An experiment reported by Search Engine Land found that revised YouTube video descriptions significantly increase the likelihood of a video appearing in answers from Google's generative features. Videos with updated descriptions outperformed those with unchanged descriptions in both AI Overviews and AI Mode.
The key metric was the rise in AI retrievals — cases where a video's content is pulled in by the generative system to build an answer. For the updated videos, this figure rose 359% compared with the control group.
The results suggest that video descriptions are no longer a secondary optimization element. For generative search systems, the description text is the primary source through which the model understands what a video contains and decides whether it is appropriate to cite it in an answer.
The practical takeaway for creators and publishers: revising descriptions is a cheap and fast way to increase the visibility of video content in AI results. The experiment highlights the gap between videos with carefully structured, substantive descriptions and those where the description has remained formal or outdated.
As generative answers are increasingly assembled from extracted fragments rather than traditional link ranking, the quality and accuracy of descriptions become a factor directly affecting the share of traffic coming from Google's AI surfaces.
