Can Google Location History Prove Your Residency?

The instinct is understandable. Your phone has been in your pocket all year, Google Maps has been quietly noting where it went, and it feels like the argument should already be won. If the state wants to know whether you spent 180 days in New York or 190, the answer is sitting in an app. Just print it.

The problem is that the thing people picture, a Google-held ledger of your movements that an auditor could go verify, no longer exists in that form. On December 12, 2023, Google announced that Location History would stop living on its servers. In Google’s wording, "your visits and routes are automatically saved to a map on each of your devices." Backups became optional and encrypted, and the post is explicit about who is locked out: Google said it would "automatically encrypt your backed-up data so no one can read it," including Google. The same announcement cut the default auto-delete period to three months, down from eighteen.

The rollout was gradual across Android and iOS through the following year, with users notified individually as it reached their accounts. It did not go quietly. By December 2024 The Register was reporting that the migration was still ongoing and that complaints about unexpectedly lost location history were piling up on Reddit and Google’s own support forums. One affected user quoted in that coverage put it plainly: "I would have been absolutely destroyed if I had accidentally selected the default option." The default option kept the most recent three months and erased the rest.

Two more facts matter before you plan a defense around this data. Timeline is off by default for a Google Account and only runs if you opted in, so a meaningful share of people have nothing at all. And the current auto-delete control offers three choices, deleting Timeline data older than 3 months, 18 months, or 36 months, plus the option to keep it until you delete it yourself. A residency audit routinely reaches back three or four years. A 36-month setting does not cover the year most likely to be examined by the time the notice arrives.

The migration was not framed as a filing-cabinet reorganization. It was designed to end law enforcement dragnets, and it worked. Writing in Forbes in October 2024, Lars Daniel described the effect precisely: moving user data from the Sensorvault to the device "would make it impossible for Google to search and return the user’s location data to law enforcement." Once the data sits on the handset, investigators "must identify the person and device from which they seek to obtain location data," which means either the owner’s cooperation or physical possession of the phone.

Google has since said so to the Supreme Court. Criminal Legal News reported on May 1, 2026 that in a March 2026 amicus brief in Chatrie v. United States, Google told the Court it "can no longer respond to geofence warrants" based on Location History, and disclosed that it had objected to over 3,000 geofence warrants on constitutional grounds, with law enforcement withdrawing more than 2,500 of them rather than litigate.

That is a genuine privacy win and a real complication for anyone who wanted to use this data as proof. A record the opposing party can independently obtain and check carries a different weight than a file only you can hand over. Telephone company records, EZ Pass logs, credit card statements, and building access swipes all arrive from someone with no stake in your tax year. Your Timeline export arrives from you, in a format you had the technical ability to edit, describing a device you control.

None of that makes it useless. It changes which category it enters in, and the categories are not a matter of opinion. They are written down.

New York publishes the manual it writes for its own auditors. Chapter VIII of the December 2021 Nonresident Audit Guidelines lists the personal records "typically requested during the course of a residency audit," and the list is worth reading as a list of things the state believes it can rely on: personal diaries and calendars, in written or electronic form; credit card statements and receipts; bank records including monthly statements, canceled checks and ATM receipts; telephone records for both the New York and non-New York residences; utility bills for both residences that may reflect regular or seasonal use; homeowner’s insurance policies to show the location where valuable items are kept; itineraries for commercial flights or flight logs for private carriers; hotel receipts; EZ Pass records for automobile usage; moving bills; and security or swipe cards providing access to office buildings.

Read it again and notice what is not there. No smartphone location app. No GPS export. No Google, by name or category. In a 107-page manual updated for December 2021, the word does not appear. The only slot a Timeline export fits into is the first one, "personal diaries and calendars, in written or electronic form," which is the taxpayer’s own record category. Everything the state considers independent on that list comes from a bank, a carrier, an airline, a toll authority, a hotel, or a building.

