Hashtags Are a Search Index, Not a Megaphone

Findability and spread are different jobs

A hashtag is a normalised identifier a user types into a caption, which a platform then treats as a key: it collects every post carrying that key and serves them when somebody asks. Structurally it is very close to a URL slug — no spaces, no punctuation, case folded for matching, one canonical form per concept — and that resemblance is not a coincidence, because both exist so a machine can group things a human wrote loosely.

That is a retrieval mechanism. Spread is a recommendation mechanism, and it works the other way round: the system decides to show a post to someone who did not ask. Confusing the two is the single most common error in hashtag advice, and it produces the behaviour everyone recognises — thirty tags stacked under a post in the belief that each is an additional loudspeaker, when each is only an additional row in an index nobody queried.

How long a post stays visible under a tag

Model the recency-ordered view of a tag, where posts appear newest first. If the tag receives V posts an hour, and the first screen shows nine of them — a three-wide grid, which is the standard layout — then your post occupies that screen from the moment it is published until nine newer posts arrive. That is nine divided by V hours, and nothing else is needed to compute it.

The numbers are worse than intuition suggests. A tag receiving 100,000 posts an hour keeps you on the first screen for 0.32 seconds. At 10,000 an hour it is 3.24 seconds. At 1,000 an hour, 32.4 seconds. At 100 an hour you get 5.4 minutes, at 10 an hour 54 minutes, and at one an hour nine hours. The curve is a reciprocal, so the difference between a very large tag and a merely large one is enormous, while the difference between two small tags barely registers.

A tag's worth is a ratio, not a volume

Visibility time on its own is not the answer, because a bigger tag also has more people looking at it. Put both in: if R people arrive at the tag's recent view each hour and scroll the first screen, the expected number who see your post is R multiplied by the fraction of an hour you spend on that screen, which is R times nine divided by V. Simplify and the hours cancel: expected exposures equal nine times R over V — nine times the tag's viewers per post published.

That result is worth pausing on, because it says something unintuitive: the size of a tag drops out entirely. What matters is the ratio of demand to supply, of people looking to posts arriving. A head tag with 200,000 viewers an hour and 100,000 posts an hour has a ratio of 2 and returns 18 expected exposures. A mid-sized tag with 5,000 viewers and 500 posts has a ratio of 10 and returns 90. A niche tag with 300 viewers and 6 posts has a ratio of 50 and returns 450 — twenty-five times the head tag, from a tag with a fraction of a percent of its traffic.

The mix follows immediately. Put a head, a mid and a niche tag on one post and the expected exposures total 558, of which the head tag contributes 18 — 3.23% of the result for a third of the tags. Replace it with four more niche tags at the same ratio and the total jumps to over 2,000. That is the whole case for a portfolio weighted towards specificity, and it is arithmetic rather than opinion. The head tag is not free either: it is the one most likely to be crowded with unrelated content, which makes it the one most likely to teach a classifier the wrong thing about your post.

What platforms have actually said about counts

The folklore says thirty. That number is dead. On 18 December 2025 Instagram's @creators account announced that it would gradually reduce the number of hashtags allowed in a caption for a post or reel to five, on the stated grounds that fewer and more targeted tags improve both the content's performance and people's experience. Any guide still recommending a stack of thirty is describing a product that no longer exists.

The direction of travel is older than that announcement. In February 2025 Adam Mosseri said publicly that hashtags do not increase reach, while calling them a good way to let people know what a post is about — a description of an index, precisely. And in December 2024 Instagram removed the ability to follow a hashtag at all, which had been the one mechanism by which a tag genuinely pushed content into feeds rather than waiting to be queried. Three separate signals, all pointing the same way.

Other platforms differ and none of them publish much, so the safe operating rule is to look up the current limit on each platform rather than carry a number across from another one. What is portable is the reasoning: if a platform tells you fewer and more specific tags perform better, that is consistent with the ratio arithmetic above, and both are consistent with hashtags being retrieval rather than distribution.

Caption or comment, banned tags, and other folklore

The belief that hiding hashtags in the first comment performs better than putting them in the caption has circulated for a decade with no platform ever confirming it. It made aesthetic sense when captions were cluttered with thirty tags; it makes no mechanical sense as a ranking claim, and a five-tag limit removes the aesthetic argument too. Now that the count is small, put them in the caption where a reader can see what the post is about — which, on the platforms' own account, is what they are for.

