VeriPLUS

AI-powered torrent verification, born right here on YourBittorrent. VeriPLUS scores every torrent against millions of known-good and known-fake examples — catching fakes and confirming the real ones automatically, so you download with confidence.

Verifying the torrent scene since 2008
480,000+Fakes caught
8,539,862Verified on YourBittorrent
12,911,675Torrents indexed
2008Verifying since
The background

In 2008 an important step towards a cleaner torrent scene was taken. The idea was simple: flag every fake torrent and confirm every real one on a massive scale, without relying on human votes. It went into action on the 15th of March 2008, when the owner of YourBittorrent set out to build something that would change the face of the torrent scene. Six months later the first beta launched on this very site — and the rest, as they say, is history.

That early system was rule-based, and by the end of 2009 its beta run had already flagged hundreds of thousands of fakes and confirmed tens of thousands of genuine torrents. Those results proved the idea worked — and set the direction for what VeriPLUS would grow into.

VeriPLUS beta results · 2009–2010
505,731Fakes flagged (beta)
90,543Torrents verified (beta)
89%Fake-detection rate
94%Real-torrent rate
How it works today

VeriPLUS has come a long way from those first rules. Today it runs on a machine-learning model trained on millions of labelled torrents. Rather than relying on any single trick, it weighs dozens of signals together to score how likely a torrent is to be fake — and it keeps learning as new fakes appear. When a torrent passes, you see the green Verified badge next to it.

1
Reputation & cross-referencing

Each torrent's infohash is checked against known-good and known-fake records across our partner sites and public sources — a strong first signal of whether it can be trusted.

2
Metadata & naming patterns

The model has learned the file names, file-lists and size patterns that fakes tend to share. Torrents that match those patterns are scored accordingly.

3
Swarm & behaviour signals

Seed/peer behaviour, tracker responses and user reports feed a final confidence score. Low-confidence cases are held as “unknown” for re-check — VeriPLUS never auto-verifies a torrent it is unsure about.

The model retrains continuously, so detection improves over time. Verified torrents from other sources are re-scored the same way — if one turns out to be fake, VeriPLUS strips its “verified” status.

The original 2010 benchmark

For historic interest, here is how the early VeriPLUS engine measured up against the alternatives of the day — the numbers that first proved automated verification could beat manual and vote-based methods.

Fake-torrent detection — success rate
VeriPLUS
92%
Manual verification
68%
Vote verification
55%
Torrents processed & verified
VeriPLUS
580k
Vertor
250k
ThePirateBay
120k
TorrentReactor
60k
Historic data: February 2010. Kept for reference — today's model processes far larger volumes.
FAQ
Why is VeriPLUS better than voting or manual verification?
  • It is automated and fast — it can score huge batches of torrents without waiting on people.
  • It is consistent: no human error, and no vote manipulation by uploaders trying to push a fake.
  • It improves over time as the model retrains on newly discovered fakes.
  • When it is not confident, it marks a torrent as “unknown” and re-checks later rather than guessing — so it never rubber-stamps a fake.
Is VeriPLUS ever wrong?

No automated system is perfect, and VeriPLUS is honest about that. When the model isn't confident either way, the torrent stays “unknown” instead of being verified, and it is re-scored as the model learns. Verification runs entirely on our servers using metadata and swarm signals — it doesn't change anything about your own download.

Available on