Verify a Reference Video's Location—or Leave It Unresolved
Verify a Reference Video’s Location—or Leave It Unresolved
A caption is a claim, not a coordinate. The clip you are about to build a pitch around arrived with a place attached to it—a town, a quay, sometimes just "somewhere on the north coast"—and it has been reposted enough times that the name now reads less like an assertion and more like a fact about the world. It is still a sentence someone typed.
The check described here is narrower than it sounds. You are not establishing the event, the date, or who filmed it. You are asking whether a specific spatial relationship visible in the clip can be reproduced at the place it claims to be, and whether a competing place fits better. If the answer turns out to be "not established," that is a usable result—better, in most pitches, than a plausible match you cannot defend in a footnote.
What follows is a method walked through on an invented clip. Both places named below are invented too. No real location, person, or address is being assessed, and the point is the shape of the reasoning, not a claim about anywhere you could visit.
Define the place claim and the safe scope of the check
Write the claim in a form that can fail. "Ashenford quay, last Tuesday" is not testable as written. "The depicted quay is at Ashenford" is. So is "the camera stood on the Ashenford quay." Those are different claims, and they fail in different ways—which is exactly why you separate them before you start looking.
Preserve the raw material first: the file, its original filename, whatever URL you first saw it at, the caption text exactly as it appears, and the date you saw it. Record the repost count at that moment if it matters to you. It does not matter to the geography. A number of reposts is a record of how far a sentence has travelled, not a second witness.
Then split the possible findings into levels, because most projects need far less precision than a researcher in the middle of a search wants to deliver:
- Which stretch of coast.
- Which named public site.
- Where the camera stood and which way it faced.
- When it was filmed.
A pitch that says "filmed on this coastline" needs the first level. A pitch that says "filmed at Ashenford" needs the second. Almost nothing needs the third, and nothing in a pitch should imply the fourth unless you can support it separately, because location never establishes capture date or authenticity.
Note also that the depicted landmark and the camera are two different objects. A building in frame tells you where the building is. It does not tell you where the person holding the phone was standing. Those can be a few meters apart or a few hundred, and the gap matters later.
Keep the exercise on a nonsensitive public landmark. Applying the same inferences to a private home, an identifiable individual, or a facility with security implications produces a different kind of harm and a different kind of publishable answer, and the method does not transfer cleanly. Related to that: do not push unredacted footage or frames from a client's unreleased material into third-party services unless someone has authorized that. The search is not worth an accidental disclosure.
The invented clip: fourteen seconds, handheld, no usable audio, a phone-app watermark in the corner. The caption reads "Ashenford quay, last Tuesday." The opening frame has water along the bottom edge, then a stone quay edge running across the frame, then the quay surface, and behind it a long shed seen along its side. The shed has a wide dark doorway at the right-hand end of the visible wall. A tall brick chimney rises behind the shed's roof, to the left of the doorway, its base hidden by the roofline and its cap clear of the ridge. The camera pans right across the clip. In the closing frame the shed has left the frame to the left, the quay edge continues to a far corner, and beyond it, across open water, sits a low headland with a dark tree line. Nothing man-made stands between the quay's far corner and the headland. There is no legible sign and no plate.
Two things worth saying out loud about that opening frame. The water in the near band means the camera was probably not standing on the quay at all. It was on a boat, on steps, or leaning out past the edge. And the dark rectangle in the shed wall is a doorway of some width; you can observe its position, not what it was for.
Inventory relationships that could discriminate between places
Most weak location work is a list of objects. Both places have a brick chimney. Both have a stone quay. Both have a shed with a slate roof. This list tells you the clip is somewhere coastal and industrial and nothing else. A familiar sign is a lead, not a conclusion, and duplicated architecture is not a coincidence you get to bank.
What discriminates is arrangement. So build a short list of relationships and their positions in the frame:
R1 — Order along the shore. The chimney appears to the left of the doorway. The two stand at different distances from the camera.
