Likes are the cheapest possible signal. They cost a viewer nothing and mean almost nothing, which is why every major recommendation system has been weighting them down for years.
What carries weight instead
Saves indicate the viewer expects to need this again. That is a strong statement about value, and it is hard to fake.
Shares, particularly direct shares to individuals, are the strongest signal available. Sending something to a specific person is a social risk — you are staking a little of your judgement on it.
Watch completion and replays on video tell the system the content held attention, which is the thing the platform is actually optimising for.
Comment threads, meaning replies to replies, count for much more than single-word comments.
What this changes about content
Content built for likes is pretty. Content built for saves is useful — reference material, checklists, methods, things with a second viewing. Content built for shares is either genuinely funny or says a thing the sharer wanted said on their behalf.
Those are different briefs, and the second two are more work.
Tagging follows the same logic
Tags describing the useful thing inside a post — a framework, a comparison, a before and after — attract the audience most likely to save. Tags describing an aesthetic attract scrollers. Both have their place, but only one compounds.