Explanation
This is the reason captions, titles and spoken words matter more than they did. These platforms transcribe audio and index on-screen text, so a video that never names its subject in words is invisible to the query traffic even if it performed well in the feed.
It also changes the value of evergreen material. A video that answers a specific question can accumulate views for a year through search, which is a completely different lifecycle from the two-day spike a feed hit produces.
The common mistake
Writing captions purely for the reader. The first line is for a person and the rest is for a query, and treating the whole caption as copy leaves the search half unwritten.