We have invested hundreds of millions of dollars to develop powerful tools to help rightsholders manage their content on YouTube. We have also accumulated over a decade of learnings on how to best balance the needs of rightsholders and creators, and continue to iterate on our tools accordingly.
Content ID automatically detects reused content in user-uploaded videos. Some rightsholders can also manually search for and claim reused content that was not automatically detected. Over 99% of Content ID claims from bteween July 2023 and December 2023 were made through automated detection, with partners' usage of manual options being exceedingly rare (fewer than 0.5% of total claims).
When partners do make manual claims on videos not detected by our systems, uploaders are more likely to dispute the claim’s validity — fewer than 0.5% of automated claims from between July 2023 and December 2023 were disputed, compared to over 0.9% for manual claims.
Over time, we've also worked to safely expand automated detection technology to more users. All copyright holders can use our webform to ask us to try to automatically prevent copies of the removed content from being reuploaded. In the second half of 2023 our automated system prevented over 4 million videos from being uploaded because they copied content from videos removed due to copyright removal requests.
Whether it be fans making reaction videos, aspiring musicians singing covers, or pirates attempting to re-stream a championship soccer game, we understand that there are countless scenarios where rightsholders need a tool that is powerful enough to quickly detect reuse and apply the appropriate policy.
Unfortunately, some individual channels attempt to circumvent the system by transforming the content they upload. This can take many forms, from shifting the pitch of audio to altering the color, orientation, or surroundings of a video. Below are some examples of the types of manipulations channels engage in.
Content ID's match technology must continually adapt to such efforts, without sacrificing its precision. In other words, in the effort to locate transformed copies of a given work, it cannot extrapolate so far that it begins to inaccurately sweep in similar works.
Original
Flipped / Mirrored
Framed
Color shifted
Aspect ratio change
Light effects
Data Collection Note:
All data in this report is collected between July 2023 through December 2023, unless otherwise annotated as below:
† Data collected as a 'snapshot' in April 2024. Some data, like rate of counter notifications and Content ID disputes, look at "trailing" events. We start with the set of claims or removals made during H2 2023, but disputes and counter notifications continue to accrue after that period because they can be made at any time. For instance, a claim made on November 30, 2023 may have a dispute made on April 1, 2024. For these data points, we have chosen to take a snapshot 3 months after the end of H2 2023.
For previous versions of this report covering Jan 1, 2021 through Dec, 31 2022, refer to the PDF Download Center.