Do hashtags still do anything?
Much less than they used to, and what they do now varies enormously by platform — because the problem they were invented to solve has largely been solved by other means.
What hashtags were for. In an era of chronological feeds and weak search, a hashtag was a user-invented index: a way to join a conversation, follow an event, or make a post findable when the platform could not understand what the post was about.
Why they matter less now:
Platforms understand content directly. Recommendation systems analyse the text, audio, images and engagement patterns of a post, and infer its topic without being told. The hashtag is redundant as a classification signal.
Feeds are recommendation-driven, so distribution comes from the system deciding who would engage, not from people browsing a tag.
Search improved, and on several platforms searching plain keywords now works better than searching tags.
Spam degraded them. Once posting thirty irrelevant popular tags became standard, tags stopped being a reliable signal of topic and platforms discounted them accordingly.
Where they still genuinely work:
Live events, where a shared tag aggregates real-time commentary that no algorithm groups as well.
Communities and niches that have adopted a specific tag as an identifier.
Professional and interest-based platforms, where users follow tags deliberately.
Campaigns and movements, where the tag is the name of the thing.
Accessibility and categorisation in some contexts, and internal search within your own account's history.
Decentralised platforms, where there is no recommendation algorithm at all and tags are one of the few discovery mechanisms available — they matter considerably more there.
What actually drives reach now: early engagement, watch time or dwell time, the relevance the system infers, and whether people send the post to someone else — which is weighted heavily on several platforms.
Practical guidance: a few genuinely relevant tags, not thirty; check what the platform's own guidance says, since several now recommend far fewer than they once did; and treat tags as description rather than as distribution.