Syndication is not conspiracy: telling a herd from a script

One story about AI and bank job cuts ran on 52 websites this week, and LSN News, my news engine, found only 2 independent voices behind it. The Microsoft outage ran on 2,272 outlets with 1,594 independent voices. When the same headline shows up on fifty websites, it is tempting to see a hand behind it. Usually there isn’t one. The outage was a herd. The bank story was syndication. Neither was a script. Here is the method I use to tell them apart, and how far the evidence goes.

Contents · 10 sections

TL;DR #

• A herd is many newsrooms covering one event independently. Syndication is one article published by a network on many sites. A script is unrelated outlets coordinated to push one frame.
• Four measurements tell them apart: independent voices, network share, timing and wording overlap.
• The Microsoft outage was a herd: 2,272 outlets, 1,594 independent voices, no network above 3%.
• This week’s AI-layoff stories were syndicated: 25 to 52 outlets from 2 to 5 voices, one network each, with no sign of coordination between networks.
• Bernays, Lippmann and Herman & Chomsky all describe influence that works mostly without a central hand.

Three things people mix up #

When coverage looks suspiciously uniform, there are three very different explanations:

• A herd. Many independent newsrooms cover the same thing at once, because it matters, or because their competitors are covering it. They write their own copy, choose their own angle, and still converge. That is normal news.
• Syndication. One newsroom writes a piece and a network publishes it on many sites: a newspaper group, a TV group, a news agency, an aggregator. The words are identical because they are literally the same article. That is a business model.
• A script. Outlets that do not share an owner or a feed push the same framing, at the same moment, in the same unusual words, because someone coordinated them. That is what people mean by propaganda or narrative steering.

The first two are everywhere. The third does happen, but it leaves different fingerprints. The trick is to look for them before reaching for the word “conspiracy”.

The method: four measurements #

For any story, LSN News lets me check four things.

1. Independent voices. Outlets that carry the same headline count as one voice, and each known outlet network counts as one. If voices are close to outlets, many people wrote independently. If voices are a small fraction, one piece is being multiplied.

2. Network share. What fraction of the outlets belong to one network? Under 5% means no group shaped the story. Over 50% means the story is mostly one group’s product.

3. Timing. A herd follows the event and the clock: newsrooms wake up, check facts, and publish over hours, region by region. Syndication arrives in bursts, often within minutes, on one network’s sites. A script would show unrelated outlets moving in lockstep.

4. Wording overlap. Identical headlines inside one network are syndication. Identical unusual phrasing across outlets with no shared owner or feed is the red flag, because that needs a common source nobody disclosed.

None of these is proof on its own. Together they grade the evidence.

Case 1: the outage was a herd #

Microsoft 365 failed on the evening of 12 November. By 17:45 UTC on Friday, LSN News counted 4,556 articles from 2,272 outlets in 55 languages in one story.

• Voices: 1,594 independent voices, about 70% of outlets. More than 2,500 distinct headlines.
• Network share: the largest network, a group of Australian regional papers, was 67 outlets, under 3%. No group dominated.
• Timing: coverage followed the sun. Spanish and Brazilian tech desks published first, at 18:30 UTC, ahead of American outlets. A wave of about 200 articles an hour on Thursday evening, a trough overnight, then a peak of 448 in one hour on Friday as Europe woke.
• Wording: each country picked its own frame. British papers wrote about councils and rail tickets, American ones asked “Is Outlook down?”, Spanish and French ones led with “global”. Where the wording was shared, it was shared inside networks, like the wire or group copy “Is Outlook down?” that ran on 163 outlets.

Even the cause shows a herd at work. About 80 articles used Microsoft’s correct wording (the load “wasn’t driven by an increase in traffic”) and about 6 used a wrong paraphrase (“unusually high demand”). A script would have produced one version. A herd produced a majority and some noise.

Verdict: documented independent coverage. Thousands of newsrooms, many angles, one event.

Case 2: the AI-layoff stories were syndicated #

Now three stories from the same week that framed job cuts around AI:

Story	Outlets	Voices	Main network	Timing
“Bank bosses face inquiry over AI and offshore job cuts”	52	2	Australian regional group, 52 of 52	about 10 hours, 2 headlines
“US-China AI race could face … mass layoffs over next three years: Report”	25	2	Big News Network, 23 of 25	about 22 hours, 2 headlines
“What to consider if company making AI-related job cuts wants you back”	45	5	MediaNews Group, 34 of 45	over 4 days, mostly one headline

All four fingerprints point the same way. Few voices. One network carrying most of the copies. Bursts inside that network. Identical headlines inside the network, and different ones outside it.

