Basic OSINT Photo Geolocation Using Landmarks, Shadows, and Signage

Save the image to your camera roll now. A forwarded photo of a crowd outside a mall arrives with a caption claiming the building is about to collapse. You have forty seconds and a phone. Before you decide whether to forward it, answer one question: where was this image actually taken? Ignore the metadata. Social platforms strip EXIF data the moment an image is uploaded. A screenshot removes it even earlier. What remains is the image itself: the signage, the shadows, the architecture, the vegetation, the language on the signs. Those are your evidence. Read them, then cross-reference what you see against Google Maps and Street View. This method outlasts any app.

EXIF metadata holds GPSLatitude, GPSLongitude, GPSAltitude, and GPSImgDirection. It is fragile. WhatsApp, Telegram, and every major social platform strip it on upload. A forwarded message, the primary vector for misinformation here, has been through at least one re-encoding that wipes the original data. A screen recording of the image removes it too. The reliable path is visual. Save the image to your phone's camera roll so you can zoom without losing resolution. Then open Google Maps in a new tab. Lateral reading, leaving the original content to check the source, begins with you not trusting a single image's own claims.

Signage is the fastest route to a location. It is designed to be read by humans moving past it. It gives you two things at once: the language and the text. Road signs here are primarily in English, with Malay, Chinese, and Tamil appearing on some. An English sign narrows the country but not the street. What you want is the small print. Look for the building name, the street name, and the HDB block number. The block numbering system is a block number plus a street name, like Blk 123 Bedok North Road. It is a precise locator on every residential block's facade. Read that number and type it into Google Maps. You get a pin. If the sign is partially obscured, zoom into the reflection on a window or the shadow cast by a nearby pillar. The text is often recoverable in the highlight.

No readable sign? The structure itself becomes the clue. The city's architecture mixes brutalist HDB flats from the 1970s, colonial-era shophouses with painted facades, and glass towers from the last two decades. Each has a signature. A flat-fronted concrete block with rectangular window slots and a void deck at ground level is an HDB. You find it here and in a few Malaysian towns. A two-storey terrace house with a five-foot way, the covered walkway along the street, is a shophouse. Its tile work and louvred windows date it. Cross-reference the style against the map. If the image shows a flat-fronted concrete block with a void deck, you are not looking at a landed property district. Google Maps Satellite View helps here. It shows the roofline and the layout that a street-level view hides. Compare the shape of the structure in the photo against the shape on the satellite image.

Shadows give you the time of day and the direction the camera was facing. Together these eliminate most of the map. The sun here sits almost on the equator. It rises close to due east and sets close to due west. A shadow pointing north means the time is around midday. A shadow pointing east means late afternoon. Find the shadow cast by a street lamp or a vertical signpost. Trace its angle. Then use SunCalc, a solar position verification tool, to match that angle to a date and time. This is chronolocation. Its primary indicators are shadows, weather, vegetation, and clothing. Long shadows stretching east mean late afternoon. A person's shadow pointing southwest means morning. That is not a precise address, but it is enough to reject any location where the sun would have been in a different position.

After you have a probable area, the vegetation and street furniture confirm or refute it. The trees are mostly rain trees, with their wide, flat canopies. The palms and hibiscus that appear in tourist photos are not everywhere. A row of palm trees along a road suggests a coastal or park area, not a dense housing estate. Street lamps, bus stops, and MRT station identifiers are precise anchors. The station codes are alphanumeric, like NS1 for Jurong East and EW24 for Jurong East. The lamp post ID is a different thing. It is a small metal plate on the post with a number and a letter. It exists so emergency services can pinpoint a location. Read that plate and you have the exact spot. The bus stop sign, with its route numbers and the service logo, narrows you to a street if the building across the road matches.

Now the method comes together. Open Google Maps. Find your candidate area. Drag the Street View pegman onto the road. Street View has covered the island since 2009. The coverage is not current everywhere, but the core districts are. You are looking for the same sign, the same facade, the same angle of a road, the same view down a lane. If the photo shows a sign on a structure at head height, Street View shows whether that sign is still there. If the sign is gone, use the satellite view to check the footprint and roofline. This is where you confirm the location or you do not. Be willing to reject your own guess. Confirmation bias is a common failure: you have a vague match, you want it to be right, so you stop looking. Find two independent matches, a sign and a building shape, before you commit.

