Skip to content
Current Zeitgeist

Scorecard

How our calls hold up

We check our own labels against what happened next, with the same data the site shows. Nothing is hand-picked: every country and every day counts.

110 days of history since Monday, 15 June 2026. Updated every week; this edition: Saturday, 3 October 2026. How the labels are set →


How long rising topics last

Most rising topics are short spikes: after 4 weeks, 10.3% are still in their country’s top 100, fewer than the 14.9% of all top topics.

Topics labelled rising in a country (not weekly or yearly repeats), checked in the same country later, next to every topic in that country’s top 100 the same day, everyday staples included.

AfterRising topics still in the top 100Still risingAll top-100 topics still there
1 week31.1% 113,484 of 364,96813.4%37.1%
4 weeks10.3% 30,468 of 295,7231.7%14.9%
12 weeks4.0% 4,360 of 108,8012.1%7.1%

By label, after 4 weeks

LabelRising topicsStill in the top 100
Breakout30,14116.9%
New68,0316.8%
Rising193,90610.5%
Back3,64514.0%

By kind of signal, after 4 weeks

SignalRising topicsStill in the top 100
Reading and searching232,7257.6%
Charts (apps, songs, films, games)60,18119.0%
Both2,81744.7%

By category, after 4 weeks

Categories with at least 500 rising topics.

CategoryRising topicsStill in the top 100All topics
Sports88,8329.0%9.2%
Movies & TV64,92114.2%13.3%
Gaming36,45319.8%28.5%
Music20,8924.6%6.2%
Politics & world affairs18,4677.2%13.7%
Travel & places10,1603.8%24.6%
Business & economy7,2356.4%15.9%
Celebrities & creators7,0293.2%4.2%
Books, arts & history4,5005.2%7.5%
Crime & justice3,0773.7%5.1%
Society & beliefs2,2512.4%2.6%
Tech & AI2,04935.4%48.4%
Nature, climate & disasters1,1016.1%8.2%
Fashion & beauty8802.3%3.6%
Science & space5932.7%4.3%

Where it’s heading next

Each month the model is tested on the last fifth of its history: of the countries it would have named (a chance of at least 20%), how many picked the topic up within a week, next to the share of all candidate countries that did.

Not yet better than chance. In its latest test the model named no country with enough confidence, and its chances were no more accurate than the base rate. Pro shows its guesses as experimental until that changes.

ModelCountries namedPicked it upAll candidatesSkill
30 Sept0––-0.002

Skill compares the model’s chances with guessing from how often each country usually picks topics up: above 0 is better, 0 or below is no better.