Week of 5–11 September 2026

Between 5 September and 11 September 2026 this project collected 110,866 articles and grouped them into 4,094 stories across seven markets. A story is a cluster of articles close enough in meaning to be the same event³; independence is counted in owners, not outlets⁴. Superscripts link into the method page, where each measurement is defined once.

Volume

Counts are after duplicate feed entries are merged², and they are not flat across the week: a weekend day runs about half a weekday¹. The owner column counts the owners that carried at least one story, which is the number the rest of the report is measured against — a story reaching six of them is a different thing in a market of thirty-three and in a market of six.

This report is written in English, but two of the markets are not — headlines from de-national and ru-domestic and ar-pan appear below as they ran, untranslated.

Of the 158 feeds polled in these markets, 3 produced nothing at all this week³².

Stories per market for every full week collected so far. The series is recomputed from the archive each time a report is built rather than copied from earlier ones, so a number here can differ from the number that was printed then³⁰.

What the week was about

The three widest stories in each market, with the words that separate each from the rest of its week⁶:

us-national. Trump makes stunning first-of-its-kind cash promise in bid to rally GOP voters at Dallas midterm convention — 27 owners, 217 articles, published over 22 hours (Trump, dividend, midterm, Republicans). Remembering 9/11, the day that changed everything — 26 owners, 361 articles, published over 24 hours (attacks, Pentagon, diesel, Vance). GOP puts Trump ‘on the ballot’ with historic midterm convention and more top headlines — 23 owners, 238 articles, published over 24 hours (convention, midterm, oil, Dallas).

uk-national. UK set to announce sanctions on illegal Israeli settlements in huge 'reset' of relations — 12 owners, 45 articles, published over 23 hours (Israeli, settlements, Miliband, Bank). Moment protesters wade into sea as rescued migrants arrive in Portsmouth — 11 owners, 46 articles, published over 22 hours (Portsmouth, migrant, Dover, protest). How did YOUR MP vote on the assisted dying bill? — 11 owners, 34 articles, published over 17 hours (assisted, dying, bill, MPs).

uk-regional. The day the world changed forever — 4 owners, 17 articles, published over 13 hours (attacks, anniversary, Pentagon, York). Andy Burnham responds to John Swinney's call for 'legal indyref2 mechanism' — 4 owners, 14 articles, published over 10 hours (indyref, Burnham, Swinney, mechanism). Man in ‘serious condition’ following Lurgan stabbing — 4 owners, 13 articles, published over 12 hours (arson, stabbing, attempted, endanger).

intl-state. Putin begins Ukraine war talks with US envoys Witkoff, Kushner in Moscow — 10 owners, 62 articles, published over 23 hours (Witkoff, Kushner, envoys, Putin). West called for Ukraine ceasefire to quell massive Russian attacks — Lavrov — 9 owners, 73 articles, published over 23 hours (Kyiv, Lavrov, Ukraine, Russia). Dialogue, diplomacy way forward to resolve conflicts, India’s Modi tells Putin — 9 owners, 55 articles, published over 24 hours (BRICS, summit, Putin, India).

de-national. Bundeskanzler Merz: AfD-Verbot löst das Problem nicht — 16 owners, 189 articles, published over 24 hours (AfD, Merz, Generaldebatte, Bundestag). Landtagswahl: AfD siegt in Sachsen-Anhalt – Debakel für die CDU — 15 owners, 497 articles, published over 24 hours (Anhalt, Sachsen, AfD, Landtagswahl). Landtagswahl: Wer gewinnt in Sachsen-Anhalt? — 15 owners, 455 articles, published over 21 hours (Anhalt, Sachsen, Landtagswahl, AfD).

ru-domestic. Покушавшемуся на жизнь военного в Саратовской области предъявили обвинение — 18 owners, 93 articles, published over 15 hours (Молдавии, Саратовской, Кочу, Николае). Число погибших при сходе лавины в Кабардино-Балкарии достигло 11 — 17 owners, 55 articles, published over 15 hours (Кабардино, Балкарии, лавины, альпинистов). МегаФон запустил первую коммерческую сеть 5G в России — 17 owners, 49 articles, published over 16 hours (запустили, Минцифры, городах, сеть).

ar-pan. الجزائر تعلن قطع العلاقات الدبلوماسية مع الإمارات — 13 owners, 54 articles, published over 20 hours (الجزائر, الدبلوماسية, قطع, الإمارات). ويتكوف وكوشنر يصلان إلى موسكو (صور) — 13 owners, 38 articles, published over 17 hours (ويتكوف, بوتين, كوشنر, ستيف). بعد مقتل 7 أشخاص.. غارات إسرائيلية جنوبي لبنان وإصابة مراسل ومصور — 12 owners, 103 articles, published over 24 hours (لبنان, جنوب, النبطية, مستشفى).

Reaching several newsrooms and being written about a lot are not the same thing³¹. The most lopsided story in each market — it travelled, and most of what was written about it still came from one owner:

The same day in seven markets

Stories are clustered inside a market and never across them⁷. The boundary is worth reading on its own, because most of what a market carries never crosses it.

