3 Comments

  • Update Feb. 2021: We've removed these test videos from YouTube to get ready for some additional experiments we'll be running. We have some other tests on different YouTube accounts still going so we can track any changes in the future.

    If you have a question about this experiment, or an idea for additional experiments, I'd love to hear from you below...
  • Thank you. Clever experiment!

    Did you do further tests? Could tags maybe help with suggested videos? Or be indexed later?
    • Alberto,

      Thanks so much for the comment.

      We've done some further tests, and we're getting the same result. We've had one test running for a very long time, and the tags haven't been indexed yet.

      Now, regarding suggested videos...it would be possible for YouTube to have different indexes, and be using tags in that index as a way to find suggested videos. There is some documentation from YouTube engineers that this was one of the original uses for tags.

      Having said that, I think that YouTube is using user signals now, instead of tags, to determine which videos to suggest. For example, they can see which videos users watch together, and from the average view duration, determine if they are a good suggestion match.

      This is all possible because of the power of computing, A.I., and the size and resources that YouTube/Google has to work with.

      And if you think about it, it's much harder to manipulate these user signals than it is to simple copy someone else's video tags. So it makes sense that they will adjust the ranking algorithm to make it more robust.

Reply