No mainstream search engine lets you upload a whole video and find where else it appears. Instead, you reverse search a video by pulling out a few distinctive still frames, called keyframes, and running each one through image search engines such as Google Lens, Bing, Yandex, and TinEye. Combined with a careful check of upload dates, this is how journalists trace viral clips back to their source and catch old footage being passed off as new.
Key takeaways
- Reverse video search means reverse image search on frames taken from the video. Pick 3–6 sharp frames that show scene changes, landmarks, or on-screen text.
- You can capture frames with a screenshot, VLC's Video > Take Snapshot, ffmpeg, or the free InVID-WeVerify browser plugin, which extracts keyframes automatically.
- Search every frame in at least two engines. Google Lens and Bing are strongest for places and news coverage. TinEye sorts exact copies by date.
- Upload dates on re-posts are meaningless. Look for the earliest copy, then check who posted it and whether the scene matches the claimed time and place.
- A video with no earlier copies isn't automatically authentic. Check for signs of AI generation or editing before you share it.
How reverse video search works
Image search engines index billions of still images, but they don't index the inside of video files in a way you can search by upload. However, videos are just sequences of images, and the frames you care about often exist online as stills too: thumbnails, screenshots in news articles, social media previews, and images posted alongside the clip. If you search the right frames, you can usually find earlier copies of the video or reports about it.
Step 1: Extract keyframes
Choosing the right frames
Not every frame is equally useful. The best search frames are:
- The first frame and each scene change. Re-uploaders often trim the start, so check frames from a few seconds in as well.
- Frames with landmarks, signs, or storefronts. These help image engines identify the location.
- Frames with on-screen text or logos. TV station graphics, watermarks, and captions lead you to the original broadcaster or creator.
- Sharp, steady frames. Avoid motion blur, heavy compression artifacts, and transitions.
Option 1: Pause and screenshot
The quickest method works on any device. Pause the video at the frame you want, switch to full screen, and take a screenshot. On Windows, press Win + Shift + S. On a Mac, press Cmd + Shift + 4. On a phone, use your normal screenshot buttons. Crop out player controls, captions you added, and app interface before searching. Our phone reverse search guide covers searching screenshots on iPhone and Android.
Option 2: VLC media player
If you have the video file, VLC (free, for Windows, Mac, and Linux) captures frames at full resolution, without the player overlay. Pause on a frame and choose Video > Take Snapshot, or press Shift + S on Windows and Linux, or Cmd + Option + S on a Mac. According to the VideoLAN wiki, snapshots save to your Pictures folder on Windows and Linux and to the Desktop on a Mac. Press E to step forward one frame at a time and find the sharpest one.
Option 3: ffmpeg (automatic, for longer videos)
If you're comfortable with a command line, the free tool ffmpeg can export a frame at every scene change:
ffmpeg -i video.mp4 -vf "select='gt(scene,0.3)'" -fps_mode vfr frame_%03d.png
Lower the 0.3 threshold to capture more frames or raise it to capture fewer. To grab one frame per second instead, use -vf fps=1. Older ffmpeg versions use -vsync vfr instead of -fps_mode vfr.
Option 4: InVID-WeVerify verification plugin
The InVID-WeVerify plugin is a free browser extension built for journalists and fact-checkers. It was developed by AFP's Medialab with European research partners and continues to receive updates. It runs in Chrome and other Chromium-based browsers, and Firefox through a manual install. Its main tools for video are:
- Keyframes: Paste a YouTube or Facebook link, or upload a file. The plugin splits the video into keyframes based on visual changes. Right-click any frame to search it on Google, Yandex, TinEye, and other engines.
- Thumbnails: For YouTube videos, this tool reverse searches the platform-generated thumbnails, which are often indexed by search engines.
- Video analysis: Shows a YouTube or Facebook video's metadata, including its upload time, plus comments, which are useful for spotting "this is old" replies.
- Metadata and Magnifier: Reads metadata from MP4 files you upload, and lets you zoom into and enhance a frame to read signs or plates.
Some advanced tools in the plugin, such as its synthetic media detectors, require registering with a professional email address. The core keyframe and search features are open to everyone.
What happened to YouTube DataViewer? Amnesty International's YouTube DataViewer was the standard tool for pulling a YouTube video's exact upload time and thumbnails. If it's unavailable when you try it, InVID's Video analysis and Thumbnails tabs provide the same information.
Step 2: Search the frames
Run each frame through at least two engines. They index different parts of the web and handle cropped or low-quality frames differently.
| Engine | How to search a frame | Strengths for video frames |
|---|---|---|
| Google Lens | Upload at google.com (camera icon), or in Chrome open the menu and choose Search with Google Lens, then drag over a paused video | Huge index, news coverage, landmarks, and products. You can select part of the frame. |
| Bing Visual Search | Camera icon in the Bing search box | Good second opinion on places and news images. Lets you crop to an area. |
| Yandex Images | Camera icon at yandex.com/images | Often surfaces Eastern European and Russian-language sources other engines miss. |
| TinEye | Upload at tineye.com (JPEG, PNG, GIF, BMP, or TIFF, up to 20 MB) | Finds exact and edited copies. Sort by Oldest to see the earliest crawl date. |
Chrome's Lens tool is particularly handy for video. You can pause a clip on YouTube or a news site, open Lens from the Chrome menu, and drag a box around the part of the frame you want, without saving a screenshot. For Google image results, the About this image panel (in the image viewer's info menu, in supported regions) shows when Google first saw the image or similar versions, which helps you date a frame.
