The Question Hidden Inside a Geofence
This series examines the constitutional treatment of geofence warrants through Chatrie v. United States and its relationship to the Fifth Circuit’s United States v. Smith. The central question is not simply whether police may draw a circle around a crime scene. It is what happens when that circle is used to compel a technology company to search for everyone who might have been there, examine some people’s movements, and identify a few of them. The practical value of this opening is a clear starting model: a geofence warrant works backward from a known place and time, and each step can change what the government is asking to see.
A crime scene, but no suspect
On May 20, 2019, someone robbed a credit union in Midlothian, Virginia. The robber carried a gun, forced the bank’s manager to open a safe, and escaped with about $195,000. Witnesses and security footage gave investigators useful facts, but not a name. The footage suggested that the robber had approached from a nearby church while appearing to use a cell phone. That detail became the bridge to a new investigative technique.
Instead of beginning with a known suspect and asking a company for that person’s records, police applied for a warrant directed to Google. The application asked for information about cell phones within a 150-meter radius of the credit union during a one-hour period around the robbery. The aim was to discover who had been near the scene and might have been the robber, an accomplice, or a witness. The magistrate issued the warrant. After Google and the police followed the process described in it, three users’ identifying information was produced. One was Okello Chatrie, whose location record showed that he entered the area about ten minutes before the robbery and then headed toward a residential part of town.
That sequence sounds narrow if it is described only as “a search near a bank for one hour.” But the phrase hides the unusual direction of the inquiry. The police did not know which account to name because they did not yet know whom to suspect. They named a place and a time, then asked a provider to look for the people who fit those coordinates. The first question was therefore not, “Does this known person’s record place them at the crime scene?” It was, “Which people should become candidates because a database says their phones were there?”
That distinction matters even when the underlying crime is serious. A person standing near a bank may be a customer, an employee, a passerby, a neighbor, or a witness. A location hit can be useful without explaining why that particular person should be investigated. The technology makes the first screen cheap and broad enough that the government can examine a population before it has individualized suspicion about any member of that population. The constitutional issue begins at that point, before the final list has been reduced to a few names.
A warrant that works backward
An ordinary search warrant usually has a forward-looking shape. Investigators develop facts about a person, place, or thing; a magistrate evaluates the application; and the warrant authorizes a defined search for evidence. A geofence warrant begins with a different fact pattern. The scene is known, but the person is not. The warrant turns a physical perimeter and a time window into a request for digital records held by a provider.
The historical Google process described in the cases made that reversal explicit. At the first stage, Google supplied a list of devices whose Location History indicated that they had been within the geofence during the specified period. The information was initially pseudonymized or anonymized: device identifiers appeared without names. That did not make the data meaningless. It still showed coordinates, timestamps, and the estimated accuracy of each point. Police could compare those movements with what they knew about the crime and decide which devices seemed worth examining further.
At the second stage, the police could ask for more location information about a smaller group. The additional record was not confined to the original circle. It could show movements before and after the crime, including travel outside the geofence. That broader context could help investigators eliminate a device that merely passed through the area or focus on one whose path appeared consistent with the crime. It could also reveal where a person had gone, or come from, beyond the place that initially justified the request.
At the third stage, police could seek identifying information for a still smaller group, including names, email addresses, and phone numbers. The logic was incremental: begin with a group, narrow the group, then attach identities. But “incremental” does not mean constitutionally insignificant. A later request can be more revealing than an earlier one, and a later choice can be made by officers using information that was not available when the original application was presented to the magistrate.
The Virginia warrant illustrates the progression. Google first returned data for 19 users. Officers reduced the list to nine and obtained broader movement information. They then reduced it to three and received account-identifying information. The numbers describe this case, not a standard limit for geofence warrants. Their importance is structural. They show how a request that begins without a suspect can produce a small set of identified people through successive decisions about whom to scrutinize more closely.
That structure also separates the roles of the participants. Google maintained the historical Location History repository and performed the technical work needed to find responsive records. Police defined the crime-related location and time, reviewed the returned material, and selected users for more intrusive requests. A magistrate approved the warrant at the outset. The later opinions disagree about how much constitutional weight to give Google’s internal search, the pseudonymized first result, and the officers’ later narrowing decisions. But they agree on the practical fact that the data changed character as the process advanced.
The process could be valuable in a case with no other lead. It could also include people who had done nothing wrong. The Fifth Circuit’s Smith opinion records concerns about false positives and innocent bystanders swept into investigations because their devices were near a crime scene. The same opinion describes delays in processing some requests and notes that not every geofence leads to an arrest. These limits do not prove that the technique is unreliable in every case. They do show why a location hit should be treated as an investigative lead, not as a completed explanation of a person’s conduct.
Three questions, not one
The Supreme Court’s 2026 decision in Chatrie supplies the first answer, but not the whole answer. The Court held that police conducted a Fourth Amendment search when they acquired Chatrie’s Location History from Google. In the Court’s view, people have a reasonable expectation of privacy in records about their cell phones’ locations, even when the records are held by a technology company and the period examined is only two hours.
That holding answers a threshold question: whether the government’s access to the information falls within the Fourth Amendment’s concept of a search. It does not automatically establish that the search was unreasonable. The Fourth Amendment prohibits unreasonable searches and seizures, so the next inquiry is whether the warrant was supported by probable cause and described the authorized search with sufficient particularity. The Supreme Court specifically left the Court of Appeals to decide whether those requirements were satisfied at each stage of the unusual, multi-step warrant.
There is also a third question: what happens to the evidence if a constitutional violation occurred? The exclusionary rule and its good-faith exception concern the remedy. The trial court in Chatrie concluded that the warrant plainly violated Fourth Amendment rights but admitted the evidence because the officers’ reliance was objectively reasonable. The Fifth Circuit reached a similar combination in Smith: it called the geofence warrants unconstitutional general warrants, yet affirmed because the investigators had acted in good faith while using a novel technique. Other Fourth Circuit judges argued that the defects were too fundamental for good faith to save the evidence.
Those three questions—search, reasonableness, and remedy—must stay distinct. If they are collapsed, the fact that a magistrate signed a document may be treated as proof that no search occurred, or the fact that evidence was admitted may be treated as proof that the warrant was valid. Neither inference follows. A warrant can authorize a search and still be challenged for its scope or factual support. A court can find a constitutional defect and still decide that suppression would not provide enough deterrence to justify its costs.
The cases also establish the series’ central perspective. A geofence is not best understood as a single circle on a map. It is a chain of decisions that moves from an unknown population to device-level location records, then to broader movement histories, and finally to identified people. The constitutional analysis must follow that chain. The Supreme Court has resolved the threshold question for historical Location History: obtaining it is a search. It has not yet resolved whether every stage of the Chatrie warrant was reasonable, whether the Fifth Circuit’s categorical general-warrant analysis is correct, or whether the evidence should ultimately be suppressed.
That is why the important question hidden inside a geofence is not simply where the boundary was drawn. It is who was brought inside the investigation, what information became available at each step, and who decided that the next intrusion was justified. The next stage of the inquiry begins with the record itself: how Location History was generated, what its repository made possible, and how a place-time request became a movement and identity search.