Get the highest-resolution file
Screenshots and re-shares compress and crop. Search engines match on visual features that survive better in the original resolution.
Provenance check
Reverse image search: the step detector scores can't replace.
No classifier, metadata reader or ELA heatmap can tell you where an image first appeared. Reverse image search can — and it's usually faster than waiting for a detector verdict anyway.
Method
A working reverse-search routine.
Run at least two engines
Google Images, Bing Visual Search and TinEye index different crawls. An image absent from one often turns up in another.
Sort by date, not relevance
You want the earliest appearance, not the most-linked one. Most engines let you filter or sort results by date.
Check the earliest source's credibility
A stock photo library, a known AI-art gallery, or a disclosed AI showcase account all answer the question differently than a first-person photo post.
Reading results
What a match — or no match — actually tells you.
| Result | What it suggests | What it doesn't prove |
|---|---|---|
| Oldest match is a stock/AI-art site | Image is stock or AI-generated, being reused out of context | Nothing about who's reusing it now or why |
| Oldest match is a personal account, years old | Likely a real, previously-posted photo | The account itself could still be compromised or misattributed |
| No matches anywhere | Either genuinely new, or too recent/obscure to be indexed yet | Not evidence of anything by itself — combine with metadata and visual review |
Combine it
Pair this with the rest of the workflow.
Reverse image search answers "where has this appeared before." It doesn't replace metadata inspection or the local checker — run all three before a decision that matters.
Worked example
A full investigation, start to finish.
Say a profile picture on a dating or marketplace account looks suspiciously polished. Save the largest version available (open the image directly, not a thumbnail). Run it through Google Images and Bing Visual Search. Google returns three results: a stock photography site, a "AI face generator showcase" gallery, and an unrelated Instagram account using the same photo as their profile picture two years earlier. That combination — appearing on an AI-art showcase site — is close to decisive: the photo is very likely a generated face being reused across multiple unrelated identities, which is a well-documented pattern in romance-scam and fake-account investigations. If instead the oldest match were a single personal account with years of consistent, unrelated posts, that would point the other way.
Tool notes
Face search vs. general image search.
Google Images, Bing, TinEye
Match on the whole image's visual features. Best for finding where a specific photo or graphic has appeared before, including on stock or AI-art sites.
Dedicated face-search engines
Some tools specialize in matching a cropped face against other photos containing a similar face, useful when the surrounding image differs but the face is reused. Treat matches as leads to verify, not conclusions — face-matching has real false-positive rates too.
Searching a frame from a video
Extract a clear frame (a moment with a clean, front-facing view works best) as a still image, then run it through the same reverse-search process used for photos.
FAQ
Reverse search questions.
Is reverse image search free?
Google Images, Bing Visual Search and TinEye all offer free reverse search with no account required for basic use.
What if the image returns zero results?
That's common for very new, very obscure, or heavily cropped images — it isn't evidence the image is fake or real by itself. Combine with metadata and visual review.
Does reverse search work on AI-generated images?
Sometimes — if the same generated image has been posted elsewhere (stock sites, AI-art galleries, prior reuse), search will surface that. A never-before-shared generated image won't have prior matches.