Hashtag Trend Radar — Rising & Emergent Tags

Most hashtag tools hand you a trending list. This one tells you which tags are actually rising, and separates a network-wide trend from one server's local noise.

Every tag arrives with seven days of daily uses and accounts, so the rise is computed, not guessed:

Field	What it answers
spikeRatio	today's uses ÷ the previous six days' mean
isEmergent	zero uses in the previous six days — a tag that did not exist until now
trend	emergent / rising / steady-high / peaked / declining
peakDayIndex	0 means still at the top; higher means it already peaked
instanceCount / isCrossInstance	how many servers saw it — the network-vs-local test
historyUses / historyAccounts	the whole daily series, newest first
usesPerAccountToday	high means a few accounts posting a lot, not broad adoption

Two dataset views: Rising now (ranked, emergent first) and Daily series (the shapes).

The one thing a seven-day window cannot do

The source gives seven days of history and no more. That means weekly ritual tags look exactly like trends. Measured on 14 of the largest servers:

Tag	Daily uses (newest first)	What it really is
#caturday	852, 7188, 626, 285, 431, 252, 313	every Saturday
#silentsunday	6762, 216, 96, 79, 66, 51, 631	every Sunday
#fragrantplaces	2824, 0, 0, 0, 0, 0, 0	genuinely new
#curiousabook	25, 2691, 0, 0, 0, 0, 0	appeared yesterday

A spike ratio alone promotes #caturday every weekend. isEmergent is the one signal a ritual cannot fake, because a recurring tag never has an all-zero history. If you want "risen in the last few days", filter on emergent.

Spike ratios need an absolute floor — and the floor is smaller than you think

A tag going from 1 use to 20 has a spike ratio of 20 and means nothing. So pair the ratio with minPeakUses, minUsesToday or minAccountsToday.

But mind the scale. Measured across 14 of the largest servers:

	single server	aggregated
median daily uses	7	23
p90	48	189
p99	—	2,691
max	666	7,188

Of 58 tags: 24 cleared a 100 peak, 12 cleared 200, 7 cleared 500, and none cleared 10,000. The whole network has roughly 3 million users, so an Instagram-scale cutoff returns an empty result rather than "the big trends". For genuinely large trends, try minPeakUses between 200 and 500 — and every run reports observedScale with the distribution it actually saw, so you can calibrate against live data instead of these numbers.

minAccountsToday is the harder floor to game: one account can post a tag many times, but it cannot be many accounts.

Cross-server confirmation

The network is federated, and each server shows trends from its own vantage point. Measured across 9 servers: 46 distinct tags, 3 appeared on all 9, and 14 appeared on only one. A tag on one server is a local phenomenon; isCrossInstance marks three or more.

Pick servers by size (topInstances), or by instanceLanguage / instanceCategory to watch a specific language or community — the official directory lists 298 servers, 175 with user counts totalling about 3 million.

Tracking your own tags

trackTags follows specific hashtags whether or not a server has them trending. Those go through the single-tag endpoint, so they are not limited to each server's curated list — a trending list is partly editorial, and "not on the list" is not the same as "nobody is using it".

How it checks itself

Check	What it catches
Too many servers failed	Below half responding, the aggregate no longer represents the network, so the run stops rather than presenting a thin sample as a network trend.
Servers answered but no tags	The trends endpoint changed shape. Reported as a structural problem, never as "no trends today".
History shape changed	The seven-day series is the denominator of every rise metric. A tag whose history loses its shape is skipped; if the directory itself is unavailable the run stops.
Unknown trend filter	Rejected rather than silently ignored.

Notes on the data

• Aggregation sums each server's daily figures. Federated views overlap, so the totals are an upper bound, not a de-duplicated count. Compare tags against each other rather than treating the number as absolute volume.
• Trending lists are curated by each server, often with moderation.
• This is a trend-detection tool. It returns tags and counts, not post content.