When a frame doesn't match anything:
- Crop tighter onto the most distinctive element, such as a building, sign, or logo.
- Flip the frame horizontally. Re-uploaders sometimes mirror videos to avoid copyright detection.
- Remove overlays. Crop away added captions, emojis, and reaction-video borders.
- Add keywords. In Google Lens, type a description, a place name, or a date in the "Add to your search" box.
Our full reverse image search guide and tool comparison cover each engine in more depth.
Step 3: Check upload dates and the earliest copy
A search that turns up dozens of copies doesn't tell you much until you sort them by time. Your aim is to find the earliest version and its original uploader.
- List every copy with its date. Note the platform, account, and posting date for each. YouTube shows the publish date in the description box below the video, TikTok shows it next to the username, and X shows it below each post.
- Get exact times where you can. Visible dates are often rounded ("2 years ago") or shown in your local time zone. InVID's Video analysis shows a more precise upload time for YouTube videos.
- Remember that publish date isn't recording date. A clip uploaded today may have been filmed years ago. The earliest upload is the latest possible recording date, nothing more.
- Look beyond video platforms. News articles, forum posts, and archived pages can predate the first upload you found. TinEye's oldest results and Google's "About this image" panel help here.
- Look at the uploader. An original uploader usually has the longest, highest-quality version, related footage from the same place, and no watermark from another account.
Our guide to finding an image's original source applies the same timeline method to photos.
Verifying a viral video: a checklist
Finding the source is only half the job. To judge whether a video shows what it claims, work through these questions:
- Is it old? Recycled footage from earlier disasters, protests, and conflicts is one of the most common forms of video misinformation. If an earlier copy exists, the new claim is wrong.
- Does the place match? Compare signs, architecture, vegetation, road markings, and the side of the road traffic drives on with the claimed location. Street-level imagery in mapping services helps confirm a match.
- Does the time match? Check the weather, season, and shadow direction against the claimed date. Historical weather records are available online.
- Is it edited? Look for cuts that skip key moments, audio that doesn't match the scene, and missing context at the start or end of the full original.
- Who is sharing it, and why? Accounts that post nothing but viral clips, or that profit from engagement, are less reliable than the person who filmed it.
- Have fact-checkers covered it? Search the claim with keywords plus "fact check." Google Lens results often include fact-check articles that used the same frames.
Video used in a scam? Fake videos of celebrities and executives promoting investments, or video "proof" sent by an online romance contact, are common. If a video is being used to ask you for money, stop and read our romance scam help guide before responding.
Deepfakes and AI-generated video
Reverse search works best on real footage that has been reused. A newly generated AI video may have no earlier copies at all, so a search that finds nothing proves nothing. In a December 2024 public service announcement, the FBI warned that criminals use AI-generated video to impersonate executives, officials, and public figures, including in real-time video calls and fake investment promotions.
Signs worth checking (none of them proof on its own):
- Distorted hands, teeth, eyes, or jewelry, especially when moving
- Faces that blur or shift at the edges, or blink unnaturally
- Lip movement that doesn't match the audio, or a voice with flat emotion
- Text in signs or on clothing that turns into gibberish
- Lighting and shadows that don't match the scene
As AI video improves, visual tells become less reliable. Detection tools, including those in the InVID-WeVerify plugin, can produce false positives and false negatives, so treat them as one input. Rely more on context: who first posted the video, whether trusted outlets confirm the event, and whether other footage from different angles exists. Our guide to spotting AI-generated images covers the same principles for still frames.
Frequently asked questions
Can I upload a video file to Google to search it?
No. Google's search by image accepts still images. Extract a frame first, or pause the video in Chrome and use Search with Google Lens to select part of the frame.
How do I reverse search a TikTok or Instagram Reel?
Pause the clip and take screenshots of distinctive moments, cropping out the app's buttons and captions. Search those in Google Lens and Bing. Watermarks showing another account's username are a strong clue that the clip was re-uploaded from someone else.
Is there a free tool that does this automatically?
The InVID-WeVerify browser plugin is free and comes closest. It extracts keyframes from a link or uploaded file and lets you send each frame to several search engines with a right-click. You still need to review the results yourself.
Why can't I find any earlier copies of a viral video?
The clip may be genuinely new, recorded privately, shared only in closed apps, or AI-generated. Try more frames and engines, search keywords from the claimed event, and check fact-checkers. A lack of matches is not evidence that it's authentic.
Sources & further reading
- Google Search Help: Search with an image on Google
- Google Search Help: Learn more about an image (About this image)
- VideoLAN wiki: Take a snapshot in VLC
- InVID project: Verification plugin
- TinEye Help: What kinds of images can I search?
- FBI IC3: Criminals use generative AI to facilitate financial fraud (2024)
Spotted an error or something out of date? Tell us — we review corrections promptly. Read how we research and update guides in our editorial policy.