A year of push notifications, from 336 news apps at once.
Six phones on a shelf kept every alert switched on while news apps from 44 countries buzzed them in 31 languages. This is what landed.
20× · an average day
-
SVT Nyheter
Sweden
RIGHT NOW · Two people dead in workplace accident in Hallsberg
-
SZ
Germany
BREAKING - Louvre Director Laurence des Cars resigns
-
28 Feb 06:16 UTC 121 apps in 27 countries
Israel and the US strike Iran · Just after 06:00 UTC, Israel announced a "preemptive strike" on Iran with American participation.
-
Berner Zeitung
Switzerland
Grandma's dishes and Southern US cuisine: These are the newest restaurants and bars in Bern.
-
Göteborgs-Posten
Sweden
Unknown images from the riots in Gothenburg • Niklas Maupoix was in the middle of the chaos when the police shot
-
Bergens Tidende
Norway
Sports · Information to BT: Lysekloster players have not received their salaries. Mohn support may end.
-
20 Mar 13:59 UTC 119 apps in 24 countries
Chuck Norris dies at 86 · TMZ reported the actor's death after a medical emergency in Hawaii and his family confirmed it within minutes.
-
ARY News PK
Dollar and Other Currency Rates in Pakistan Today - Nov. 12, 2025
-
Lrytas LT
Former Finance Minister K. Vaitiekūnas answered what he thinks about his successor T. Valis
-
StN News
Germany
Breaking News · 2026 FIFA World Cup: These are the opponents Germany will play against in the group stage
-
12 Apr 04:38 UTC 110 apps in 22 countries
Orbán loses Hungary's election · Hungarians voted in what European outlets billed as the year's most important EU election.
-
Sydsvenskan
Sweden
The pharmaceutical factory in Limhamn · Chemists played a trick to fool the inspectors – hid chemicals • “They won't see how messy it is.”
Every card is a notification a phone actually received, translated into English. Twelve of them, arriving at the average pace of the year. Hold the cursor over a card to stop the stream. A red rail marks a story cluster, one event that 110 or more apps pushed, and that line is our summary of the cluster, not any one app’s wording.
Everything below comes from the same file. First, how it was collected.
How the notifications were collected
Every notification in this dataset was recorded from a phone that received it. Four steps take it from the lock screen to a row in the file.
6 phones with 336 news apps installed and every alert switched on. Everything in this dataset was collected from them.
Four fields the moment an alert lands: app, title, body, timestamp, to the second. That timestamp is delivery, not publication, so every clock on this page is a clock on when a pocket buzzed.
31 languages in, English out, the original kept beside the translation. 77.6% of the year arrived in something other than English, so most of what you will read quoted here is a machine's English rather than the newsroom's own words.
Pushes about the same event become one story. 38,794 stories drew alerts from two or more apps; 1,846 drew ten or more.
- Period
- 1 Aug 2025 to 31 Jul 2026 · 365 days
- Notifications
- 537,376 · counted at the moment of delivery
- Apps
- 336 · 98 live on 180 days or more
- Countries
- 44 · the app's country, not the reader's
- Languages
- 31 · counted by the language the push arrived in
- Archive
- 4,815,066 · from 713 apps since March 2021; this report is one year of it
The next question is which apps do the pushing, and they are nothing like each other.
Who pushes most
The loudest app in the dataset, RaiNews.it, sent 59.7 pushes a day for a year. The median app that ran most of the year sends fewer than 10.
21,782 notifications in twelve months puts the Italian public broadcaster's news app at the top of the list. Second place, Breaking News, is an unbranded aggregator with a generic package id; it stays in the count because a real app interrupted a real phone 17,763 times. And the Nordics are over-represented all the way down: eight of the twenty loudest apps are Nordic, marked in green below.
Where would your desk sit?
Across the 98 apps that were live on at least 180 days, counted only on the days they were live, the median sends 9.7 pushes a day. A quarter send fewer than 5.9. One in ten sends more than 26.0, a different denominator from the chart above, so it gets its own axis.
Put your own number in.
A tenth of the field sends fewer than 3.9 a day; a tenth sends more than 26.0, and the chart stops measuring there.
Drag the handle below, or drag straight on the shape.
Louder than 50% of the 98 apps live 180 days or more. That is exactly the median, 9.7 a day. About 49 of the 98 push more often; at your rate the phone buzzes every 148 minutes, day and night.
The full per-app distribution, with the working, is in how many pushes a news app sends per day.