The photo may not be new. Reverse image search tells you faster than any other tool. Google Images lets you upload a photo and find every other page where it appears. TinEye, the oldest, indexes image content as opposed to surrounding text. Yandex Images is the best at finding cropped and recoloured versions. That matters when someone has taken an old photo and applied a filter. Upload the image, not a link. Look for the earliest occurrence. If the earliest occurrence is a news article from a different country or a stock photo site, the image has been recycled. Google reverse image search supports JPEG, PNG, GIF, BMP, WebP, and SVG files up to 20 MB. TinEye does too. You are unlikely to hit a size limit. This is the fastest way to learn whether the photo is a cheapfake, media altered by cropping out context or mislabelling an old photo.

When you have narrowed the search to this city, the local systems make the final confirmation straightforward. The six-digit postal code, like 238859, is unique to a building, not a street. It appears on almost every facade. Type that code into Google Maps and you get the exact block. The HDB block number, read off the side of a residential block, works the same way. The MRT station code, like NS1 or EW24, is printed on station signs and on the side of trains. A photo of a station exterior gives you the code. Match the code to the station, then walk the surrounding streets on Street View to find the exact structure. The method stops being about the photo and starts being about the physical layout of a small, dense, well-documented place. If you have the code, you have the location within a minute of opening the app.

Geolocation becomes open source intelligence photo verification when you apply it to a claim, not just a curiosity. The method is the same. The stakes are higher, and the failure modes are specific. The first failure is verification bypass: sharing a claim with a comment like "is this true?" without waiting for an answer. That is amplification with extra steps. The second is source confusion: believing a claim because the person who forwarded it is trusted, without checking whether that person is the original source. The third is single-source settlement: accepting a claim after finding one source that confirms it. The Bellingcat guide "How to geolocate a photo or video," published on 26 January 2023, walks through the same steps: read the signs, check the shadows, match the building. The difference is you are now doing it to decide whether to report a POFMA direction or a police case, not just to settle a bar bet.

Geolocation confirms where a photo was taken. It does not confirm whether the event depicted actually happened. A photo of a flooded road here that actually shows a flooded road in Jakarta has been geolocated to Jakarta. End of matter. But a photo of a crowd outside a structure that is genuinely in this city does not tell you why they are there, or whether the structure is about to collapse. That distinction is the difference between a fact and an interpretation. Most people go wrong here. The language on the signage and the architectural style are neutral evidence. The caption is not. Verify the image, not the caption. Keep the two separate in your head.

What Each Tool Answers

No tool does this for you. Google Lens is a visual search engine that identifies objects and landmarks. It does not geolocate a photo. Jeffrey's Image Metadata Viewer, launched in 2006, shows EXIF data if it survived. It almost never does on a forwarded image. Forensically, launched in 2015, does error-level analysis and clone detection. It can show if a photo has been manipulated. It will not tell you where it was taken. SunCalc is a solar position calculator, not a map. The trap is thinking that a powerful tool is sufficient. Each tool answers one question. You have to know which question to ask. Geolocation is a method, not an app.

The most common mistake is skipping straight to reverse image search and stopping when it returns nothing. A photo never posted online returns no results. That is not evidence of authenticity. The second is tool over-trust: uploading an image to a single AI-detection tool and treating the result as definitive, without understanding the tool's error rate. The third is correction blindness: seeing a fact-check but dismissing it because it comes from a source you distrust. The fourth is the AI realism trap: believing an image is authentic because it looks real, when AI generation can now produce photorealistic synthetic media. None of these are fixed by a better tool. They are habits of mind. The only fix is the discipline of lateral reading. Open a new tab. Search for the organisation named in the caption. Check whether it exists. Check whether it has said anything about the claim. That is the whole job.

Uploading a suspicious image to a public reverse image search service may reveal your identity or your location. Think before you upload. If the image is sensitive, do not upload it to a public service. Google and TinEye retain the images you upload. They do not make them public, but you are trusting a third party with your search. The alternative is the visual analysis method, which requires no upload at all. Read the sign. Check the shadow. Match the building. You never send the image anywhere. That is the most privacy-preserving form of verification. It is also the most reliable, because you are not depending on a database that may not have the image. When you do use a reverse image search, use a cropped version that removes any faces or identifying details. The image is no longer yours once you upload it.

