InvestigationU

OSINT

Reverse Image Search: Four Engines, Four Answers

These four engines do not share an index and do not use the same matching. Running one and stopping is the commonest mistake in image work.

EngineStrongest atUse it for
Google LensObjects, places, text, productsIdentifying what is in the picture
Bing Visual SearchBroad web coverage, cropping toolsA second opinion, and searching part of an image
YandexFaces and near-duplicatesFinding the same person elsewhere
TinEyeExact and edited copies, with datesWhere an image first appeared

Which one for which question

“Where else is this exact photo?” — TinEye. It matches copies rather than similar-looking things, and it sorts by oldest, which tells you where an image originated.

“Who is this person?” — Yandex, by a wide margin. Its matching on faces and near-duplicates outperforms the others consistently.

“Where was this taken?” — Google Lens, which is strongest at landmarks, signage and reading text inside the image.

“Is this stock or stolen?” — TinEye first for the earliest appearance, then Bing for breadth.

1

Check the file before you search it

Run it through the EXIF viewer first. If it has not been through a social platform it may still carry the camera, the date, a serial number and GPS coordinates — which answers the question without any searching at all.

Take the hash at the same time. If the image matters, you want a fingerprint of it as received.

2

Crop before you search

Engines match the whole frame. If the useful part is a tattoo, a logo, a sign or a face in a group, crop to it and search the crop separately as well as the whole image.

Bing’s visual search lets you drag a box over part of an image and search only that region, which is the fastest way to do this.

3

Search the variants

Try the original, a cropped version, a mirrored version, and a version with any watermark cropped out.

Mirroring matters more than people expect — profile pictures get flipped constantly and a flipped copy will not match on some engines.

4

Search the face separately

Crop tightly to the face and run that alone through Yandex. Backgrounds dominate matching and a tight crop removes the noise.

5

Follow the earliest result, not the best one

TinEye sorts by date. The oldest appearance is usually the source, and the source is what tells you whether a profile photo was taken by the person or lifted from a stock library or somebody else’s account.

A profile photograph whose earliest appearance is a 2016 stock listing is a fake account, and that is a finding in itself.

6

Compare candidates properly

When you have two images you think are the same, do not eyeball it. Put them through the image comparison tool — identical hashes mean the same file, and a high perceptual score means the same photograph re-saved.

What a negative result means

Very little. No match means the image is not in that engine’s index, which is the normal state for an ordinary photograph taken by an ordinary person. It is not evidence the photograph is original, and it is certainly not evidence the person is real.

Do not upload sensitive images to search engines. A client’s photograph, an image from an active case, anything involving a minor — uploading it hands a copy to a third party you have no agreement with. For those, work from the metadata and the content of the frame instead.