The loudest country is not the United States
An average Swedish news app sends 10.3 alerts a day, 3.1 times its American counterpart. Denmark and Norway come next. Divide each country's pushes by the number of its apps we carry and by the days in the year, and the United States lands 14th of 19 on intensity while sitting 2nd on raw volume. Volume follows the roster: we carry 53 American apps and 19 Swedish ones. Switch the chart to total pushes and watch the flags rearrange.
Volume is one question, and the easy one. The harder question is when newsrooms send, and there the countries that look alike on this chart keep different hours.
When each country pushes
Put twelve months of alerts on each country's own clock and the middles of the day nearly touch: 14 of the 16 pass their halfway mark, the hour by which half of a country's year has landed, inside a window of 69 minutes, between 13:43 and 14:52.
Every notification is timestamped in UTC. Here each country's traffic is moved onto one representative time zone, Rome time for Italy, New York time for the United States, and counted as a share of that country's own year, so Sweden's 71,780 alerts and New Zealand's 7,011 sit on the same scale. A perfectly flat day would put 4.2% of the year in every hour; the hairline on every card below marks it.
The same day, sixteen ways
One distribution per country, each scaled to its own busiest hour, so the shapes are comparable and the heights are not. The hairline across every card is a flat day at 4.2%; the tick under the axis marks local night, midnight to 06:00.
Italypeak07:00Switzerlandpeak07:00
Finlandpeak07:00
NZpeak15:00
Brazilpeak11:00Swedenpeak19:00Germanypeak16:00
USpeak12:00
Denmarkpeak07:00
Spainpeak09:00
Francepeak08:00Norwaypeak10:00
Belgiumpeak12:00
Indiapeak15:00
UKpeak11:00
Israelpeak16:00Each card is scaled to its own busiest hour, so a tall bar means "busy for this country", not "busy compared with the next card". The flat-day hairline is the one thing that is the same on all 16.
Where the markets differ
The middle of the day is the same everywhere. The start of it is not, though it is closer than you would guess: every one of the 16 working days is above a flat day by 08:00, and six of them already are at 06:00. Italian desks put 7.2% of their year into the 06:00 hour; Spanish ones put 0.2% into it and do not cross a flat day until 08:00. Two hours is the whole spread.
The evening end spreads the same two hours: six desks are back under a flat day after 20:00, and three are still above it at 22:00, Norway, Denmark and the United Kingdom. Between those two edges sits a working day that no market has any particular claim on.
Peak hours are a weaker statistic than they look. Four countries push hardest at 07:00, Italy, Switzerland, Finland and Denmark, while France peaks at 08:00, Spain at 09:00, the United Kingdom at 11:00, the United States and Belgium at 12:00. But eight of the 16 have a second hour within a quarter of a point of their busiest, which on these volumes is a rounding difference; only three, Italy, India and Spain, win their own day by more than half a point.
One country's busiest hour falls after 16:00: Sweden's, at 19:00. Swedish apps do not have an evening peak so much as two peaks, 12 hours apart and the same height, 19:00 beats 07:00 by 0.02 points, about 14 alerts out of 71,780. Switzerland runs a quieter version of the same shape: its 17:00 is 98% of its 07:00.
The night is where the rows stop looking alike. 22.5% of Israeli alerts arrived between midnight and 06:00, in a year of war, and Israel is the flattest distribution of the sixteen by a distance: its busiest hour is 1.3× a flat day, its quietest 0.5×, a ratio of 2.6 where Italy's is 55. At 03:00 it is still running at 3.7% of its year, while nine of the 16 are under 0.1%. The United States is second at 9.6%, but read that with care, four American time zones are stacked on one clock here, so a 21:00 push in Los Angeles is counted as midnight in New York, and part of both the night share and the long American afternoon is geography rather than scheduling. Brazil, on three zones, carries the same caveat. Belgium carries none: 0.1% of its year lands before dawn, which is another way of saying its desks go home.
Weekends run quieter but not quiet, 1,566 pushes on an average weekday against 1,237 at the weekend, and everything above averages the two together. Each country also has its own deep dive, which splits them and marks the windows where a push faces the least competition.
When a market pushes is one house rule. How it writes the push is another, and the markets differ just as sharply there.
59 characters in Denmark, 120 in Italy
Two slots and a screen that stops after two lines: that is the whole form. 12.6% of the year does not use the title slot at all, which leaves the body doing all of the work. How much room it gets is a house rule, and the houses sort by country.