The sign is too small, too blurry, or too far away. The shadow is ambiguous because it is overcast. This is the failure case, and most people give up here. Do not. Zoom into the image in sections, left to right, top to bottom. Look for a detail you missed: a reflection in a shop window, a manhole cover on the road, a bus stop sign, a fire hydrant, a street lamp with a distinctive shape. These small details geolocate a photo because they are not what the photographer wanted you to see. A bus stop sign has the service number and the name of the stop. It is a precise locator. If you cannot read the text, look at the shape of the sign itself. It is standardized across the island. If you have nothing, move to the vegetation: the species of tree, the shape of the leaves, the density of the undergrowth. The more detail you extract, the more likely you are to find the one that matches.

Your phone has everything you need. Google Maps, Google Lens, and the browser are pre-installed on most devices. Take a photo of a place you know. Wait a day. Then geolocate your own photo using only the visual evidence and a map. This is the most effective training because you know the answer. You can tell when your reasoning is right and when it is wrong. Start with a building you can see from your window. Then a street corner. Then a park. Then a place you have visited once. Each time, write down the clues you used: the sign, the shadow, the building shape. Check your answer. After a week, you will geolocate a photo of a random street faster than you can type the address into a search engine.

Never forward a photo you have not geolocated yourself. Even if you agree with the caption. The forwarding is the harm. A message that goes out with a false location or a false date is the same as a lie, intended or not. Verification speed matters, but it is secondary. Check before you share. Next time someone sends you a photo, open Google Maps. Zoom into the area the caption claims. Look at the street names. Then open Street View and walk the block. If the buildings match, forward it. If they do not, do not. This is the difference between a person who repeats claims and a person who verifies them. It takes two minutes. The next photo that lands in your chat: geolocate it before you share it, or do not share it.

What is the first thing I should do with a photo I want to geolocate?

Save it to your camera roll so you can zoom without losing resolution. Then open Google Maps in a new tab. Do not rely on the version in the message. It is compressed and may have lost detail.

Why does the EXIF data never work when someone sends me a photo?

Social platforms strip EXIF data on upload. Messaging apps often re-encode images, removing the metadata entirely. The EXIF standard itself is reliable, but you never see it on a forwarded image.

How do I know if a shadow points east or west?

The sun rises in the east and sets in the west. A shadow that points toward the road in the morning points west. In the afternoon it points east. Trace the shadow's angle and use SunCalc to confirm the time of day.

What is the most reliable single clue in a photo?

A readable HDB block number is the most precise. It maps to exactly one building. The MRT station code, like NS1 or EW24, is a close second. If you have either, you have the location.

Can I geolocate a photo that was taken indoors?

Yes, but you are limited to the signage inside the building. That is often a shop name or a brand logo. A photo inside an MRT station will show the station code. A photo inside a shopping mall will show a shop sign you can search for.

What if the photo is a deepfake or AI-generated?

That is a different problem. Geolocation assumes the photo is real but mislabelled. If you suspect AI generation, look for inconsistencies in the shadows, the number of fingers, or the text on the signs. AI still struggles with these. The method on this page will not work on a purely synthetic image.

How do I verify the date the photo was taken?

Use chronolocation: the shadow length and direction, the weather, the vegetation, and the clothing of any people in the photo. The inter-monsoon periods, late March to May and October to November, bring heavy rain. A dry photo in that window is a clue about the date.

Key Facts at a Glance

• EXIF standard version: 2.32 (2019 revision)
• Google reverse image search file size limit: 20 MB
• TinEye reverse image search file size limit: 20 MB
• Google Maps Street View coverage started: 2009
• MRT station identifier format: Alphanumeric, e.g., NS1, EW24
• Postal code format: Six-digit numeric, e.g., 238859
• HDB block number format: Blk + number + street name
• Lamp post ID purpose: Emergency location identifier
• SunCalc solar position tool launched: 2013
• Jeffrey's Image Metadata Viewer launched: 2006
• Forenically image forensics tool launched: 2015
• Google Lens monthly active users: 500 million+ (2022)