The carrier records are the ones auditors go get themselves. Writing in The CPA Journal in November 2016, Corey L. Rosenthal and Fred R. Komarow described the practice: the Department of Taxation and Finance routinely obtains cellphone records directly from telecommunications carriers, and those records "will generally include details for every call made for each cellphone number on an account, including connection date, time, duration of the call, and most importantly, the location of the cellular antennae used to complete the transmission." The state does not need your phone. It needs your carrier, and it has a subpoena.

The pattern holds in the other states this post touches. California’s Franchise Tax Board opens with an Information Document Request and can subpoena cell phone records, credit card statements, and travel records to reconstruct physical presence. Illinois auditors, per ResidencyIQ’s own dossier research, reconstruct day counts using credit card statements, toll records, airline records, and utility usage, and cross-check the General Homestead Exemption against a claimed out-of-state domicile. In every case the reconstruction is built from third parties first.

The deeper issue is what Timeline actually records. It is not a log of coordinates. It is Google’s interpretation of a sampled signal, and the file admits as much in its own fields.

In the Semantic Location History format that Takeout produced for years, a place visit is not asserted, it is scored. Each visit carries a locationConfidence value, a separate visitConfidence value, and a placeConfidence string drawn from LOW_CONFIDENCE, MEDIUM_CONFIDENCE, HIGH_CONFIDENCE, or USER_CONFIRMED. Movement between places is stored as an activity segment with its own confidence field valued LOW, MEDIUM, HIGH, or UNKNOWN_CONFIDENCE, plus an array of candidate activities each carrying a probability from 0 to 100, where the probabilities sum to no more than 100. The community reference that documents this format notes that Google has never explained how the three confidence fields differ from one another.

Sit with USER_CONFIRMED for a moment. That value exists because Google expected to be wrong often enough that it built a mechanism to ask you. A file that distinguishes between places it inferred and places you corrected is telling you, in its own schema, that the rest are inferences.

The on-device era changed the container without changing the character. Current exports are organized around semanticSegments, alongside rawSignals and a userLocationProfile, and the export path now runs through the Maps app on the device rather than a web download. The structure is different. The epistemology is the same. Timeline says a device probably visited a place. It does not say a person stood on a specific side of a state line at 11:58 p.m.

Practitioners who fight day counts for a living have been documenting how location data goes wrong for years, and the failure modes they catalog apply to any location record, phone-derived or app-derived.

On June 29, 2020, Timothy P. Noonan and Andrew W. Wright of Hodgson Russ described cell records as "the go-to third-party source of day count proof" and then spent the piece explaining how to distrust them. Their central example is data trailing: "Data trailing happens when an application on your cell phone continues to distribute a GPS address without confirming the cell phone’s location with the network." It shows up as suspiciously regular repeating entries, 12:05pm, 1:05pm, 2:05pm, with the bytes-up and bytes-down columns both sitting at zero, meaning nothing was actually transmitted. The phone was reporting a stale position, not a current one. They also flag microcells, noting that "a regular cell site will almost never be someone’s home address. A microcell will," and that a portable base station moved without updating its registered address will place you where it used to be.

The cross-border problem is the one that matters most in the corridors where residency is contested. A New York antenna with a stronger signal can put a New Jersey or Connecticut taxpayer in New York on paper, on an evening they never crossed the line. Accuracy also degrades with network traffic, population density, weather, and whether the person is stationary or moving, per the CPA Journal analysis. A purpose-built GPS tracking application, that article notes, produces something closer to a continuous data stream at roughly 25-foot accuracy. Google Timeline is not that. It samples, then infers.

And there is the category error underneath all of it. A device is not a person. The Nonresident Audit Guidelines already warn auditors about what they call false indicators, listing "credit card purchases in New York by children, phone calls by housekeepers, and children or relatives staying at the New York address as a guest of the taxpayer when he may not be in New York." A phone left charging in a drawer at the Manhattan apartment while you are in Palm Beach generates exactly that class of error, except this time it is generating it against you in a file you produced.