A lock screen gives a push three bands of attention: an app row the newsroom does not control, a title slot it often skips, and a body it can fill until the phone stops printing. Every house rule in this act is made inside those three.
The title slot is smaller than it looks. 12.6% of the year carries no title at all: the sentence arrives with nothing above it. And when the slot is filled, it usually holds a label rather than a headline. "Leader", "Sports", "RIGHT NOW", the masthead itself. The story rides on the sentence in the body, which is why its length is the constraint that moves.
How much room a sentence gets
Half of everything a newsroom sends falls between 64 and 109 characters: one clause of what happened, one clause of where. The mean, 89, sits 4 characters right of the median; the long pushes pull and the short ones cannot pull back. The year-wide figure hides the thing an editor would actually want to know, which is that the sentence gets a different amount of room in every market.
Italy averages 120 characters, past the 75th percentile of the whole year (109). Behind that average: 16 apps and 30,891 pushes. That is 61 more than Denmark, the shortest market here with more than one app.
The two ends of the chart are doing the same job with twice the words. Danish apps average 59 characters and Norwegian 64: a subject, a verb, a place, and the tap carries the rest. A push in Italy arrives at 120 already summarized, and Italy is the market that puts 10.5% of its entire year into the 07:00 hour, where an alert is closer to a morning digest than to an interruption. How long a push runs is set before the story arrives.
The habits
Only 5.2% of the year says a word that means breaking in its own language. An alert that has already vibrated a pocket may not need to announce that it is urgent, and counted across the whole year, the other habits an editor argues about are rarer than the argument suggests too.
- No title at all 12.6%
- A question mark 6.4%
- A word meaning “breaking” 5.2%
- An emoji 4.7%
- An exclamation mark 2.5%
Words meaning breaking are matched in the original language, before translation. And the emoji row reads as a rounding error, until it is split by country, below.
Bars are the share of each country's own pushes that carry at least one emoji, against the whole year's rate in blue.
-
France
26.9%
≈12,300
-
Belgium
20.5%
≈3,900
-
Austria
14.4%
≈750
-
Spain
13.2%
≈2,600
-
Germany
5.2%
≈1,600
-
Brazil
3.9%
≈530
-
Israel
3.4%
≈260
-
United States
3.3%
≈2,100
-
Norway
2.2%
≈680
8 more markets sit under one push in a hundred and are left off the bars: Turkey, the Netherlands, Italy, Switzerland, India, Sweden, the UK and Finland, 0.9% down to 0.0%.
Nearly half of the pushes that carry an emoji are French. France accounts for about 12,300 of the roughly 25,000, and Belgium is the only market anywhere near its rate. Language alone does not sort them: Austria at 14.4% and Germany at 5.2% write in the same one and disagree by a factor of 2.8, and Spain at 13.2% sits 3.4 times above Brazil at 3.9%. Switch the bars to counts and the United States climbs: at 3.3% it still puts more emoji-carrying alerts on lock screens than Germany at 5.2%, about 2,100 against about 1,600, because it sends so much more of everything.
3.9 minutes in Stockholm, 28 in New York
A story breaks, one app pushes it, and the clock starts on everybody else. Swedish and Norwegian apps answer in under four minutes. American apps take 28 minutes. In India the first push has the story to itself for 71.7.
We timed that gap market by market, counting from the first push at zero: the clock stops when a second app in the same market follows. Each bar below is one market's median. Pool the whole dataset instead, market boundaries taken out, and the median is 16.5 minutes, a number that matches almost nothing below it, since the markets run from under four minutes to over an hour.
3.9
min
Sweden median
In Sweden the second alert lands after a median of 3.9 min. In the United States it takes 28.0, 24.1 min longer, 7.2× the wait. Timed on 2,608 and 2,016 stories.
Sweden · median 3.9 minUS · median 28.0 min
Tap any market to move it into the red lane. Your last pick slides to blue.
In Stockholm a lead lasts about as long as it takes to type the follow-up. In New York a first push has closer to half an hour of clear air, long enough to decide that being second with something better beats being first with something thin.
A median hides a great deal. Pooled across the dataset, a quarter of second alerts land inside 3.1 minutes and a quarter take more than 64.4; that spread is the whisker on the pooled row. Every bar on the chart is the midpoint of a distribution that wide.
The rest of the field does not turn up in a pack either. The median wait for the third app is 51.5 minutes, roughly 3× the wait for the second, the dashed rule. By the time most of the field arrives, the story has been on lock screens for the better part of an hour.