R2 — Occlusion. The shed's roof ridge crosses the chimney's shaft; the base is hidden, the cap clear. The chimney is on the far side of the shed from the camera.
R3 — Termination. Past the quay's far corner there is open water and then the headland. No built structure intervenes.
R4 — Count. One straight quay edge with one visible corner. No second arm, no breakwater, in frame.
R1 is the only one of the four that can reject a candidate from the whole water side of a quay at once, but only under a condition that is easy to lose. Take two things standing on the same line parallel to the shore—the same distance back from the water. For a camera on the water side of that shore, the left-to-right order of those two things in the frame does not change as the camera moves along the water. Standing on a boat south of a quay that runs east–west, looking north, east is to your right wherever you are along it. Move a kilometer west and it is still true. Two objects the same distance from the camera swing through the frame together, so their order holds.
The condition is the same-distance part. Put one of the two nearer the water than the other and the nearer one's bearing swings faster as you move along; past the sight line joining the pair, the order reverses. The doorway sits in the shed's water-facing wall, the chimney stands behind the shed, and the two are not the same distance from anything on the water. The order you see in the clip is an observation about the clip, not a property of the shore it shows. It becomes a test of a candidate place only if that place's own imagery puts the second feature the same distance back from the water—which is a thing to read off the views, not to assume.
Which side of the frame the order lands on is a second premise, from a second source. From water south of a quay, facing north, east falls right. From water north of it, facing south, the same order runs the other way. Read the water side off the candidate's chart before you convert a compass direction into a frame side, or the deduction rests on something nobody wrote down.
R2 is weaker, because occlusion depends on where you stand. It is a useful corroborating observation and a bad basis for a rejection.
Across the list, keep observed shape separate from inferred identity. "A tall brick stack, square in section, with a slight taper" is an observation. "The old mill chimney" is an inference borrowed from a story about the place. Record them in different columns, or at least in different sentences.
And record the gaps as gaps. The shed's west end is not in frame at any point in the clip. The doorway's interior is dark. The seaward face of the quay is not visible. You will be tempted, later, to fill these from what you expect a quay to look like. Do not, because the gaps are where your two candidates will separate.
Test alternative hypotheses using independent material
Build at least two candidates. Do not let the caption supply both, or you will spend the afternoon comparing a place to itself. In this invented case the alternative, Carrow Bay, came from the headland profile and the coastline type—a plausible harbor sixty kilometers along the same shore—not from the caption.
Then go looking for views you obtained yourself, and record what each one is: who published it, the date attached to it, the resolution, and the area it covers. A published map and a dated aerial photograph are different instruments and answer different questions.
Independence is the part people skip. If the same still frame appears on six sites, you have one view with six addresses. Check the watermark, the earliest upload date, and whether the sites are mirrors of each other. This is a different job from tracing an image back to its original poster—knowing who first posted something does not tell you what the caption's place name is worth—but the independence check borrows the same habit.
For each candidate, hold the date next to the claim it is being asked to support. Imagery from a period other than the clip's has to be treated as evidence about that period, not this one.
Against invented Ashenford: a public harbor chart and an aerial view dated 2023. The chart puts the harbor water south of the quay, shows a shed on the quay, and shows a brick chimney standing east of the shed, between the shed and the harbor's eastern approach, with open hardstanding between the two structures. In the aerial, the stack's base and the shed's quay-side wall stand the same distance back from the water.
Against invented Carrow Bay: a 2019 aerial at low resolution, and a 2011 street-level photograph taken facing northwest, away from the water. The aerial shows a shed with a chimney behind it, north of the roof line—reproducing R2. It does not resolve the shed's south wall, so the doorway's position relative to the chimney cannot be read off it. The 2011 photograph shows the wrong elevation entirely.
Which means Carrow Bay cannot be tested on R1. That is a coverage gap, not a match and not a mismatch. Missing coverage and contradicted arrangement are different findings and they license different sentences later. Write down which one you have.
