Guide
One sentence carries this page: on a map screenshot the identifying detail is the geometry of the whole frame rather than a value in one place, so the only sound redactions are covering the entire map pane or not capturing it at all — and hiding the ends of a route is known not to hide where the route went.
The pin is not the location
Start with an inventory of what is still in the picture after you have covered the marker, the address bar and the labels. Every item here is drawn rather than written, so no amount of covering text touches any of it.
The road network. The strongest identifier in the frame and the one with no text in it at all. Junction angles, block lengths, the shape of a dead end, the kink where an old lane meets a new estate. Residential layouts are far less repetitive than they look.
Natural and drawn features. A coastline, a river bend, a lake edge, the outline of a park, a golf course, a reservoir, a cemetery. These are the most distinctive shapes on most maps and the least likely to be covered, because they do not look like data.
The scale bar and the zoom level. These convert the shape into real distances, which is what turns a suggestive picture into a matchable one. A north arrow or compass does the same job for rotation.
The provider's styling. Tile colours, label typeface, icon set and the way building footprints are drawn all identify which map product the screenshot came from, and therefore which dataset to compare it against.
Icons without their labels. A hospital cross, a station symbol, a school, a cluster of three cafes on one corner. The pattern of symbols is itself a fingerprint, and covering the words beside them leaves the pattern.
The route line. Discussed on its own below, because it is the case people most often think they have handled.
The pin, by contrast, carries almost nothing the rest of the frame does not already carry. It is the most satisfying thing to cover and the least useful.
Hiding the ends of a route does not hide where it went
This is the part of the page that is not a matter of opinion. It has been measured, more than once, on exactly this mechanism.
Fitness tracking services have a purpose-built feature for the problem: an endpoint privacy zone, which conceals the portion of a published route that falls within a chosen radius of a place you designate as sensitive, typically home or work. It is the same idea as dragging a black box over the two ends of a route in a screenshot — except done by the service that holds the full track, with a defined radius, at the data level rather than the pixel level. If the idea worked anywhere, it would work there.
At USENIX Security 2018, Wajih Ul Hassan, Saad Hussain and Adam Bates of the University of Illinois Urbana-Champaign analysed a month-long snapshot of public posts from a popular fitness tracking service — 21 million posts across 3 million users — and observed that 16.5% of users had endpoint privacy zones switched on. They then developed an attack that infers a user's protected location from the information still visible in those public posts, and reported that 95.1% of moderately active users were at risk of having their protected location extracted. The paper states that the affected companies were notified and had incorporated the proposed countermeasures by publication, so the specific services are not necessarily in that state today.
The obvious follow-up is whether this takes an attacker with research tooling. It does not. At CHI 2022, Jaron Mink, Amanda Rose Yuile, Uma Pal, Adam J. Aviv and Adam Bates ran an online study with 603 participants who were shown exercise routes with a privacy zone applied and simply asked to click where they thought the hidden location was. For the most common privacy zone size, 68% of guesses fell within 50 metres of the hidden location when participants were shown just three activities — roughly the footprint of one residential property, from ordinary people clicking on a map. Participants shown three activities were also more confident than those shown one, which is the uncomfortable part: the inference is not a lucky guess, and the people making it can tell when they have got it.
Those studies are about a specific feature on specific services, not about your screenshot, and that distinction matters. But the direction of the inference transfers exactly, and your screenshot is in a weaker position for three reasons. A box you dragged has no defined radius, so you are eyeballing how much to hide rather than setting it. The visible route runs right up to the edge of the hole, and its heading at that boundary points into it. And a rectangle laid over a radial problem tends to leave the approach visible on two sides. If a feature designed for this job leaks the endpoint to a casual observer with three samples, a rectangle over the end of a line is not going to do better.
One corollary is worth stating plainly, because it is the trap: this does not get better with repetition, it gets worse. Each additional route through the same covered area narrows the hidden region further — which is precisely why three activities beat one in the study. Two map screenshots of different trips from the same home, posted a week apart, are not two independent disclosures. They are one disclosure with two samples, and the covered area does not grow back between them.
What actually works
In rough order of effectiveness.
1. Do not capture the map. Most map screenshots are taken to show something that is not the map: a fare, a delivery confirmation, an ETA, an error message, a receipt, a charge you are disputing. Capture the window or the region that contains that thing and leave the map pane outside the frame. This editor has no crop, no resize and no rotate — every operation replaces pixels inside a shape you draw, and the export keeps your source's exact pixel dimensions — so deciding the frame at capture time is by far the cheapest control you have.
2. Screenshot the text view instead of the map view. Nearly all of these apps have a list, a receipt, a trip detail page or an activity summary that states the same facts in words. Words sit in places and can be covered one at a time. A map cannot. Switching screens before you capture converts an unsolvable redaction into an ordinary one.
3. Cover the entire map pane with one black box. If the map has to be in the frame — you are showing the layout of a dialog, or the panel will not render without it — then treat the whole pane as the sensitive region and fill it. One rectangle drag in Black Box mode, edge to edge, past the pane borders rather than up to them. This is the only partial-frame redaction of a map that is actually sound, and it is sound precisely because it is not partial with respect to the map.
4. If the geometry is genuinely the point, control the audience rather than the image. Sometimes you do need to show a route: disputing a delivery that never arrived, documenting a near miss, explaining a detour to a claims handler. You cannot redact a route into safety. What you can do is send it to one named recipient through a channel you chose, rather than posting it somewhere it will be indexed and archived, and say in the message what you have left visible and why.
