Reverse Image Search: Image Verification and Source Tracing

Images travel faster than context. Reverse image search tells you where an image has appeared before, which is usually enough to know whether a "breaking news" photo is actually from three years ago and a different country. This tutorial covers the four engines that complement each other, and when to use which.

Who this is for

Beginner

Journalists verifying user-generated content, researchers checking profile photos, civic investigators documenting viral claims.

What you'll need

A browser.
The image you want to check, saved locally at the highest resolution you can find.
Optional: a second browser profile for engines that behave differently when logged out.

How it works

Each engine indexes images by computing a perceptual hash or feature vector and comparing queries against its index. They differ sharply in what they see: Yandex leads on faces and on Eastern European and Russian sources; Google Lens leads on objects, text, and recent indexing; Bing is strong on product and stock imagery; TinEye is strong on exact duplicates and oldest-known appearance. Use all of them.

Step-by-step walkthrough

Save the original image. Right-click > Save image as. Avoid screenshots when the original file is available — screenshots lose resolution and strip metadata.
Strip and check metadata first. Before any reverse search, run ExifTool:
```
exiftool suspect.jpg
```
Camera model, timestamp, and GPS can short-circuit the whole investigation. See the metadata extraction tutorial for more depth.
Run the four engines in parallel.
- Google Lens: https://lens.google.com/ — upload or drag in.
- Yandex Images: https://yandex.com/images/ — the camera icon.
- Bing Visual Search: https://www.bing.com/visualsearch.
- TinEye: https://tineye.com/. Open each in a new tab. Do not rely on any single result.
Sort TinEye by "oldest". TinEye's "oldest" sort is the fastest way to find the earliest known publication, which is usually the original source.
Crop and re-search. If the image is a composite or has distracting elements, crop to the most distinctive subregion and re-run. For faces, crop tightly; for scenes, try a distinctive background element (a street sign, a building, a logo).
Combine with a text search. If an engine returns a caption, search that caption in quotes. Verified captions from established outlets are a stronger anchor than any hash match.
Geolocate if needed. If the image shows an outdoor scene, combine reverse search with geolocation — match architectural features, signs, and vegetation against satellite and Street View. Bellingcat's public writeups are the best worked examples of this workflow.
Document the result. Record for each engine: URL of the top three matches, timestamp, and screenshot. Note discrepancies between engines — they are frequently meaningful.

Common pitfalls

Engine bias. A single-engine null result means nothing. Different engines see different corners of the web.
Low-resolution input. Reverse search on a compressed thumbnail returns noise. Get the highest-resolution source before searching.
Skipping the crop. Full-image searches miss matches where the subject appears in a different composition.
Trusting "first seen" dates. Engines report when they indexed a copy, not when the image was created.
Face-search services that scrape without consent. Several engines specialise in face matching against scraped profile photos. Their use is regulated or prohibited in many jurisdictions and their accuracy is uneven. Treat results as leads, never as identifications.

Verifying your findings

An image is verified when at least two independent sources can be identified and corroborated for time, place, and subject. Pair reverse-search hits with metadata, geolocation, and where possible direct confirmation from a party in the image. Document the chain as described in the analysis phase guide.

Apply this in practice

The verifying a viral image case study walks through a complete reverse-search workflow from a raw claim to a documented finding. For extended worked examples of image-led investigation, see the Epstein Revealed investigation series.