To find the original source of an image, search the full picture in at least two reverse image search engines, open the matching pages, and follow credits and links back until you reach the creator, the agency or the first credible publication. The earliest result you find is a strong lead, not proof: it is only the earliest copy in the indexes you checked, and it may not be the original or the copyright owner.
The short version: eight steps
- Save what you started with. Keep the image file, the page where you saw it, the caption and today's date.
- Write down your question. Are you looking for the creator, the first publication, a licence, a bigger copy, or whether a caption is true? Each needs different evidence.
- Search the full image in two engines. Google Lens for broad coverage, TinEye for exact copies and dates. Add Bing or Yandex if both come up short.
- Search the text. Any watermark, credit line, asset number, sign or caption phrase goes into a normal text search in quotes.
- Open the candidates, not the thumbnails. Read the credit, caption and links on each page.
- Follow the chain backwards. If a page credits or links to another source, go there. Repeat until the trail stops.
- Compare the image itself. Check crops, mirroring, colour, edges and small details to confirm it is the same photograph, not one from the same shoot.
- Record a bounded conclusion. "Earliest copy I could find" or "credited to X on the agency's own listing" is better than an unsupported "original found".
Decide which "original" you need
People mean different things by "original source". Naming yours at the start stops you from stopping too early, or chasing evidence you do not need.
| Your question | What counts as an answer | Best evidence |
|---|---|---|
| Who made it? | Photographer, illustrator or studio | Creator's portfolio, agency listing, credit line on a primary publication |
| Where did it first appear? | Earliest credible publication | Dated pages, archived copies, publisher records |
| Can I reuse it? | The rights holder and licence terms | Stock or archive listing, licence text, written permission |
| Is the caption true? | The original context of the picture | Original caption and article from the first publisher |
| Is there a bigger copy? | A legitimate higher-resolution file | Creator or library download, press kit |
An early repost can settle chronology while saying nothing about authorship. A licensed agency copy may name the photographer even though it is not the earliest indexed page. If your real question is a caption or a news claim, use the companion guide on verifying a photo before you share or publish it.
Keep the image URL and the page URL
These are two different links and you need both.
- Page URL: the address in your browser bar while you view the article, post or profile. It preserves the title, caption and surrounding claim.
- Image URL: the address of the image file itself. On desktop, right-click the picture and choose Copy image address (Chrome, Edge) or Copy Image Link (Firefox, Safari). On a phone, long-press the image and look for Copy image address or Open image in new tab; labels vary by browser.
Do not confuse Copy image (copies the pixels) with Copy image address (copies the link). A few things to know about image URLs:
- An image URL does not need to end in .jpg, .png or .webp. Many sites serve images through long URLs with resizing parameters. Removing parameters such as
?w=400sometimes returns a larger version. - A
blob:URL only exists inside your current browser tab. A search engine cannot fetch it, so download or screenshot the image instead. - Signed cloud links (Google Drive, Dropbox, S3 with tokens) can expose access to private files. Do not paste them into a public search box.
Record the access date beside both URLs. Pages change, and a later reviewer needs to know what you saw and when. If a page has gone, say so; a cached thumbnail in a results grid does not confirm what the old page claimed. Try the Internet Archive's Wayback Machine for an archived copy.
Search with more than one engine
No engine indexes the whole web, and they are built for different jobs. Google's own help page explains how to search with an image.
| Engine | Strongest at | Useful feature for source work |
|---|---|---|
| Google Lens / Google Images | Broad coverage, objects, places, text in images | The Exact matches filter shows pages with the same picture |
| TinEye | Exact and edited copies of one image | Sort results by Oldest, Biggest image or Most changed |
| Bing Visual Search | A second large index, products | Shows "pages that include this image" |
| Yandex Images | Some regions and sites others miss | Lists matching sizes of the same image |
Start with the full, uncropped image. If results are dominated by the wrong thing (a busy background, a logo, a face), run a second search on a crop of the part you care about. For a side-by-side look at the engines, see Google Lens alternatives for tracing image sources and the tool selection worksheet.