One caveat, because it changes how you read the slow end. This measures the alert desk, not the newsroom. We see push decisions and never publication times. A market that batches its alerts into two or three editorial moments a day looks slow here even when both newsrooms had the story online seconds apart. Nor is it simply a matter of clock discipline: Italian apps put 10.5% of their year into a single local hour and take 44.5 minutes to the second alert, Swedish apps put 8.1% into theirs and take 3.9. Read the bars as a map of alerting habits rather than of who wrote first.
Being quick as a market is not the same as being quick as an app. Which names collect those first minutes is the next act.
Who gets there first
In most markets the app with the most firsts is a tabloid or a free daily; in Italy it is the public broadcaster.
For every story that at least three apps in one market pushed, the app whose alert arrived earliest takes the first. That leaves each outlet with two numbers. Firsts counts races won. First-rate, firsts as a share of the stories the outlet pushed at all, says how often it wins when it turns up. The blue tick in the panel below marks what an outlet's rate would be if firsts simply followed how often it took part.
Sweden
6 of 19 apps · baseline 25.1%
Aftonbladet took 523 firsts where its share of the pushing predicts 367, the widest margin here, and the most firsts in the market. Furthest the other way: Göteborgs-Posten, 1,092 stories pushed, first on 16.8%, 91 races below what attendance predicts.
- 1 Aftonbladet 523 of 1,460 35.8% +156
- 2 Expressen 440 of 1,283 34.3% +118
- 3 Omni 300 of 1,432 20.9% −60
- 4 SVT Nyheter 258 of 1,302 19.8% −69
- 5 TV4 Nyheterna 210 of 1,055 19.9% −55
- 6 Göteborgs-Posten 183 of 1,092 16.8% −91
Whiskers are 95% intervals on the rate; where one crosses the blue tick, the gap to the baseline is inside the noise.
France
6 of 20 apps · baseline 24.3%
BFMTV took 394 firsts where its share of the pushing predicts 307, the widest margin here, and the most firsts in the market. Furthest the other way: franceinfo, 1,029 stories pushed, first on 14.2%, 104 races below what attendance predicts.
- 1 BFMTV 394 of 1,262 31.2% +87
- 2 Le Figaro 373 of 1,429 26.1% +26
- 3 Le Parisien 312 of 1,237 25.2% +12
- 4 TF1 Info 212 of 908 23.3% −9
- 5 Ouest-France 156 of 693 22.5% −12
- 6 franceinfo 146 of 1,029 14.2% −104
Whiskers are 95% intervals on the rate; where one crosses the blue tick, the gap to the baseline is inside the noise.
United States
6 of 53 apps · baseline 23.0%
NYPost took 254 firsts where its share of the pushing predicts 215, the widest margin here, and the most firsts in the market. Furthest the other way: Breaking News, 1,186 stories pushed, first on 19.6%, 41 races below what attendance predicts.
- 1 NYPost 254 of 933 27.2% +39
- 2 Breaking News 232 of 1,186 19.6% −41
- 3 AP News 228 of 930 24.5% +14
- 4 NBC NEWS 148 of 643 23.0% ±0
- 5 CBS News 103 of 462 22.3% −3
- 6 BBC 87 of 410 21.2% −8
Whiskers are 95% intervals on the rate; where one crosses the blue tick, the gap to the baseline is inside the noise.
Switzerland
6 of 27 apps · baseline 31.1%
20 Minuten took 401 firsts where its share of the pushing predicts 306, the widest margin here, and the most firsts in the market. Furthest the other way: Nau, 562 stories pushed, first on 19.6%, 65 races below what attendance predicts.
- 1 20 Minuten 401 of 986 40.7% +95
- 2 Blick 330 of 842 39.2% +68
- 3 Aargauer Zeitung 153 of 526 29.1% −10
- 4 St. Galler Tagblatt 110 of 473 23.3% −37
- 5 Nau 110 of 562 19.6% −65
- 6 24 heures 88 of 448 19.6% −51
Whiskers are 95% intervals on the rate; where one crosses the blue tick, the gap to the baseline is inside the noise.
Finland
6 of 19 apps · baseline 27.7%
Iltalehti took 225 firsts where its share of the pushing predicts 166, the widest margin here, and the most firsts in the market. Furthest the other way: Keskisuomalainen, 337 stories pushed, first on 17.2%, 35 races below what attendance predicts.