The match behind that last column is deliberately loose⁸: the claim being made is “nobody else ran this”, so it should fail towards saying a story travelled. At the strict threshold the same shares are 26%, 47%, 23%, 58%, 21%, 6% and 17%, in the order above.

172 events ran in more than one market this week, and 122 of them crossed a language boundary³³. The table below shows only those that reached three markets or more; most crossings are between two, so a story picked up in English and in German on the same day is counted here and not shown there.

Where an event did cross, the wording is the interesting part. The same day, the same event, as each market ran it:

Saturday 5 September

Monday 7 September

Friday 11 September

The other side of the same measurement is what a market carried and nobody else did⁹:

intl-state is the row worth reading in full: the state-funded broadcasters are kept as their own market precisely because they are a labelled example of coordinated promotion, so the stories only they carried are the ones to look at.

How far stories travel

Owner counts here are corrected for wire copy: two owners running one agency text under two mastheads count once between them⁵.

How long a story lasts

A week-long story is a chain of daily clusters rather than one object¹⁰. Ordering is by how much of this week a story occupied, and it has to hold at least 3 days of it to appear¹¹. The column worth reading is the last one: a story that opens wide was already everyone’s on day one, and one that opens narrow had to travel first¹².

Counting owners per day says how wide a story got; the rows above say how it filled up¹³. It is one story, though, not a claim about how these owners behave¹⁴.

Who moves together

A high score means two owners run the same stories more often than their volumes explain, which is not the same as coordination¹⁵. What it cannot say is whether either is following the other; for that the next section uses the clock. Only pairs above a score of three appear, which is where pairs thrown up by chance alone stop appearing²⁹.

One market can be checked against something already known³⁴. The outlets in intl-state answer to different governments, and three pairs of them answer to the same one — Anadolu Ajansi with TRT, China Media with People's Daily and Rossiya Segodnya with TASS. The measure puts those pairs 1st, 2nd and 3rd of 44 in that market.

How stories move between newsrooms

Direction is what separates the mechanisms, and it is the part a co-occurrence count cannot see¹⁶.

Measured over the whole archive rather than this week, and only after two controls: each owner’s own speed is subtracted, and the pair’s tightest quarter of stories is dropped¹⁷. Timing is the noisier of the two measurements in this report, so the bar here is the same three and worth taking literally²⁹, and a pair also has to hold up in each half of the archive counted on its own³⁵.

The same measurement can be asked of a single owner: is it upstream of the newsrooms it shares stories with, or downstream of them¹⁸?

The two ends read differently, and neither is an accusation¹⁹.

What the headline said first

Every version of an article’s text is kept, so a headline that changes after publication is visible. This section counts the four English-language markets only, not all seven²⁸. This week 1,974 headlines were rewritten there, 917 of them substantially — less than half the words survived — and the median rewrite landed 4.2 hours after the article first appeared. Not all of that is a newsroom changing its mind²⁰: requiring the two versions to be recognisably the same sentence, changed within a day, leaves 676 — and those are the ones below.

This is the one place in the corpus where an editorial decision is observed twice — the same newsroom, the same story, two framings hours apart²¹.

• Warner Bros — “Rescuers in Nepal pull Chinese national alive from tunnel 10 days after floods” → “September 5, 2026 — Updates on rescue efforts in Nepal 10 days after catastrophic floods”
• Nexstar — “Duffy issues full ground stop after plane overruns Miami runway” → “At least five dead, others hospitalized after plane overshoots Miami runway”
• Associated Press — “German chancellor faces tough times after first AfD state election win” → “German leader under pressure after far-right state election triumph”
• BBC — “UK pupils rank highly again in international reading, maths and science tests” → “England up in reading, maths and science rankings as Scotland and Wales fall behind”
• TRT — “Paris probes claim female Eiffel Tower staff were hidden from Hindu delegation” → “Eiffel Tower reopens after staff walkout over removal of women employees”
• Lebedev — “How long will Heathrow and Gatwick flights chaos last? Passengers face days of disruption” → “When will flights at Heathrow and Gatwick get back to normal?”

The examples are drawn from stories that at least 4 owners carried, not from the largest rewrites²².

Rewriting is not evenly spread. Emerson Collective changed the headline on 44% of what it published this week; across the owners with enough output to measure the rate runs from that down to a percent or two²³.

What the rewrites do not show is a softening of language. Quotation marks appear in 168 rewritten headlines and disappear from 91; words that hold a claim at arm’s length — so-called, alleged, reportedly — move in 22 and out of 14. At that size neither is a finding. Distancing is thin enough in headlines that it has to be measured across the whole corpus rather than inside its edits, which is the next section.

Words that have to be quoted

This project is named after the Overton window, and everything above measures something else: the shape of distribution. The window is answered by how a word is marked, not by how far it travels²⁴.

Counted over the four English-language markets only²⁸. The separation is an order of magnitude and it is not subtle. Nothing about the count knows which words are contested — the list is fixed in advance, and it has to be²⁵. Read the shares as a level, not as a movement²⁶: the window is the change in them over months.

Method

Every measurement above is defined once, on the method page — the superscripts throughout this report link straight into it. Reports are generated without human editing and without a language model²⁷.