Why the earliest result is not always the original
The earliest result is only the earliest result in the places you searched. Older copies can be missing because:
- the first publication was in print, a private group, a messaging app or a paywalled archive;
- the original page was deleted, moved or redesigned and lost its old date;
- the creator's portfolio was rebuilt recently, so it now carries a newer date than the reposts;
- the search engine never indexed that site, or dropped it.
Use the phrase "earliest located copy" unless you have creator or publisher records that support "original". Keep your search date in your notes, because a later search may surface pages that were invisible today.
What the dates actually mean
A date next to a search result, a page's updated date and the date a photo was taken are separate facts.
| Date you see | What it usually describes | What it does not prove |
|---|---|---|
| Date in a search result | When the page was published, updated or crawled | When the photo was taken or first published |
| Article publication date | When that article went live | That its images are new; articles often use file or stock photos |
| "Updated" note | A later edit to the page | When the image was added |
| TinEye "first found" | When TinEye's crawler first saw that copy | That nobody published it earlier |
| EXIF capture date | The camera clock setting, if metadata survived | Accuracy; clocks are wrong and metadata is editable |
Write narrow sentences: "This image appears on a page dated 3 March 2024" is accurate. Save "first published" and "taken on" for evidence that says exactly that, such as an original caption with a date or an agency record.
Use watermarks, credits and text in the picture
Watermarks
Copy the watermark exactly: creator name, agency, handle or asset number. Search it as text, in quotes, with a short description of the scene. If it names a stock agency, use that agency's own search or asset ID lookup. A matching listing usually gives the contributor, the asset ID and licence options.
A watermark is a lead and a likely rights notice, but not conclusive. Reposters add their own watermarks, and aggregator accounts stamp their logo on other people's work. Do not remove a watermark to get "cleaner" search results; search the image as it is. If the text is unreadable, look for a larger legitimate preview rather than "enhancing" it with an AI tool, which invents characters.
Text inside the image
Signs, posters, captions, shirt slogans and asset codes often find a source faster than visual matching. Transcribe only what you can read and mark uncertain characters. An unusual phrase or event title beats a common word like "sale". A poster for a one-off event can be easy to find by its title even if its background photo appears on hundreds of pages, and the event page often carries the correct credit.
Many sites repeating one caption is not independent confirmation.
Photo credit or person's name?
A name near a photo can belong to the photographer, the agency, the uploader or the subject. Look at the role words:
- "Photo by", "Photograph:", "/ Agency" or a name in small type under the image usually credits the creator.
- "Courtesy of" usually names who supplied it.
- The article byline belongs to the writer, who may have used an unrelated stock image.
- A name in the caption sentence ("Sam Lee at the opening...") usually names the subject.
"Alex Morgan / Agency B" under a travel photo most likely credits the photographer, not the traveller. If the page does not name the person shown, leave them unnamed; source verification can be complete without identifying everyone in the frame.
Check metadata and Content Credentials
If you have the original file (not a download from social media), its EXIF data may hold the camera model, capture date, copyright field or creator name. ExifTool reads all of it; most photo viewers show the basics under Info or Properties. Most social platforms and messaging apps strip metadata on upload, so an empty field means "not available here", not "never existed". Metadata is also easy to edit, so treat it as supporting evidence.
Some cameras, editing apps and publishers now attach Content Credentials based on the C2PA standard. Upload the file to Content Credentials Verify to see any signed history of who created and edited it. Missing credentials are normal and prove nothing. If you suspect the image is generated or heavily edited, use the AI image and deepfake checks alongside this workflow.
Find a higher-resolution copy the right way
Creators and agencies keep full-size files; reposters keep shrunken copies. So a bigger legitimate copy is often your best lead.
- In TinEye, sort by Biggest image. In Google Lens, open Exact matches and look at the listed dimensions.
- Open the largest candidates and check who publishes them and how they credit the image.
- Compare real detail, not just pixel dimensions. Wallpaper sites often host upscaled copies that are larger but blurrier.
- Download through the site's normal options, and check the licence before reusing it.
