# How to Find the Original Source of an Image

> Run the picture through at least two reverse image search engines, such as Google Lens and TinEye, then look for the oldest and largest copy. Check that copy's page for a photographer credit, an upload date and intact metadata. The original is usually the earliest, uncropped, highest-resolution version with a named author.

- URL: https://fake-ai.com/images-and-video/find-original-image-source/
- Language: en
- Updated: 2026-10-19
- Publisher: Fake AI (OnlineBuilders LLC, https://onlinebuilders.biz)

Old pictures have a habit of coming back. Flood photos from one country resurface during floods in another. Protest images from 2015 get reposted as if it were taken this morning. None of that needs AI. Real pictures with false dates do the job just as well, and they cost nothing.

The fix is the same in both cases: find the first appearance. Once you know who took a picture and when and where, most questions answer themselves.

## Why the first upload matters more than the pixels

Pixel analysis tells you whether an image was edited. The first upload tells you what the image is. Real photos with false captions are misinformation, and no artefact check will catch it. Strange-looking photos that a named photographer published with full EXIF data in 2017 are almost certainly not AI, because the tools to fake it convincingly did not exist then.

Provenance also protects you from the opposite mistake. People increasingly dismiss real photos as "AI" because they look too dramatic. Finding the agency original with a photographer's byline ends that argument.

## The workflow, step by step

**From a forwarded picture to its first appearance**

1. **Get the best copy** Ask for the file itself, not a screenshot. Save it at full size.
2. **Clean it up** Crop away borders, captions, emojis and platform frames. Keep the main subject.
3. **Search two or three engines** Google Lens, Bing Visual Search and TinEye. Note every match with a date.
4. **Sort by age and size** Look for the oldest copy and the one with the highest resolution.
5. **Read the page** Caption, photographer credit, agency, date, location. Check the archive if the page changed.
6. **Check metadata** If you find a downloadable original, read its EXIF and Content Credentials.

Cropping matters more than people expect. Screenshots of a post with the image inside give the search engine a picture of a tweet. Cut it down to the photo, and if text has been printed over the subject, try a second crop that avoids the text.

If nothing turns up, the copy you have may be mirrored. Fakers sometimes flip images precisely to break reverse searches. Flip it horizontally in any photo editor and search again.

## Which search engine does what

Each engine has its own index and its own idea of "similar". Google Lens tends to return visually related scenes and products, plus places, which is great for identifying a landmark and less great for finding exact copies. TinEye matches copies of the same image, including resized, cropped and colour-adjusted versions, and does not do facial recognition. Bing Visual Search sits somewhere in between.

**Reverse image search engines compared**

| | Google Lens | Bing Visual Search | TinEye |
|---|---|---|---|
| Finds exact and edited copies | partly | partly | yes |
| Identifies places and objects | yes | yes | no |
| Shows visually similar scenes | yes | yes | no |
| Free without an account | yes | yes | yes |
*Our assessment based on each service's documented features, October 2026.*

Use the engine that fits the question. Or use all three. "Where was this taken?" is a Lens job. "Who published this first?" is a TinEye job, backed up by Lens.

## Reading the metadata of the copy you find

Once you have a candidate original, download it from the source page, not from a social network. Then read its metadata. ExifTool by Phil Harvey is the standard free tool, and it reads EXIF, GPS, IPTC, XMP and C2PA data ([ExifTool](https://exiftool.org/)). On the command line:

```
exiftool -a -G original.jpg
```

Look at the camera model, the capture time, the software field and any GPS coordinates. Agency photos often carry an IPTC caption and a copyright line. If the software field names an editing app, that is normal for published photos and says nothing about manipulation on its own.

Two cautions. Metadata can be edited, so treat it as a claim, not a proof. And the absence of metadata is meaningless on social media copies, because the platforms strip it. Signed Content Credentials are much harder to tamper with than plain EXIF. Our guide to [Content Credentials](https://fake-ai.com/images-and-video/content-credentials-c2pa/) explains how to read them.

## When the source page has changed or vanished

Search results often point to pages that no longer show the image, or that now carry a different caption. Before you give up on a match, paste the URL into the Internet Archive's Wayback Machine and look at the earliest snapshot. Archived captions are frequently more honest than the current one, because they were written before the picture became controversial.

Dates on web pages deserve the same scepticism as metadata. Content management systems often show the date of the last edit, not the first publication. Comment timestamps, archive snapshots and the date in the image's own URL path (many sites store uploads in folders like /2019/06/) are harder to fake.

## Traps that send people to the wrong source

The most common mistake is stopping at the first big website. Large aggregators, wallpaper sites and stock libraries copy images at scale and often rank above the original. Being listed on a stock site does not make the uploader the photographer. Some contributors upload pictures they did not take, and the libraries cannot check every file.

The second trap is trusting the highest-ranked result instead of the oldest. Search engines rank by relevance and popularity. You have to sort or compare dates yourself.

The third is circular sourcing. Ten news sites that all credit "social media" are one source, not ten. Follow the credit chain until you reach a person or an agency with a name.

## Signs that you have found the original

**Signs of a likely original**

- [ ] **Oldest** dated copy among all search results
- [ ] **Largest** resolution, often wider than the viral crop
- [ ] A named photographer, agency or account that posts similar original work
- [ ] A caption with place and date that fits the visible details
- [ ] Metadata or Content Credentials that match the caption
- [ ] Other photos from the same scene, shot seconds apart

The last point is underrated. Photographers rarely take one frame. If the account also has the shot before and after, with slightly different angles, you are almost certainly looking at the person who stood there.

## Video and AI images need extra steps

For video, extract keyframes and search those. The InVID-WeVerify browser plugin, funded under the EU's Horizon 2020 programme, does this and bundles keyframe extraction, reverse search and forensic tools in one place. It runs mainly in Chrome.

For images that might be fully AI generated, there may be no earlier copy at all, because the picture didn't exist until someone made it. Complete absence of prior copies, combined with a brand-new account as first poster, is a warning sign in itself. Continue with our checklist on [how to tell if an image is AI generated](https://fake-ai.com/images-and-video/how-to-tell-if-image-is-ai/), and for clips see [how to spot a deepfake video](https://fake-ai.com/images-and-video/how-to-spot-deepfake-video/).

If your team checks images every day, for insurance claims, marketplace listings or user uploads, the searching can be partly automated. Studios such as [OnlineBuilders](https://onlinebuilders.biz/solutions/data-analytics) build pipelines that flag reused or altered images before a person looks at them. For a single picture, three searches and five minutes of reading are usually enough. You will find more checks for pictures and clips in our [image and video section](https://fake-ai.com/images-and-video/).

## FAQ

### Which reverse image search is best?

There is no single winner. Google Lens is strong on objects, places and products, while TinEye focuses on finding copies of the same picture and lets you compare them. Use at least two engines, because their indexes overlap only partly.

### Can I reverse search a video?

Not directly. Take screenshots of a few distinct frames, or use the free InVID-WeVerify browser plugin to extract keyframes, then search those stills.

### Why does the search find nothing for an obviously viral photo?

The copy you have may be cropped, mirrored or covered with text, and search engines need time to index new uploads. Crop to the main subject, remove borders and try again a day later.

### Is it legal to reverse search a photo of a person?

Searching where an image appears online is generally fine. Using face search tools to identify strangers raises privacy and data protection issues, which is why mainstream search engines keep such features limited.

Photo: Friedrich Haag, CC BY-SA 4.0 (https://commons.wikimedia.org/wiki/File:2015_04_08_009_Analogfilme.jpg)