What is not on that list: zooming out, zooming in, cropping to a fragment, blurring the pane, pixelating the pane, rotating it, or covering the pin and the labels and calling it done. Each of those feels like progress and none of them removes the geometry. Zooming out is the most tempting and among the worst — a wider frame trades legible street names for coastlines, motorway interchanges and river systems, and those are far more distinctive shapes than a residential street. Cropping to a small unlabelled fragment is the next most tempting, and a single unusual junction can be more identifying than a labelled view of a regular grid.
The panel beside the map leaks the map
Cover the pane and you are not finished, because the text around it usually reconstructs most of what you just hid.
Distance and duration. A distance from a known point describes a circle. Two distances describe a pair of points. A distance plus a direction describes a place. If either end of the trip is already public — an office, a shop, a restaurant you tagged last month — the other end collapses to a short list.
ETA and the clock. The same distance in different units, and the current time on the status bar completes it.
Departure and arrival times. Together these bound the speed, which implies the mode of travel, which sets a plausible radius around a known endpoint.
Recent searches, saved places and shortcuts. An autocomplete dropdown left open, a recents list, a Home or Work entry. These name places outright, and they are often the part of the screen nobody looks at twice.
Partial addresses. A street name with the number boxed out, a postcode with the last characters covered, a neighbourhood or suburb label. Partial is the operative word, and it is usually enough.
The account name or avatar. This is what attaches all of the above to a specific person rather than to an anonymous trip.
The arithmetic there is a general problem rather than a map problem: a redaction fails whenever something still visible lets the covered value be recomputed. It is set out properly in when visible totals let someone recompute a redacted number.
What the editor above does and does not do here
Three modes — Black Box, Blur, Pixelate — each available as a rectangle, an oval or a freehand lasso. Every operation replaces the pixels inside the path you drew and nothing else; there is no whole-image operation of any kind, and no crop, resize or rotate. The canvas is created at your image's exact pixel dimensions and displayed scaled to fit, so what you see is usually smaller than what you are editing. Undo restores stored pixel snapshots in draw order, Clear redraws from the original image the page kept, and Download PNG writes the canvas to a file named hideshot-<timestamp>.png.
For maps, two of those three modes are worse than useless, and the actual numbers say why. Blur downsamples the selected area by a fixed factor of ten and scales it back up. Drag it across an 800-pixel-wide map pane and you have resampled that pane to roughly 80 pixels across — and 80 pixels of width is ample to preserve a route line, a motorway, a river and the general direction of the grid as recognisable shapes. Pixelate is worse still: its block size is the shorter side of your selection divided by twelve, with a floor of six pixels, so a large selection gets correspondingly large blocks and the output is a mosaic about twelve blocks high. Twelve blocks is a coarse sketch of the route, and a coarse sketch of the route is the identifier. Both modes exist to discard fine detail while keeping coarse structure — a sensible trade on small text, and precisely the wrong one here, because on a map the coarse structure is the secret.
That leaves Black Box, which fills the path with flat #111111 and has no strength setting, so there is nothing underneath to work back from. Rectangle is the right shape for it: a map pane is a rectangle, and the lasso invites you to trace carefully around something you meant to keep, which is how half-covered maps get published. One mechanical caution — a rectangle drag under about four image pixels, or an oval under six, is discarded silently, so a small box over a stray label can simply fail to appear. Check each one is actually there before you download; on a dark map style, also check the box is visible against the tiles rather than blending into them.
Nothing here is persisted anywhere. No account, no saved presets, nothing written outside the open tab — the image, the undo snapshots and the retained original live in that tab's memory and go when it closes. And your image is never sent anywhere: the editor contains no fetch call, no XMLHttpRequest, no WebSocket and no address of any kind, which also means the page makes no geolocation request, has no idea where you are, and could not look anything up about your map even if it wanted to.
The underlying difficulty — an identifier that is not shaped like a region, and therefore cannot be covered by drawing a shape over it — is not unique to maps. The general case is set out in identifying details that a box cannot cover.
Common mistakes and misconceptions
“I covered the pin.” The pin is a marker, not a location. Everything that located it is still drawn around the hole where the pin was.
“I covered the address bar and every label.” You have made the place slower to search for, not harder to match. Road geometry contains no text, so text redaction does not reach it.
“I blurred the map, it is unreadable.” Unreadable and unidentifiable are different properties. A factor-of-ten downsample keeps the route and the grid as shapes, and shapes are what you were hiding.
“I zoomed out first so it is less specific.” A wider frame swaps street names for coastlines and interchanges. Those are more distinctive, not less.
“I cropped it down to a piece with no labels.” A junction is a fingerprint. A fragment of an irregular layout can beat a labelled view of a regular one.
“I covered both ends of the route, so the middle is fine.” On the purpose-built version of that idea, ordinary study participants put 68% of guesses within 50 metres of the hidden place after seeing three activities. A box you dragged by eye is weaker than the feature they defeated.
“It is only one screenshot.” Each further map of a trip through the same covered area narrows it. Posts accumulate and the hidden region never grows back.
“It is my own location, so it is my risk to take.” Shared maps carry other people by default — the friend whose house the trip ended at, the driver, the colleague in the carpool. And a route to a clinic, a shelter, a place of worship or a lawyer's office discloses a category about someone, not merely a coordinate.