One warning from that Hodgson Russ piece deserves to be repeated without softening. The authors state that "it is never acceptable to advise a client to turn off their phone, or leave their phone at home, while in New York." Curating a location record to manufacture absences is not clever planning. It is fraud, and the pattern it leaves behind is more conspicuous than the days it was meant to hide.

In a New York statutory residency case the burden sits entirely on the taxpayer, and the standard is clear and convincing evidence. The guidelines, quoting the Tax Appeals Tribunal in Matter of Julian H. and Josephine Robertson, DTA No. 822004, describe how that burden can be met: through "testimonial evidence, documentary evidence, or a combination of the two." The manual then lists the routes that have worked. A combination of documentary evidence and affidavits showing an overall living pattern succeeded in Matter of Armel. Contemporaneously maintained diaries or calendars supported by credible testimony succeeded in Matter of Moss. Credible testimony was held sufficient in Matter of Avildsen, where the Tribunal reversed an ALJ who had ruled that testimony alone could not carry the burden absent corroborating documents.

Moss is the case worth studying if you are planning to lean on a Timeline export, because the record there was flawed in exactly the ways yours will be. The Department challenged the taxpayer’s business diaries on the grounds that "the entries were not always clear as to the taxpayer’s location and some were not contemporaneous." The Tribunal accepted them anyway, because the diaries were supplemented by credible testimony and travel reports. Gaps and ambiguity did not sink the record. Corroboration saved it.

The losing side is just as instructive. In Matter of Roth the Tribunal found the taxpayer’s diaries "illegible" and "meaningless" in determining days in and out of New York. In Matter of Rubin, the taxpayers were held statutory residents of both New York State and City because neither maintained contemporaneous records of their whereabouts and their testimony lacked specificity. And the Tribunal in Matter of R. Michael Holt, DTA No. 821018, set out the general expectation these audits run on: they "are very fact intensive and require specific evidence through substantiating contemporaneous records to show a taxpayer’s whereabouts on a day-to-day basis during each year in question. Such records could include not only day calendars but airline tickets, restaurant and hotel receipts and credit card statements."

The guidelines also show what corroboration looks like in practice, in an example that maps directly onto how a Timeline export will be received. If phone bills show calls from a Connecticut home from early morning until late afternoon and nothing places the taxpayer in New York, "it would be reasonable to conclude that the taxpayer was not in New York." That presumption "would not apply if phone bills for the New York residence showed calls on the same day," where the taxpayer has not complied with record requests, or where the taxpayer has "demonstrated a pattern of being in both locations on the same day."

So the honest answer to the question in the title is this. Google location history is admissible in spirit and useful in practice as a contemporaneous record of where your device was. It is not self-proving, it is not independent, and standing alone it looks like a diary you generated. Paired with flight itineraries, hotel receipts, card transactions, and toll records that were created by someone else and happen to agree with it, it becomes the spine that organizes the rest. That is the same structure that won Moss and Robertson, and its absence is what lost Roth and Rubin.

One more thing a Timeline export cannot do is answer a question the state is not asking. Day counting is decisive in some jurisdictions and close to beside the point in others, and the corridor you are in determines how much weight this file can carry.

New York is the jurisdiction where it matters most directly. The statutory residency test is a bright line, and someone moving from New York to Florida while keeping a New York home is standing on the wrong side of a threshold that gets counted in whole calendar days. There, a defensible day count is the case. New York is also, per ResidencyIQ’s dossier research, one of the two most aggressive exit-audit states in the country, with document-intensive audits that commonly run 12 to 24 months.

California is the opposite shape. There is no bright-line day count. The Franchise Tax Board applies a closest connection analysis, and the domicile factor list practitioners still call the Bragg factors after Appeal of Bragg, 2003-SBE-002, weighs where your ties actually sit. Someone moving from California to Texas can have an immaculate Timeline export showing 300 Texas days and still lose on a house, a business, a doctor, and a set of club memberships that never moved. The statute of limitations posture sharpens this. For filed part-year or nonresident returns California has four years to assess, but if no California return was ever filed for a year the FTB believes you were a resident, there is no statute of limitations at all under R&TC section 19057(a).