- 1 Iltalehti 225 of 600 37.5% +59
- 2 Ilta-Sanomat 169 of 548 30.8% +17
- 3 Helsingin Sanomat 153 of 504 30.4% +13
- 4 Keskisuomalainen 58 of 337 17.2% −35
- 5 Aamulehti 38 of 238 16.0% −28
- 6 Hufvudstadsbladet 37 of 227 16.3% −26
Whiskers are 95% intervals on the rate; where one crosses the blue tick, the gap to the baseline is inside the noise.
Norway
6 of 12 apps · baseline 30.9%
Dagbladet took 185 firsts where its share of the pushing predicts 145, the widest margin here, and the most firsts in the market. Furthest the other way: NRK, 343 stories pushed, first on 23.9%, 24 races below what attendance predicts.
- 1 Dagbladet 185 of 470 39.4% +40
- 2 Aftenposten 99 of 378 26.2% −18
- 3 NRK 82 of 343 23.9% −24
- 4 VG 74 of 215 34.4% +8
- 5 DN 33 of 130 25.4% −7
- 6 E24 27 of 82 32.9% +2
Whiskers are 95% intervals on the rate; where one crosses the blue tick, the gap to the baseline is inside the noise.
Spain
5 of 12 apps · baseline 30.6%
EL PAÍS took 234 firsts where its share of the pushing predicts 147, the widest margin here, and the most firsts in the market. Furthest the other way: El Mundo, 432 stories pushed, first on 22.5%, 35 races below what attendance predicts.
- 1 EL PAÍS 234 of 481 48.6% +87
- 2 Diario ABC 135 of 470 28.7% −9
- 3 El Mundo 97 of 432 22.5% −35
- 4 SUR 15 of 142 10.6% −29
- 5 La Voz de Galicia 11 of 81 13.6% −14
Whiskers are 95% intervals on the rate; where one crosses the blue tick, the gap to the baseline is inside the noise. One of the six entered fewer than 30 races and is left out: ElDiario.es (5 of 17).
Italy
5 of 16 apps · baseline 30.0%
RaiNews.it took 150 firsts where its share of the pushing predicts 117, the widest margin here, and the most firsts in the market. Furthest the other way: Leggo, 282 stories pushed, first on 20.9%, 26 races below what attendance predicts.
- 1 RaiNews.it 150 of 390 38.5% +33
- 2 Corriere.it 111 of 360 30.8% +3
- 3 Il Mattino 61 of 218 28.0% −4
- 4 Leggo 59 of 282 20.9% −26
- 5 Il Sole 24 Ore 26 of 107 24.3% −6
Whiskers are 95% intervals on the rate; where one crosses the blue tick, the gap to the baseline is inside the noise. One of the six entered fewer than 30 races and is left out: Corriere Edicola (11 of 26).
United Kingdom
6 of 12 apps · baseline 27.9%
Sky News took 108 firsts where its share of the pushing predicts 84, the widest margin here, and the most firsts in the market. Furthest the other way: GB News, 223 stories pushed, first on 17.5%, 23 races below what attendance predicts.
- 1 Sky News 108 of 302 35.8% +24
- 2 The Sun 70 of 212 33.0% +11
- 3 Mirror 60 of 242 24.8% −8
- 4 The Telegraph 50 of 215 23.3% −10
- 5 GB News 39 of 223 17.5% −23
- 6 DailyMail 34 of 100 34.0% +6
Whiskers are 95% intervals on the rate; where one crosses the blue tick, the gap to the baseline is inside the noise.
Belgium
6 of 12 apps · baseline 29.7%
Nieuwsblad took 100 firsts where its share of the pushing predicts 60, the widest margin here, and the most firsts in the market. Furthest the other way: Gazet van Antwerpen, 263 stories pushed, first on 16.3%, 35 races below what attendance predicts.
- 1 Nieuwsblad 100 of 203 49.3% +40
- 2 VRT NWS 89 of 195 45.6% +31
- 3 Gazet van Antwerpen 43 of 263 16.3% −35
- 4 RTBF Actus 33 of 109 30.3% +1
- 5 Het Belang van Limburg 29 of 183 15.8% −25
- 6 L'Avenir 20 of 104 19.2% −11
Whiskers are 95% intervals on the rate; where one crosses the blue tick, the gap to the baseline is inside the noise.
Germany
6 of 20 apps · baseline 30.8%
Tagesspiegel took 69 firsts where its share of the pushing predicts 55, the widest margin here, and the most firsts in the market. Furthest the other way: DW, 87 stories pushed, first on 24.1%, 6 races below what attendance predicts.
