From a region to a standpoint.
A street-level position
The standpoint the frame was captured from, with the direction the camera was facing.
Corroboration
The independent sources that confirmed it, each one named, so the claim can be checked without rerunning anything.
A refusal, when it is right
Where the frame supports a region but not a street, it says so and stays at the wider radius rather than inventing precision.
Street furniture
Poles, bollards, drainage, kerb cuts and paving, matched against the region.
Sightlines
The geometry between foreground and background fixes a standpoint a single object cannot.
Legible text
Where words resolve to a real place. Text alone never counts as an address-level cue.
Reverse geocoding
The proposed point is checked back against what is actually there.
Address-sized claims require GPS, a verified landmark, resolved place text, or several independent visual matches.
Unsupported results are widened and demoted rather than presented with false precision.
The standpoint is where the camera was, which is not a claim about who was holding it.