Two real risks do exist. Some tags are blocked or restricted, usually because they have attracted content that violates policy, and putting one on a post can limit that post's distribution — check any unfamiliar tag by opening it before you use it, and treat an empty or warning-laden results page as a reason not to. And a tag that describes your post badly is worse than no tag at all, because the index entry it creates is wrong: it brings people who wanted something else, who then do not engage, which is a signal every ranking system reads.

One small accessibility point that costs nothing: capitalise the first letter of each word inside a multi-word tag. The platform matches case-insensitively so it changes no behaviour, and a screen reader announces the words separately instead of attempting the run-together string. It is the same discipline that makes a URL slug readable, applied to the other place a machine-normalised identifier gets shown to a human.

The branded tag is the exception, because you own the container

Everything above treats a hashtag as a query somebody else might run. A branded tag inverts that: it is a container you create, name and fill, and its value is not the traffic it intercepts but the collection it accumulates. Nobody searches it by accident, which is exactly the point — the arrivals are people who saw it on your packaging, in your campaign, or on another customer's post, and who came deliberately.

The ratio arithmetic makes the case starkly. Model a branded tag receiving one post every two hours from 120 viewers an hour: the ratio is 240 and the expected exposures per post are 2,160, against 18 for a head tag. That is not because the branded tag is big; it is because nothing else is competing for its screen. A container with almost no supply and a small, deliberate demand is the most efficient shape a tag can have, and it is the one shape you can create yourself rather than compete for.

Two conditions make it work. The name has to be unambiguous enough that nobody else fills it accidentally — check it before you print it on anything, because a tag you already share with an unrelated community cannot be reclaimed. And there has to be a reason for anybody but you to post under it, which usually means a customer's own content has somewhere better to go than a mention that disappears. Get those right and you own a small, permanent, self-filling index. Get them wrong and you own a tag nobody types.

Frequently asked questions

Do hashtags increase reach?

Instagram's own head said they do not. In February 2025 Adam Mosseri stated publicly that hashtags do not increase reach, describing them instead as a good way to let people know what a post is about. Instagram had already removed the ability to follow a hashtag in December 2024, which was the one mechanism by which a tag pushed content into feeds without being queried. What a hashtag buys is retrieval: it makes the post findable to somebody who runs that query, and the expected number of those people is computable — nine times the tag's viewers per post published, on a recency-ordered view showing nine posts a screen.

How many hashtags should I use?

Check the current limit on each platform, because the number everyone repeats is out of date. On 18 December 2025 Instagram's @creators account announced it would gradually limit captions on posts and reels to five hashtags, stating that fewer and more targeted tags improve both performance and the reader's experience. The thirty-tag advice that still circulates describes a product that no longer exists. The arithmetic agrees with the platform anyway: since a tag's return is its viewers per post published, five well-chosen specific tags beat thirty generic ones by a wide margin.

Are popular hashtags better than small ones?

Almost always worse, and the model shows why. On a recency-ordered tag view showing nine posts a screen, a post stays on the first screen for nine divided by the tag's posts per hour: 0.32 seconds at 100,000 posts an hour, against 5.4 minutes at 100. Multiply by hourly viewers and the expected exposures reduce to nine times the tag's viewers per post published — a ratio, not a size. A head tag with 200,000 viewers and 100,000 posts an hour returns 18 exposures; a niche tag with 300 viewers and 6 posts returns 450, twenty-five times more. In a three-tag mix totalling 558 expected exposures, the head tag contributes 3.23%.

Should I put hashtags in the caption or in the first comment?

In the caption. No platform has ever confirmed that the comment placement performs better, and the belief survives mainly because it once solved an aesthetic problem: thirty tags made a caption unreadable. With a five-tag limit that problem is gone, and the tags belong where a reader can see what the post is about — which is what the platforms themselves say hashtags are for. While you are there, capitalise the first letter of each word inside a multi-word tag: matching is case-insensitive so nothing changes mechanically, and a screen reader will announce the words separately instead of one run-together string.

Is a branded hashtag worth creating?

Yes, because it is the one tag whose supply you control. Model a branded tag with 120 viewers an hour and one post every two hours: the viewers-per-post ratio is 240 and the expected exposures per post are 2,160, against 18 for a head tag. Nothing about that comes from size; it comes from having almost no competition for the screen. Two conditions make it work — a name unambiguous enough that no unrelated community fills it, which you must verify before printing it anywhere, and a reason for customers to post under it. Fail either and you own a tag nobody types.