- 1 Tagesspiegel 69 of 177 39.0% +14
- 2 DER SPIEGEL 40 of 141 28.4% −3
- 3 SZ 26 of 99 26.3% −4
- 4 WDR aktuell 26 of 88 29.5% −1
- 5 DW 21 of 87 24.1% −6
- 6 Handelsblatt 20 of 64 31.3% ±0
Whiskers are 95% intervals on the rate; where one crosses the blue tick, the gap to the baseline is inside the noise.
Denmark
4 of 7 apps · baseline 31.3%
TV 2 Nyheder took 58 firsts where its share of the pushing predicts 45, the widest margin here, though B.T. wins more races outright, 63 to 58, off a bigger docket. Furthest the other way: Jyllands-Posten, 99 stories pushed, first on 15.2%, 16 races below what attendance predicts.
- 1 B.T. 63 of 163 38.7% +12
- 2 TV 2 Nyheder 58 of 143 40.6% +13
- 3 DR Nyheder 43 of 166 25.9% −9
- 4 Jyllands-Posten 15 of 99 15.2% −16
Whiskers are 95% intervals on the rate; where one crosses the blue tick, the gap to the baseline is inside the noise. Two of the six entered fewer than 30 races and are left out: Berlingske (1 of 5), KristeligtDagblad (0 of 7).
The names in front are tabloids, free dailies and rolling-news television: Aftonbladet and Expressen in Sweden, 20 Minuten and Blick in Switzerland, BFMTV in France, Sky News in Britain, B.T. in Denmark, Dagbladet in Norway, NYPost in the United States. (Country here is a property of the app, not of the newsroom, which is why BBC sits in the American six.) Attendance buys the count and speed buys the share; the outlets in front have both. The app that pushed the most French stories in this set is Le Figaro, 1,429 of them, and it still lost the year to a rival working off fewer stories: BFMTV took 394 firsts to Le Figaro's 373.
Public broadcasters are the mirror image. franceinfo pushed 1,029 of the French stories, more than most of its rivals, and was first on 14.2% of them. In Sweden, where the second alert lands 3.9 minutes after the first, being the reliable second source is a position rather than a failure.
Not every broadcaster takes that role. RaiNews.it is first on 38.5% of the Italian stories it pushes, and it is also the most prolific app on the whole rack at 59.7 alerts a day. VRT NWS is 31 up in Belgium, behind only Nieuwsblad. Denmark is the only market where the most firsts and the highest rate point at different apps, and they part by 1.9 points across the 306 races the two entered between them: B.T. 38.7%, TV 2 Nyheder 40.6%. Either whisker swallows the gap.
Reading the gutter against the lane, the minutes and the rates do not line up. Among Swedish apps the second alert lands 3.9 minutes after the first, and the quickest Swedish outlet still wins only 35.8% of what it pushes. Among Italian apps the second takes 44.5 minutes, and RaiNews.it wins 38.5%. A tight chase does not make the first mover weak, and a slow one does not make it strong.
One limit, and it changes how to read every rate here. This is push timing, not publication time. A desk that publishes first and alerts on the hour reads as slow; a desk that fires an alert off a wire line and writes the story afterwards reads as fast. The artefact being timed is the alert itself, SZ's Louvre push, no title, all of it: “BREAKING - Louvre Director Laurence des Cars resigns”. The baseline has a limit of its own: it treats a market's six as interchangeable from story to story, so an outlet that only pushes the big pile-ups meets more rivals in each of its races than the model allows.
Firsts are won inside one market. The next question is which stories were big enough to make every market run at once.
The biggest story reached 121 of 250 apps
At 06:16 UTC on 28 Feb 2026 the first alert we caught about the Israeli and American strike on Iran left a newsroom. 121 apps in 27 countries ended up pushing it, 48.4% of every app that pushed anything that month, and more than any other event of the year. The median event that three or more apps thought worth an alert reached 4 apps.
Clustering lets us ask a better question than which app pushed most: how many separate newsrooms decided the same thing was worth an interruption? Rank the year's events by that answer and the top twelve are below, each one measured against the field that was pushing at the time.
Showing story 01 of 12, Israel and the US strike Iran, 28 Feb, 121 apps in 27 countries.
Pick a story to open it. Prev and Next, or the arrow keys, step through the twelve.
Israel and the US strike Iran
Just after 06:00 UTC, Israel announced a "preemptive strike" on Iran with American participation. Explosions in Tehran, a state of emergency in Israel, then Iranian missile and drone barrages that reached US bases in Bahrain and Qatar and closed Gulf airspace. Six of the year's 25 biggest clusters are shards of this one morning; the count here is the largest single shard.
