng. Rep. 807 (C.P. 1765) (Camden, L.C.J.) (“The great end, for which men entered into society, was to secure their property.”); see also James Otis, “Argument Against Writs of Assistance” (Feb. 1761) (“A man’s house is his castle; and whilst he is quiet, he is as well guarded as a prince in his castle.”). ↩