Illinois sits between the two, and it produced the cleanest demonstration that day counts do not decide these cases. In Cain v. Hamer, 2012 IL App (1st) 112833, 975 N.E.2d 321, retired snowbirds who declared Florida domicile in 1995 spent 1,666 days in Illinois against 1,700 in Florida over 1996 through 2004, an average above 183 Illinois days a year, with roughly $1.9 million in Illinois tax and penalties at stake. They were held to be Florida residents. The court weighed Florida licenses, voter registration, a Florida declaration of domicile, club spending, and credit card data showing 73 percent of expenditures and 61 percent of transactions outside Illinois. Anyone moving from Illinois to Florida should read that as both reassurance and warning: the day count alone neither convicts nor acquits.

If you want to see what your own export actually says before deciding how much of your case rests on it, the Google Timeline residency importer reads the file in your browser and reports days per state per year against each jurisdiction’s own statutory threshold. Nothing is uploaded. It will also show you, very quickly, how many days in your year Google has no opinion about.

The single highest-value action is the least interesting one. Open Google Maps, go to Timeline, then Location and privacy settings, and look at the auto-delete control. If it says three months, every year you might be audited on is already gone or going. Change it to keep the data until you delete it, or to 36 months if you prefer a ceiling, and understand that neither choice retroactively restores what was already purged.

Then export what you have, now, rather than in the second year of an examination. The export lives on the device, and devices are lost, replaced, wiped, and stolen. Enable the encrypted backup if you want it to survive a new phone. Save a dated copy of the export somewhere durable, and save it once a year rather than once, so that the file itself has a history of having existed before anyone asked for it.

After that, stop treating the export as the case and start treating it as an index. For every month it describes, pull the records that were created by someone else and would independently corroborate it: boarding passes and flight itineraries, hotel folios, card and bank transactions with locations, toll records, building access logs, utility usage at both residences. Where those agree with Timeline, you have the Moss structure. Where they contradict it, you have found the days that would have been fought over anyway, and you have found them while the receipts are still retrievable and your memory of the trip still works.

And keep the arithmetic in mind rather than the app. Presence for any part of a calendar day is a day in most day-count states. Air travel around midnight, a lunch across a state line, and a Sunday evening return are exactly the events a sampled location record handles worst and an auditor handles most literally.

The Mobility Map records days and nights across states as they happen, against each jurisdiction’s own threshold, so the year is counted while it is being lived rather than reconstructed from whatever survived a retention setting. Evidence Vault holds the categories the state’s own record list asks for: calendars, travel itineraries, financial records, and property and residence documents. AuditIQ surfaces thin days and retained-tie exposure, and advisor sharing lets a CPA or tax attorney review the chronology directly.

This article is informational and does not evaluate any individual’s tax situation. ResidencyIQ is not a law firm or accounting firm; work with a qualified CPA or tax attorney on your own residency change or audit.

Google, "Updates to Location History and new controls coming soon to Maps" (December 12, 2023), is the source of the on-device storage change, the quoted language about visits and routes being saved to a map on each of your devices, the automatic encryption of backed-up data so that no one including Google can read it, and the change of the default auto-delete period from 18 months to three months: https://blog.google/products/maps/updates-to-location-history-and-new-controls-coming-soon-to-maps/.

Google Maps Help, "Manage your Google Maps Timeline," is the source of the statement that Timeline is off by default for a Google Account and can only be turned on by opting in, that the data shown on Timeline comes directly from the device, that backups are stored as an encrypted copy on Google’s servers, and that the auto-delete control offers deletion of Timeline data older than 3 months, 18 months, or 36 months in addition to keeping it until the user deletes it: https://support.google.com/maps/answer/6258979.

The Register, "Google Timeline location purge causes collateral damage" (December 13, 2024), is the source of the gradual and ongoing rollout, the user reports of unexpectedly lost location history on Reddit and Google’s support forums, and the quoted user reaction to the default option: https://www.theregister.com/2024/12/13/google_timeline_purge/.

