What is a hashtag, and what are hashtags used for?

In a caption or post, the platform recognises a valid # token and turns it into a link. Clicking that link opens the platform’s results for the tag, rather than searching the entire web. #UrbanGardening is one tag; #Urban Gardening normally tags only Urban because the space ends the token. Exact punctuation and character rules vary by platform.

A hashtag can contain a keyword, but it is not the same thing as a keyword. A keyword is a term someone searches for or a phrase describing a subject. A hashtag is a marked-up label with a platform-specific destination. Writing ‘container gardening’ naturally in a caption explains the subject even when no hashtag is present.

Use capital letters to separate words in multiword tags, such as #AccessibleTravel. This improves readability and can help screen readers interpret the words. Do not assume that changing capitalisation creates an entirely separate discovery topic; platform matching rules determine that.

There is no single hashtag algorithm. Tagged search, feed recommendations and trending conversations are different surfaces. A tag can identify the subject without making a platform recommend the post to additional people.

Distinguish a hard limit from a recommendation and an editorial starting point. A limit says what the product accepts. A recommendation is platform guidance, not a controlled experiment proving an optimum. A starting point is something to test with your audience. None establishes a universal number that increases reach.

Instagram’s December 2025 announcement said captions for posts and Reels would gradually be limited to five hashtags. Social Media Today reproduced the announcement and its advice to use fewer, targeted tags. That is reported platform guidance, not our own reach study; account interfaces can reflect staged rollouts. Do not build a current strategy around the former 30-tag allowance.

A discovery hashtag names a subject that people might explore, such as #ContainerGardening. A branded hashtag names a business, product or campaign. Its primary purpose is to collect related contributions, make a campaign recognisable or help an existing audience find posts. A new branded tag has no built-in audience just because it is unique.

For a fictional community project, #RooftopGardenDay could collect participant photos while #UrbanGardening describes the wider topic. One post can use both. Check existing uses before adopting a campaign tag: a seemingly original phrase may already mean something different. Do not claim ownership of a generic topic label.

Stuffing is adding many repetitive, generic or unrelated tags to chase visibility. On YouTube there is a documented consequence: more than 60 hashtags causes all of them to be ignored, and over-tagging or misleading tags can lead to removal from uploads or search. Other platforms have their own product limits and policies; do not transfer YouTube’s threshold to them.

Outside an explicit policy, ‘every extra tag triggers a reach penalty’ is not established. The more defensible concern is relevance: an unrelated label can attract the wrong viewers, obscure the caption and make the post look indiscriminate. If those viewers skip it, weak engagement is a plausible consequence, not proof of a universal hashtag penalty or a shadowban.

A gardening tutorial tagged with unrelated celebrity names is not better described. A concise caption saying which plants, climate and container size the tutorial covers gives people useful context. Add tags that match that context rather than trying to substitute metadata for the answer.

Their role as a dependable reach lever has diminished on several recommendation-led platforms. Instagram’s move to fewer, targeted tags and TikTok’s emphasis on personalised viewing signals are consistent with that shift. They do not establish a comparable percentage decline across platforms, and we have no cross-platform experiment that would justify one.

A tag can still help with a niche topic, a live conversation or a campaign collection. What it cannot do is guarantee distribution when viewers do not choose or enjoy the content. ‘Hashtags are dead’ is as unhelpful as ‘more hashtags means more reach’: both collapse distinct search, labelling and recommendation functions into one claim.

To test usefulness, compare similar posts across several publishing cycles, hold the topic and format as steady as possible, and separate search visits from recommendation traffic where analytics allows it. Watch audience relevance, not just total views. This is an observational test: topic interest, timing and content quality still confound the result.

Inside a platform, a hashtag can create a direct route to related posts. Broader search may also use caption text, titles and other content signals; a person searching ‘balcony tomato care’ does not necessarily type a hashtag. On Google, adding a hashtag is not a documented general-purpose ranking boost. Google may index accessible public platform pages, but crawlability and the page’s content still matter.

Where this matters for search: distinguish topic labels from measured discovery. The relevant service reference covers social media publishing and per-platform reporting; website crawlability and indexation belong to the separate SEO discipline.

For the web-search side, see the explanation of crawlability, indexation and search intent. A social hashtag is not a substitute for an accessible page answering the query.

YouTube Help, ‘Find playlists & videos using hashtags’, documents hashtag result links, the up-to-three title display and the more-than-60 enforcement rule. TikTok Support, ‘How TikTok recommends content’, describes personalised recommendation signals. LinkedIn Help, ‘Post and share content on LinkedIn’, is the ordinary-post reference; it does not establish the small-count test suggested here.

Social Media Today, ‘Instagram Implements New Limits on Hashtag Use’ (December 2025), quotes Instagram’s five-tag announcement. That evidence is secondary reporting of a platform statement. The suggested TikTok, LinkedIn and X test counts are editorial starting points, not documented optimal counts or findings from a DigeeSell experiment.