War in the Middle East. "Death to America!", "Death to Israel!" - thousands mourn the death of Ali Khamenei | LIVE TEXT/VIDEO
US-Israel: Preemptive strike on Iran hits US bases in Qatar, Bahrain, the Emirates, and Kuwait. Defense Minister Israel Katz announced, declaring "an immediate state of emergency across the country." Media reports: Khamenei is in a safe place. NYT reports: US attack on Iran more extensive than the one in June. Explosions in several cities across the country.
Chuck Norris dies at 86
TMZ reported the actor's death after a medical emergency in Hawaii and his family confirmed it within minutes. Four apps pushed in the same minute, at 13:59 UTC, and inside five minutes the story was on lock screens across Europe.
Chuck Norris has died. The famous actor was 86 years old and was hospitalized with health problems
Norris died at the age of 86 yesterday morning, after being hospitalized for the last 24 hours on the island of Kauai. Norris's family confirmed the death in a post on his personal Instagram account, calling it a "sudden passing." They said he was "reconciled" and with his family at the time.
Orbán loses Hungary's election
Hungarians voted in what European outlets billed as the year's most important EU election. Péter Magyar's Tisza party ended Viktor Orbán's 16 years in power on record turnout, and congratulations from European leaders followed the next morning.
Nicușor Dan and European leaders, congratulatory messages for Peter Magyar. Viktor Orban lost power in Hungary after 16 years
Viktor Orban lost power in Hungary after 16 years, and the opposition led by Peter Magyar won the parliamentary elections. The vote ended with a record turnout and a major change on the political scene. The counting of over 98% of the votes showed that the center-right, pro-EU Tisza party, led by Péter Magyar, won 138 seats. That is, a crucial two-thirds majority in the 199-member parliament, ahead of Orbán's Fidesz party, which won 55 seats.
Hormuz ultimatums and a rejected ceasefire
With the Strait of Hormuz blockaded, mediators put a 45-day ceasefire on the table. Iran rejected it, Trump rejected it too, then set deadlines on Truth Social and told Iran to "open Hormuz, you crazy bastards", warning on 7 April that "an entire civilization will die tonight". Iran broke off the talks.
Media: Tehran threatens to leave the entire Gulf in the dark. All communication channels with the US closed.
VIDEO - From threats to a proposed agreement, Trump sets yet another ultimatum. On Truth, the US president: "Open Hormuz, you crazy bastards," otherwise power plants and bridges will be hit. The deadline is 8:00 PM Washington time.
The war reaches the Gulf
Israeli strikes on Tehran and Beirut were answered by Iranian missiles and drones over the Gulf: at least eight dead in the UAE, explosions in Doha and Dubai, airspace closures, and Macron convening back-to-back defence councils in Paris.
No sampled wording can be pinned to this cluster with confidence, so the card carries none.
The Islamabad peace talks collapse
Iran's delegation left Pakistan without meeting the US envoys, and Trump cancelled Witkoff and Kushner's trip with the words "too much work", later claiming Iran offered better terms "within 10 minutes" of the cancellation.
No sampled wording can be pinned to this cluster with confidence, so the card carries none.
Secret Service kills intruder at Mar-a-Lago
A man who appeared to be armed was shot dead after entering the secure perimeter at Trump's Florida estate. Reuters moved it first; ninety apps pushed within the hour.
He tried to enter Trump's Florida resort. The sheriff: "He placed the can on the ground and then raised the rifle to firing position."
A man who "appeared" to be armed was shot and killed by Secret Service after "unauthorized entry into the secure perimeter at Mar-a-Lago," agency says
Andrew Mountbatten-Windsor arrested
Police arrived at Sandringham on the former prince's birthday and arrested him on suspicion of misconduct in public office, the BBC reported first. He was released the same evening.
No sampled wording can be pinned to this cluster with confidence, so the card carries none.
Shots at the correspondents' dinner
Gunfire in a hallway at the White House Correspondents' Dinner sent journalists under tables and Trump and the first lady out of the room. A suspect was arrested on the spot, the dinner was postponed, and Trump gave a late-night statement praising the Secret Service.
No sampled wording can be pinned to this cluster with confidence, so the card carries none.
UEFA threatens to boycott the World Cup
After The Times revealed Gianni Infantino's plan to sell part of the World Cup to private investors, every European federation lined up behind a boycott threat, throwing the tournament into open crisis within an afternoon.
No sampled wording can be pinned to this cluster with confidence, so the card carries none.
