Free lesson

Where Did This Picture Come From? A Two-Minute Reverse Search

Where Did This Picture Come From? A Two-Minute Reverse Search — Spot Deepfakes and AI Fakes: Check Photos, Videos and Voices. Start with origin, not a diagnosis; Choose a tool and prepare the image; Read matches as leads, not ready-made answers; When the first search finds nothing; Check the claim separately from the picture; The two-minute pass and its…

Demo edition. The materials are undergoing editorial review.

Topics covered

  • Start with origin, not a diagnosis
  • Choose a tool and prepare the image
  • Read matches as leads, not ready-made answers
  • When the first search finds nothing
  • Check the claim separately from the picture
  • The two-minute pass and its limit

Start with origin, not a diagnosis

When a picture carries a surprising claim, first ask: where did it appear before, and in what context? That is often more useful than trying to decide immediately whether AI made it. A real photograph may have been moved from another year, country, or event. An AI-generated picture may have first appeared on its creator’s page with an honest label.

Use a reverse image search: give a search service an image file or image address, and it looks for visual matches and pages containing them. A result does not automatically establish the photographer, capture date, or truth of the caption. It gives you a trail to inspect on the pages themselves. The aim of a quick first pass is to find the earliest available version, read its caption, and compare it with the current claim.

Choose a tool and prepare the image

Start with one service. Google Lens accepts an uploaded file and may return pages containing the same or similar images. Bing Visual Search also accepts a file, address, or photo. TinEye is useful for exact and altered copies and can sort by when its crawler first saw a match. That is a TinEye indexing date, not proven evidence of the first publication date.

On a phone, save the picture or take a screenshot, then open it in an available service. On a computer, upload the file or use your browser’s contextual image search. If the first index finds nothing useful, check a second one because services cover different pages. Do not upload private or intimate images without permission: the file goes to an external service, whose own handling rules apply.

Read matches as leads, not ready-made answers

First separate exact or near-exact copies from pictures that merely share a subject. Check the positions of people, objects, and background details. A crop, resize, or overlaid caption does not necessarily make it a different source image.

Open several matches. Record the page date, credited photographer or agency, caption, and address. TinEye’s “oldest” sort means the earliest detection by its crawler, so compare that with the date shown on the page. An old match can disprove “taken today,” but the oldest copy you find may still be a repost.

Fictional example. A chat labels a flooded-platform photograph as today’s emergency in City A. The popular results repeat that caption. An older 2019 news page shows the same platform, names City B, and credits a photographer. The narrow conclusion is that this photograph does not show today’s emergency in City A. It does not establish whether City A currently has a flood.

When the first search finds nothing

Change the query before drawing a conclusion. Crop away the chat frame, caption, emoji, and large empty areas; retain a distinctive building, object, or part of the scene. Reddit discussions among OSINT practitioners provide a useful practical point: even a small change to the selected search area can produce different matches. Search the complete scene and distinctive sections separately.

Try a cleaner copy if the screenshot is blurred. If the picture appears mirrored—for example, visible text runs backwards—flip it horizontally and search again. Add a text query containing visible words, the event type, and the claimed location. Check another search index.

No match does not prove authenticity or AI generation. The picture may be new, private, poorly indexed, heavily altered, or synthetic. The sound interim record is “no earlier version found with these searches,” not “original confirmed.”

Check the claim separately from the picture

Reverse search mainly answers “where else has this picture appeared?” By itself, it does not establish who captured it, where and when it was captured, whether it was edited, or whether the current caption is accurate. Open a new search and check the event itself: place, date, names, and specific claim. Look for independent reporting from news agencies, responsible public bodies, or fact-checkers rather than many accounts repeating one text.

Save an auditable trail:

  • what you checked and where it reached you;
  • the exact claim;
  • the earliest page found, its date, credit, and caption;
  • what independent sources report;
  • verdict: confirmed / probably real / don’t know / probably fake / fake;
  • action: do not share, correct the post, or take no action.

“Don’t know” is appropriate when evidence is insufficient. A link and a screenshot showing the date let another person repeat your check.

The two-minute pass and its limit

For a rapid check, make a short pass: save the picture, run one reverse search, select exact matches, open several early pages, and compare their captions with the current claim. If the trail immediately leads to an older event or the page of an AI-image creator, that may be enough reason not to forward the current post.

Do not treat two minutes as a guarantee of finding the original. When the stakes are significant or results conflict, continue with another index, several crops, dates checked on the actual pages, and a separate event search. One common mistake is treating the first or most popular result as the original. Another is treating dozens of identical posts as independent confirmation. They may all derive from one unverified post. Stop only at the narrow conclusion the trail actually supports.

Practical assignment

Trace one viral picture

Continue to learning

Course outline

  1. Test Your Eye: How Well Can You Really Spot AI?
  2. Seconds: Is This Worth Checking?
  3. Where Did This Picture Come From? A Two-Minute Reverse Search
  4. Anatomy of an AI Image: Clues, Not Verdicts
  5. Real Photo, False Story: Cheapfakes
  6. Where and When Was This Taken?
  7. Video: Keyframes, Seams and AI Clips
  8. A Voice on the Phone: Clones and the Callback Rule
  9. The Video Call From “Your Boss”
  10. Labels and Watermarks: What Provenance Can and Can't Tell You
  11. AI Detectors: One Piece of Evidence Among Many
  12. “That’s a Deepfake!” When People Lie About Fakes
  13. Breaking News, Elections and Disasters: The First Hours
  14. If Someone Makes a Deepfake of You or Someone You Love
  15. Protect Your Voice and Face
  16. How to Correct a Relative in the Family Chat
  17. Your Verification Protocol: 30 Seconds, 5 Minutes, 30 Minutes