Another frame from the same shoot is not the same picture: compare hand positions, expressions and background details. Upscaling a small copy does not recover the original detail; if you upscale for display, label the result as edited and never use invented texture or text as evidence.
Special cases
Collages, memes and before-and-after images
Each panel can come from a different source. Save the whole collage with its caption, then crop each panel into its own file, numbered, without altering the panel content. Search each one separately and record a source per panel. Do not apply one panel's credit to the others. For a before-and-after image, finding both photos may show they were taken in different places or years, which matters more than either match alone.
Book covers
Record title, author, publisher and edition first, because editions change artwork. Search the whole cover to find the edition, then search a crop of just the photo or illustration to find the underlying image. The copyright page often credits the cover designer and the image source separately; the photographer is not the designer of the whole cover. If you have an image and want the book, a whole-image search usually lands on bookshop or library listings with the title beside it.
Logos, artwork and objects
Keep the full photo for context and crop the logo, signature or label without redrawing it. Search the crop and any legible lettering separately, then confirm against the brand's official site, a museum collection or the artist's catalogue. A generic mountain emblem matches many outdoor brands; a matching wordmark plus product category is a much stronger clue. Identifying a logo or painting does not prove that a particular item is authentic, original or valuable. For artwork in a room photo, crop to the artwork only; you do not need to know whose room it is.
Landmarks in your own travel photos
Check your own records first: album date, itinerary, and location data in the original file (exported or messaged copies often lose it). Then compare distinctive architecture, signs and terrain with public reference photos from several angles. Label a likely landmark as likely until the details and your timeline agree. Use this for your own photos and public landmarks only, never to work out where a person lives or is.
Archive and public-domain images
Search any caption, accession number or institution name visible near the image, because old photos exist as many scans with different borders and contrast. Then open the archive's own catalogue record, compare the full frame and read the rights statement for that specific item. Old does not mean free to use. Wikimedia Commons and Openverse are useful for finding openly licensed versions, but check each file's licence page.
Stock photos
Reverse search finds where a picture appears; the stock library's own search confirms the asset ID, contributor and licence. Match the exact composition, since one shoot produces many near-identical frames with different IDs. An unwatermarked copy elsewhere grants none of the library's rights.
Pick the method for your job
| You are... | Lead with | Watch out for |
|---|---|---|
| A photographer checking reuse | TinEye plus Google Lens Exact matches, tested on 3 photos with known publications | Licensed clients publishing on other domains; empty results are not proof of no reuse |
| Checking product photos | Copy search for the photo, visual search for the product | Same product, different photo; similar product, different item |
| Tracing an illustration | Whole composition plus a crop of a distinctive detail | Shared style is not copying; look for identical small irregularities |
| Tracing a frame from a video | A clean still, searched in two engines | See reverse searching a video frame |
Label your confidence and keep a worksheet
Plain labels explain more than invented percentages.
| Label | Use it when |
|---|---|
| Exact copy, credited source | Same image, and the page names a creator or agency you could check |
| Earliest located copy | Same image, oldest credible page you found, no creator record |
| Possible derivative | Same composition but cropped, flipped, recoloured or only a small preview |
| Similar, not the same | Same scene or subject, different photograph |
| Unresolved | Nothing reliable found, or evidence conflicts |
Keep one row per search: input filename (number your crops), engine, candidate page URL, access date, what the page actually shows, and a separate conclusion column. "Same composition; page credits Studio A; no publication date" is a useful entry. "Original found" is not. If a better source turns up later, add a row explaining why the conclusion changed rather than overwriting the old one.
When the image shows a person
A copy-oriented search traces the photograph. If what you need is other photos of the same face, for example to check whether a profile picture belongs to someone else or to find where your own photos are posted, that is a different search; face search vs. reverse image search explains the difference. Facehunch runs that kind of search for $4.99 per search and shows each match with its source link and a similarity score, which you then verify with the steps above. Use it to check someone you are already talking to or to find your own photos online, never to identify or locate a stranger.
For general picture tracing without a face involved, start with a reverse image search and follow the chain until the evidence stops.
Written by Facehunch with AI assistance and checked against the cited sources. Our editorial policy.