The sequence being followed here—collect visible clues, process them, search for candidate places, check the candidates against independent views—is the one set out in the geolocation guide published by Exposing the Invisible (https://kit.exposingtheinvisible.org/en/geolocation.html). The method is the durable part. Its descriptions of particular search services age quickly, and I would not read them as a promise about what any tool does today.
Try to disprove the leading match
Now pick the relationship that has to hold if the caption's place is right, and go looking for the version of it that cannot.
If the clip is Ashenford, the chimney must appear to the left of the doorway. At Ashenford the chimney stands east of the shed entirely, which puts it east of the doorway as well, and the aerial puts both the same distance back from the water. The chart puts the water south of the quay, and the opening frame's near band of water puts the camera on the water side of the shore it shows, facing that shore. So a camera at Ashenford, in the clip's position, faces north with east on its right, and the chimney must appear to the right of the doorway—from every camera position on that water, because the two features stand the same distance back from it. The arrangement observed in the clip cannot be produced at the Ashenford quay.
Two things follow. First, the cameras-on-the-water caveat from earlier stops mattering here, with its condition carried rather than dropped. A boat off the quay does not rescue the match, because the order of two features the same distance back from the water holds for every camera on that side of it. The vantage question changes what a camera can see; it does not reverse the order of two things equidistant from it. Second, this is a positive contradiction, not an absence. You are not arguing that some expected feature is missing from the frame. You are arguing that a feature stands in the wrong place.
Two rescue attempts are worth running anyway, because the discipline is the point.
Could the shot have been filmed from land instead, from behind the sheds? That would flip the left-right order. It would also put the water beyond the building rather than in the near band—and the opening frame has water along its bottom edge. And a camera behind the shed, facing the shore, has the shed's north wall toward it, while the doorway sits in the water-facing wall on the far side of the building. The shot would have to show a wall the clip does not show. The rescue fails twice over, on the water and on the doorway.
Could the chart and the aerial be showing one stack when a second, unmapped stack stands west of the doorway? It depends on what ground the views cover, and that is a thing you can state. The chart and the 2023 aerial take in the quay frontage and the strip of ground immediately behind it, from the western approach to the eastern one. A stack standing west of the doorway—at the frontage, or just behind the shed where the roof would hide its base—falls inside that ground, and neither view shows one. That is a contradicted arrangement, not an unseen one.
What the views do not resolve is the ground further inland, past the strip the aerial covers. A stack there cannot be excluded by these views, and that belongs in the file as a coverage gap, filed separately from the rejection and not used to support it.
What genuinely could overturn the rejection is a change in the structures, a mistake in the imagery's date, or a re-reading of the views that puts the stack and the doorway at different distances back from the water—in which case the order test rejects nothing and you are left holding a coverage question instead. This is why the caption's date is not a side issue. The clip says "last Tuesday"; nothing in the clip establishes that. So the rejection you record is conditioned: for the period covered by the available views, and on the readings those views support, the Ashenford arrangement contradicts the clip.
Record what you rejected and why, and record the observations you did not lean on. R4 is a good example. The chart of Ashenford shows a breakwater east of the quay, and no such structure appears in the clip. That looks like another mismatch, and it might be, but it is also exactly the kind of thing that sits outside a fourteen-second pan. Do not let it into the rejection. Do not delete it either. An inconvenient observation that survives your reasoning belongs in the file with a note about why it was not decisive.
Write the narrow geographic conclusion
The result of this exercise is three sentences, and they do different jobs:
Contradicted. The published views put Ashenford's stack and the shed's doorway the same distance back from the water, and put the water south of the quay; from every camera position on that water the stack stands to the right of the doorway. The clip puts it on the left. For the period those views cover, the arrangement in the clip is not Ashenford's.
