Short answer: A GPS-verified photo records the device''s reported position at the moment of capture and binds it to the image server-side, then compares it against the expected site. Typical outdoor smartphone accuracy is 3–10 meters; indoors and in dense urban settings it degrades badly, so location should be treated as a strong signal rather than an absolute gate.
How the verification runs
- The session requests location permission from the browser at join.
- At each capture, the device reports coordinates plus an accuracy radius.
- The server records that alongside its own timestamp and the image hash — the device cannot rewrite it afterwards.
- The coordinates are compared to the site anchor (the job or claim address, geocoded) and a distance plus confidence score is produced.
- The result is displayed with the photo and written to the audit trail.
Accuracy, realistically
| Environment | Typical accuracy | Implication |
|---|---|---|
| Open outdoor | 3–10 m | Reliable address-level verification |
| Suburban with tree cover | 10–25 m | Usually fine; can straddle adjacent lots |
| Dense urban canyon | 20–100 m | Multipath error; verify by street view shot too |
| Indoors | Wi-Fi/cell derived, 30–150 m | Confirms neighborhood, not the unit |
| Basement / steel structure | Often unavailable | Expect misses; do not hard-fail |
Where it fails — and what to do
- Permission denied. The most common cause of a missing location. Prompt clearly at join and explain why.
- Indoor capture. Anchor the session with an outdoor address shot at the start, then treat interior captures as inheriting that anchor.
- Large sites. A 200-acre solar farm needs a radius, not a point. Set the geofence to the site boundary. See geofencing.
- Spoofing. Mock-location apps exist. This is why location is one signal among several — device attestation, session binding, and rebroadcast detection cover the rest. See tamper detection.
- Legitimate off-site work. Equipment brought to a shop is not fraud. The system should flag the discrepancy for a human, not block the capture.
Design rule: advisory over blocking
Programs that hard-fail on missing GPS end up with field users who cannot complete legitimate work. The pattern that holds up: capture always succeeds, missing signals produce an advisory warning, active contradictions (coordinates hundreds of miles from the site) produce a hard flag, and the reviewer sees the reason. See GPS accuracy FAQ.
Privacy
Location is personal data in most regimes. Collect it only during sessions, disclose it in the join consent, retain it under the same policy as the media, and allow deletion. See GDPR and CCPA.
What the coordinates are actually derived from
"GPS" on a phone is shorthand for a fused position. The browser''s geolocation API returns whatever the operating system considers its best estimate, and that estimate blends several sources:
| Source | How it works | Typical error |
|---|---|---|
| GNSS satellites | Time-of-flight from GPS, Galileo, GLONASS, BeiDou | 3–10 m outdoors |
| Wi-Fi positioning | Nearby access point MACs matched against a database | 20–50 m in built-up areas |
| Cell tower triangulation | Signal from known tower positions | 100 m to several km |
| IP geolocation | Network registration data — a fallback, never a proof | City-level at best |
The accuracy radius the browser reports tells you which of these dominated. A radius under 15 metres is a satellite fix. A radius of several hundred metres is a network fix and should be labelled as such in the record rather than presented as a location.
Reading the confidence score
A distance from the site anchor is only half the picture; it has to be read against the accuracy radius. A capture 40 metres from the anchor with a 5-metre radius is a real discrepancy. The same 40 metres with a 300-metre radius is noise.
- Verified — anchor is inside the reported accuracy circle.
- Near — outside the circle but within a tolerance you set for the property type; a large rural parcel needs a wider tolerance than a townhouse.
- Advisory — no fix, denied permission, or a radius so wide the reading is uninformative.
- Contradicted — a precise fix that places the capture somewhere else entirely. This is the only state that deserves an alarm.
Anchoring a session properly
Most location disputes are avoidable with thirty seconds of process at the start of the call. The pattern that works across trades, claims and lending:
- Geocode the job or claim address before the session and store it as the site anchor.
- Open with an outdoor establishing shot: street number, meter, or building entrance, captured outdoors where the fix is strongest.
- Move inside and continue; interior captures inherit the anchored session rather than being judged individually.
- Capture a closing exterior shot if the inspection ran long, so the session is bracketed in time and place.
Because Virtual Inspection Pro sessions run in the browser, the guest is granting permission once at join rather than installing anything — which is the single biggest driver of whether you get a location at all.
Spoofing, and why location is never the only control
Mock-location tooling is freely available on both major mobile platforms, and developer options can feed arbitrary coordinates to any app that simply asks for a position. Any vendor claiming GPS alone proves presence is overselling. Location earns its place as one signal inside a stack:
- Live session binding — the frame came from an active call, not a file picker.
- Server-side time — the device clock is never trusted.
- Content credentials, so later edits are visible. See C2PA.
- Rebroadcast and screen-replay detection. See tamper detection.
- Human review of contradictions, because legitimate off-site work exists.
Worked example: a two-property dispute
A carrier receives roof damage photos for a duplex. The captures sit 18 metres from the anchored address with a reported accuracy of 6 metres — enough to place them on the neighbouring unit. Under a hard-fail rule the claim stalls. Under an advisory rule the adjuster sees the discrepancy, sends a targeted retake request for one establishing shot of the street number, and resolves it in a single message instead of a second dispatch. That is the whole argument for advisory design: it turns ambiguity into a cheap follow-up rather than an expensive truck roll.
FAQ
Does the photo have to contain GPS in its EXIF?
No — and it should not be trusted there. The server-side record is the authoritative one.
Can a guest disable location mid-session?
Yes, and that change is recorded. Captures after that point show as unverified.
Is Wi-Fi positioning good enough?
For confirming the general area, yes. For proving a specific unit in a complex, no — pair it with an address shot.
How accurate is GPS on a phone for property inspections?
Roughly 3–10 metres outdoors with a clear sky view, degrading to tens or hundreds of metres indoors, in basements, and between tall buildings. Good enough for address-level verification, not for identifying a specific unit without a supporting shot.
Can GPS location on a photo be faked?
Coordinates written into EXIF can be edited trivially, and device positions can be mocked. That is why the authoritative record is captured server-side during a live session and combined with time, session identity, and content credentials.
What happens if the field user denies location permission?
Capture still succeeds and the photo is marked as location-unverified. Blocking capture on a missing signal simply stops legitimate work from being recorded.
Does GPS verification work indoors?
Not reliably. Anchor the session with an outdoor shot at the start and let interior captures inherit that anchor.
Is collecting location data a privacy problem?
Location is personal data under most regimes. Collect it only during sessions, disclose it in the join consent, retain it on the same schedule as the media, and support deletion — see GDPR and CCPA.
Keep reading
- C2PA content credentials — signed provenance for field photos.
- Chain of custody for inspection evidence — a practical guide.
- Photo fraud in insurance claims — how verified capture closes the gap.
- Remote claims adjusting with live video — workflow and cycle-time impact.
- Virtual inspection checklist — permit-ready capture sequences.