Lars Daniel, Forbes, "Google To Stop Giving Location Evidence To Law Enforcement" (October 8, 2024), is the source of the Sensorvault migration description and the quoted conclusion that the change would make it impossible for Google to search and return a user’s location data to law enforcement, and that investigators must instead identify the person and device: https://www.forbes.com/sites/larsdaniel/2024/10/08/google-to-stop-sharing-location-data-with-law-enforcement/.

Richard Resch, Criminal Legal News, "Beyond Geofence Warrants: Chatrie, Reverse-Location Surveillance, and the Fourth Amendment Fight Over Digital Dragnets" (May 1, 2026), is the source of Google’s March 2026 amicus brief statement to the Supreme Court that it can no longer respond to geofence warrants based on Location History, the disclosure that Google objected to over 3,000 such warrants on constitutional grounds, and that more than 2,500 were withdrawn: https://www.criminallegalnews.org/news/2026/may/1/beyond-geofence-warrants-chatrie-reverse-location-surveillance-and-the-fourth-amendment-fight-over-digital-dragnets/.

The Location History Format reference documents the Semantic Location History schema and is the source of the placeVisit fields locationConfidence and visitConfidence, the placeConfidence enumeration of LOW_CONFIDENCE, MEDIUM_CONFIDENCE, HIGH_CONFIDENCE, and USER_CONFIRMED, the activitySegment confidence values of LOW, MEDIUM, HIGH, and UNKNOWN_CONFIDENCE, the candidate activity probability range of 0 to 100 summing to no more than 100, and the note that the distinctions among the confidence fields are undocumented: https://locationhistoryformat.com/reference/semantic/.

New York State Department of Taxation and Finance, Nonresident Audit Guidelines (December 2021), is the source of the Chapter VIII list of personal records typically requested in a residency audit, the absence of any smartphone, GPS, or location-application category from that list, the clear and convincing standard and the Robertson formula of testimonial evidence, documentary evidence, or a combination of the two, the summaries of Matter of Armel, Matter of Moss, Matter of Avildsen, Matter of Roth, and Matter of Rubin, the Matter of R. Michael Holt (DTA No. 821018) passage on substantiating contemporaneous records, the Connecticut phone bill corroboration example and the conditions under which the presumption does not apply, and the false indicators passage: https://www.tax.ny.gov/pdf/2021/misc/nonresident-audit-guidelines-2021.pdf.

Timothy P. Noonan and Andrew W. Wright, Hodgson Russ, "Another (!) Note on Cell Phone Records" (June 29, 2020), is the source of cell records being described as the go-to third-party source of day count proof, the definition and red flags of data trailing including the zero bytes-up and bytes-down pattern, the microcell versus regular cell site distinction, and the statement that it is never acceptable to advise a client to turn off their phone or leave it at home while in New York: https://www.hodgsonruss.com/Noonans-Notes-Blog/another-note-on-cell-phone-records.

Corey L. Rosenthal and Fred R. Komarow, The CPA Journal, "NY Residency Audits and Electronic Data Records" (November 2016), is the source of the Department obtaining cellphone records directly from carriers, the quoted description of what carrier records contain including the location of the cellular antennae, the factors that affect accuracy, and the description of a GPS tracking application producing an almost continuous data stream at roughly 25-foot accuracy: https://www.cpajournal.com/2016/11/23/slt-ny-residency-audits-electronic-data-records/.

Appeal of Bragg, 2003-SBE-002, the California four-year and unlimited assessment periods under R&TC section 19057(a), the New York exit-audit posture and 12 to 24 month audit duration, and the Illinois reconstruction methods come from ResidencyIQ’s own dossier research, with underlying citations on the California, New York, and Illinois residency guides. Cain v. Hamer, 2012 IL App (1st) 112833, 975 N.E.2d 321, supplies the Illinois day counts of 1,666 against 1,700, the roughly $1.9 million at stake, and the credit card evidence of 73 percent of expenditures and 61 percent of transactions outside Illinois.