Hantavirus on the MV Hondius
Three passengers died in an outbreak on a cruise ship off Cape Verde. The WHO confirmed cases aboard and warned human-to-human transmission was possible, ports refused the ship for days, and it finally docked in Tenerife under quarantine on 10 May.
No sampled wording can be pinned to this cluster with confidence, so the card carries none.
Supreme Court voids Trump's tariffs
The US Supreme Court ruled the administration's global tariffs illegal. Trump called the ruling a "shame" and promised new 10-percent tariffs by other means before the day was out.
WASHINGTON — The Supreme Court struck down President Donald Trump's far-reaching global tariffs on Friday, handing him a significant loss on an issue crucial to his economic agenda. The 6-3 decision centers on tariffs imposed under an emergency powers law,...
The 50% line is half the apps that were pushing that month. None of the twelve reaches it.
Read the list and the year is lopsided. The Iran war supplies four of the twelve, the 28 Feb strikes, the Gulf barrages, the Hormuz ultimatums, the collapse of the Islamabad talks. Trump supplies three more: the intruder shot at Mar-a-Lago, the gunfire at the correspondents' dinner, the Supreme Court voiding his tariffs. The rest is what lock screens are otherwise for: a dead action hero, an election in Hungary, an arrest at Sandringham, a hantavirus outbreak on a cruise ship, and a football federation threatening to walk out of a World Cup. Ten of the twelve fall between February and April, the months the war ran hottest. An earlier round of the same conflict, the twelve days of June 2025, falls just before this window opens; we told that one alert by alert.
Reach can be counted two ways, and they disagree. 81 apps carried the World Cup boycott threat into 25 countries; 110 apps carried the Hungarian election into 22. Switch the panel to countries and the two change places. One is a few desks in most markets, the other most of the desks in a few. And three days apart in February, the arrest at Sandringham and the shooting at Mar-a-Lago landed almost on top of each other: 89 apps and 90 apps, 21 countries each.
The bars are not in the order of the list. 119 apps is fewer than 121 and a larger share of the field, 49.6% of everything pushing in March 2026 against 48.4% in February 2026. Measured that way, the biggest story of the year is the death of Chuck Norris.
No event in the year got past 121 apps, and that was still 48.4% of the field. The twelfth story reached 73, or 29.2%. Nothing gets pushed everywhere.
The eight busiest days
Story-level clustering hides the days when everything lands at once. These are the eight days the phones recorded most, and measured against an average day in their own month only one of them is unusual: 29 Jul ran 3.3 times its own July. The other seven sit between 18% and 40% above their month, busy, not exceptional. Open a day to read the wording that carried it.
| Date | Pushes | Against its own month | Most-pushed wording that day |
|---|---|---|---|
| 29 Jul 2026 | 3,216 | 3.3× |
28 apps
|
| 17 Mar 2026 | 3,172 | 1.3× |
93 apps
|
| 18 Mar 2026 | 3,133 | 1.3× |
41 apps
|
| 11 Feb 2026 | 3,098 | 1.4× |
57 apps
|
| 19 Mar 2026 | 3,007 | 1.2× | , No wording from this day is in the text sample. |
| 20 Feb 2026 | 2,994 | 1.4× |
68 apps
|
| 20 Mar 2026 | 2,968 | 1.2× |
119 apps
|
| 23 Mar 2026 | 2,908 | 1.2× |
60 apps
|
That is the year as the lock screens told it: 537,376 interruptions from 336 apps, and twelve stories that most of them agreed on.
Read end to end, the year says one thing plainly: almost nothing about a push notification is global. The clock a market keeps, how fast it answers itself, how long it writes and whether it reaches for an emoji are all set locally, market by market. The file below is the evidence, for anyone who wants to check it against their own desk.
Download the dataset
Every table on this page, as one CSV. Released under Creative Commons Attribution 4.0: reuse it, redraw it, sell work built on it, credit AO.news and link back.
AO.news (2026). A Year of Push: 537,376 news notifications from 336 apps [Data set]. https://ao.news/data/year-of-push-notifications
- Window
- 1 Aug 2025 to 31 Jul 2026
- Notifications
- 537,376
- Apps
- 336 in 44 countries
- Format
- Aggregate tables, long format; no notification text
- Questions
- hello@ao.news
Watch the next story break, in the live deck.
Open the example deck: hundreds of news apps, clustered by story, updating in real time. Newsroom access unlocks the full notification text, Slack and the API. Write to hello@ao.news.