Possible but untested. Carrow Bay reproduces the chimney-behind-shed relationship. Its imagery does not show the shed's south wall, so the one relationship that would separate it from the field cannot be checked there.
Unresolved. Where the clip was actually filmed. Also unresolved: the camera's exact position, the direction of the pan in world terms, and the date.
Then list what you are explicitly not claiming, because a pitch will otherwise absorb it by implication. Not the event. Not the date. Not the maker. Not what the caption means, or why someone wrote it. Not authenticity. The location check and the authenticity question are separate jobs, and clearing one does not clear the other.
Do not attach a confidence percentage to any of this. A number like "70 percent confident in the Ashenford rejection" implies a measurement that no amount of reposting produced, and it tends to survive into the pitch as a fact while the coverage gap behind it quietly falls away. Write down what you looked at, what each source could actually show, and what you concluded. That is more persuasive than a score and it is checkable.
For the project itself: a contradicted caption does not invalidate the pitch. Depending on what the pitch needs, you can drop the clip, or describe it as an unattributed circulating clip of an unnamed quay and use it for what it visibly shows—a working quay with a shed and a stack, filmed from the water—and build the location claim on material you can stand behind. If the pitch is about how a caption travels, the contradiction is your opening rather than your problem.
The strongest sentence available at the end of this is the negative one, and it is worth keeping narrow: on the publicly available views, the caption's place is ruled out for the period those views cover, and nothing else is established. What would take that back is a dated image showing Ashenford's chimney standing west of the shed at the time of the clip; a record of a second stack on ground those views do not cover; a re-reading of the views that puts the stack and the doorway at different distances back from the water, which ends the order test rather than reversing it; or evidence that the imagery you used describes a later arrangement than the clip. Until someone produces one of those, the honest line in the pitch is that the clip's location is unknown, and the caption is not a source.
Frequently asked questions
Is the clip in this piece a real video from a real quay?
No. The clip, Ashenford, and Carrow Bay are all invented, and the piece states that no real location, person, or address is being assessed. The point of the walkthrough is the shape of the reasoning—how a caption's place claim is defined, tested against independent imagery, and either contradicted or left unresolved—not any claim about a place you could visit.
Why can the chimney-to-doorway order only reject a candidate place under a condition?
The left-to-right order of two features stays stable as a camera moves along the water only while both stand the same distance back from the water. Move one nearer the water and its bearing swings faster, and past the sight line joining the pair the order reverses. At Ashenford the chart and the 2023 aerial put the stack and the shed's quay-side wall the same distance back, so from every camera on the water side the stack must appear to the right of the doorway. The clip puts it on the left, which is what turns the observation into a rejection.
What is the difference between a coverage gap and a contradicted arrangement?
They license different sentences. At Ashenford, the published views place a feature in the wrong position relative to the doorway, which is a positive contradiction. At Carrow Bay, the chimney-behind-shed relationship is reproduced, but the imagery never resolves the shed's south wall, so the one relationship that would separate it cannot be checked there. That is missing coverage—not a match and not a mismatch—and it needs to be filed as its own finding.
Could the shot have been filmed from land behind the sheds instead of from the water?
The piece runs that rescue and finds it fails twice. Filming from behind the shed would put the water beyond the building rather than in the near band, and the opening frame has water along its bottom edge. It would also face the shed's north wall, while the doorway sits in the water-facing wall on the far side; the shot would have to show a wall the clip never shows.
Why does the piece refuse to attach a confidence percentage to the result?
A number like "70 percent confident in the rejection" implies a measurement that no amount of reposting produced, and the piece notes that such a figure tends to survive into the pitch as a fact while the coverage gap behind it quietly falls away. Writing down what was looked at, what each source could actually show, and what was concluded is described as more persuasive and checkable. What would take back the rejection is named instead: a dated image of the stack standing west of the shed, a record of a second stack on ground the views do not cover, a re-reading that puts the stack and doorway at different distances back from the water, or evidence the imagery describes a later arrangement.