A popular video does not stay where it was posted
A single video can be reposted on several platforms, uploaded under a different title, cropped to change its look, or edited before it goes up again. Finding those copies by hand is slow, and the easy shortcuts get it wrong.
Searching by title misses uploads with new titles. Searching by thumbnail returns videos that look alike but contain different footage. To see how far one video travels, we ran Revideo Deep Search on MrBeast's Don't Pop the Balloon, published on 10 July 2026.
The shared Revideo report recorded 124 confirmed copies out of 136 matches checked, across YouTube, Instagram, TikTok, Facebook and web sources. It kept 12 similar-looking results out of the copy count, because looking similar is not the same as being a copy.
- Original video
- Don't Pop the Balloon
- Creator
- MrBeast
- Published
- 10 July 2026
- Tool used
- Revideo Deep Search
- Matches checked
- 136
- Confirmed copies
- 124
- Similar, excluded
- 12
- Sources
- YouTube, Instagram, TikTok, Facebook, web
Fifty-one seconds, one giant balloon
The investigation started with Don't Pop the Balloon, a vertical Short published on MrBeast's official YouTube channel on 10 July 2026. Three people, a field, and a balloon that keeps growing overhead. The scenes are distinctive, which made them a clean reference to compare every candidate against.
We wanted more than exact reposts. The goal was to find every place the same footage appeared, including versions with a different frame, title, caption or soundtrack, and to see which results the evidence actually supported.
Reuploads are rarely exact duplicates
People who repost a video often change it just enough to slip past an ordinary search.
A search on the original title will not find a repost with a new one. A thumbnail match alone does not prove the video itself is the same. For creators, publishers and content-protection teams, the real job is to find possible copies, compare each one with the original, and collect enough evidence to act. That is what Revideo Deep Search is built for.
Two stages: find everything, then prove it
Retrieval. Revideo fingerprinted the video and searched Google, Bing and Yandex, then YouTube, TikTok, Instagram, Facebook and more than 100 other platforms. Anything that might be a match became a candidate.
Verification. Each candidate was then checked against the original using up to six signals. The report shows, for every result, which checks matched, which did not, and which could not run.
THUMBranThumbnail
Is the cover image a real frame of the original?
VISranVisual matching
The footage itself, compared every 2 seconds.
AUDranAudio fingerprint
Is it the same recording, not just similar sound?
METAn/aMetadata
Titles and descriptions.
TXTn/aSubtitles
Spoken words and subtitle tracks.
OCRn/aOn-screen text
Captions, watermarks and labels read from frames.
For this video, three signals carried the verification: thumbnail, visual and audio. The other three were not applicable to this search. The original has no subtitles or on-screen text to compare, so the copies were confirmed on picture and sound, the two signals hardest to fake.
124 confirmed copies out of 136
The 91.2% figure describes this one report. It is not a general accuracy rate for Revideo across all videos.
Where the matches were found
YouTube Shorts carried most of the candidates, but the video had also spread to Instagram Reels, TikTok, Facebook and the open web. Someone checking only the original platform would have missed more than a quarter of them.
Counts cover all 136 candidates, the 124 confirmed copies and the 12 similar results together.
How fast it spread
- MrBeast publishes the original Short.
- The earliest copy in the report appears on YouTube, about a day later.
- New copies keep going up for weeks. Results in the report carry dates through July, August and September.
- One Deep Search finds and verifies all of them.
A one-time manual check in July would have missed most of these. Reuploads are not a single spike, which is why ongoing monitoring matters as much as the first search.
What individual results show
Each result card shows the source, title, uploader, thumbnail, the six verification signals and a confidence score. Many results show 90% confidence based on three of six signals. Others match on fewer. The score and signals guide review; they are not, on their own, proof of legal infringement.
Same footage, different packaging
Copied footage stays recognisable even when everything around it changes. Here is what we saw in this report, and what Revideo is built to catch.
Unchanged reposts
The full video uploaded again, frame for frame, often under a new title. Strong visual and audio matches confirm these even when nothing else matches.
Cropped and reframed
The vertical clip zoomed, cropped, or placed inside split-screen and side-panel layouts. Comparing many frames, not just the cover, keeps these from slipping through.
New titles and overlays
Retitled with #mrbeast tags, bare dates or prize claims, with new covers and caption text. Title search alone would miss these entirely.
Dubbed versions
New audio over the original picture. Visual matching still finds the footage when the soundtrack has changed.
No single signal handles every kind of change equally well. The value of Deep Search is that it runs several comparisons in one investigation, so a candidate is judged on its footage and sound, not its title or thumbnail.
What "confirmed copy" actually means
Every confirmed copy opens into an evidence panel. Here is one, a YouTube Short whose title is just a date: July 13, 2026. Nothing in its title points to MrBeast. The evidence does.
Why the misses matter as much as the hits
Revideo did not call all 136 candidates copies. Twelve were listed as similar-looking and kept out of the count.
Two videos can share a giant balloon and the same faces and still be different recordings. Counting every lookalike as a copy creates false positives, wastes investigators' time, and can damage a creator's credibility if a report is filed against the wrong video. Separating the two makes the result something you can act on. Individual results may still need human review before any claim is made.
From a pile of links to an investigation
Cross-platform discovery
Candidates from YouTube, Instagram, TikTok, Facebook and the web in one report, instead of a separate manual search on each.
Beyond titles and thumbnails
Matches are judged on the footage and the sound, so a new title or cover does not hide a copy.
Match-level evidence
Source URLs, previews, signals, confidence, side-by-side playback and audio fingerprints for every result.
A structured review
Copies first, lookalikes separate, filters by platform, so you know which cases deserve follow-up.
Ready for protection workflows
Share the report, keep watching the video for new uploads, and use the evidence when contacting an uploader, checking licensing, or filing a platform report.
Finding a copy does not remove it or establish that it is unlawful. It gives you the evidence to decide what to do next.
Not just for the biggest channel on YouTube
This video came from a high-profile creator, but the same problem hits anyone whose videos travel. If one Short from MrBeast was copied 124 times, smaller creators are being copied too, with far less chance anyone notices.
What the evidence supports
This case study is based on a Revideo Deep Search run on 13 October 2026 and shared on 14 October 2026. The report recorded 136 checked matches: 124 confirmed copies and 12 similar-looking results, across YouTube, Instagram, TikTok, Facebook and web sources.
Limitations of this investigation
- Results reflect the search at the time it ran. Videos may since have been removed, made private, or changed.
- The report documents matches. It does not establish permission or licensing status for any individual result.
- Platform counts cover all 136 candidates, not only the 124 confirmed copies.
- Modification types are described from results we reviewed, not assigned to every result.
- We did not measure financial loss, lost revenue, or whether any takedown took place.
- A match does not automatically mean copyright infringement. Authorisation, licensing, context and local law all affect how a result should be treated.
Finding copies is the first step
One Deep Search recorded 124 confirmed copies of a video MrBeast published, out of 136 matches checked, across five source types. It showed the evidence for each, and it kept 12 lookalikes out of the count.
Copies hide behind new titles, new platforms and new framing. Matching the footage and the sound is how you find them anyway.
For creators and teams who work with video, that discovery is where monitoring, evidence collection and next steps begin.