What we do not see is the script signature: different owners pushing the same frame at the same time. The three stories come from three unrelated networks, in different countries, on different days, and one of them argues the opposite of the others. The MediaNews story is about companies regretting AI cuts and offering jobs back. If someone were steering an “AI is taking the jobs” narrative, they did a poor job of it.

Verdict: documented syndication, no evidence of coordination. The fairest phrase is that these outlets “publish the same headlines as” each other, because they publish the same articles.

What the library says #

I went to my library for the classics on propaganda, because they are more careful than their reputation.

Edward Bernays wrote Propaganda in 1928, and he did not hide what he was doing. Its first chapter opens: “The conscious and intelligent manipulation of the organized habits and opinions of the masses is an important element in democratic society.” Bernays was a press agent. A core part of his craft was to place the story at the source, so that independent newsrooms would carry it in their own words. My reading, worth remembering: a skilled influence campaign tries to look like a herd, not like syndication. Identical copy on fifty sites of one group is the opposite of what a skilled operator wants.

Walter Lippmann’s Public Opinion, from 1922, explains the more common mechanism. Under deadline, editors copy the shape of the leading coverage: “It is here that newspapers influence each other most deeply.” And he rejected the conspiracy reading of his own critique. The press, he wrote, is “not so universally wicked, nor so deeply conspiring” as its harshest critics claimed, but “very much more frail” than democratic theory admits.

My library does not hold Herman and Chomsky’s Manufacturing Consent itself, only its reference shelf, which quotes the book’s own summary: the media serve power “by reliance on market forces, internalized assumptions, and self-censorship, and without overt coercion.” One of their five filters is sourcing: newsrooms depend on the institutions that feed them cheap copy. Whatever you think of their politics, their model is explicitly not a conspiracy. It is a structure. Syndication is one of the structures they would point to.

Grading the evidence honestly #

Here is how I would state the findings, in plain words:

• Documented: the outage coverage was broad and mostly independent. The AI-layoff stories above reached many outlets through a few networks.
• Inferred: in this week’s data, the “AI and jobs” theme looks bigger in outlet counts than in independent voices. Count the voices before you count the outlets.
• Not shown: any coordination between networks, or any intent behind the syndication. Republishing deals, editorial calendars and cost-cutting explain everything I see.
• We don’t know: whether this pattern is new. LSN News only has data from 7 November 2026, about one week. I cannot compare it with last year.

The measurements also have blind spots. Two outlets rewriting the same agency copy with different headlines count as two voices, so syndication is if anything under-counted. A script by competent people would rewrite each copy and could slip past these four checks; catching that needs reading, not counting.

Checklist: herd, syndicate or script? #

When a story seems to be everywhere:

1. Count voices, not sites. Open three copies from different-looking sites. Same text? One voice.
2. Find the owner. Footers, “originally published by”, the same byline. One network is syndication.
3. Check the clock. Did it spread with the event, or in one burst on one group’s sites?
4. Look across owners. Only identical, unusual wording from unrelated outlets at the same time deserves the word “coordinated”.

Frequently asked questions #

Is news syndication the same as propaganda? #

No. Syndication is one article published on many sites by a group, agency or aggregator. It is a business model. Propaganda needs a deliberate attempt to shape opinion, which syndication by itself does not show.

How can I tell if news coverage is coordinated? #

Look for identical, unusual wording or framing from outlets that share no owner or feed, appearing at the same time. Identical text inside one network is syndication, not coordination.

Was the Microsoft outage coverage steered? #

No evidence of that. About 70% of outlets wrote their own copy and the largest network was under 3% of the story.

Were the AI-layoff stories coordinated? #

The data shows three stories multiplied by three unrelated networks, on different days, one of them arguing the opposite. That is syndication. Nothing shows coordination between them.

What is the propaganda model? #

Herman and Chomsky’s 1988 model in Manufacturing Consent: five filters (ownership, advertising, sourcing, flak and a common enemy) that bias news through structure and market forces, without overt coercion.

The lesson #

Uniform coverage is not, by itself, evidence of a hidden hand. Count the voices, find the owners, check the clock, and only then look for the rare signature of real coordination: unrelated outlets saying the same unusual thing at the same moment. Most of the time you will find a herd or a syndicate. Both are worth knowing about. Neither is a conspiracy. The outage story page shows what a herd looks like, and the bank job cuts page shows a syndicate; one click each.