The corpusThe licenceThe 21The aggregateThe graphSourcesHow to reach themThe vaultMethod
What twenty-one editors said, counted
Twenty-one of the most senior people in news were asked the same question in the same week and answered it in public. Read one at a time these are opinion pieces. Counted together they are a picture of what this industry currently believes is worth saying — and, more usefully, of what it does not say.
Beta, agent-produced. The fetching, extraction, classification and drafting here are done by software agents against frozen bytes. Nothing on this page is a legal opinion, and there has been no legal review. If you are deciding whether you may republish one of these pieces, read the terms on the piece itself and ask the publisher — that is the point we are making.
We hold all twenty-one pieces and republish none of them. Every article is linked to its original on worldnewsday.org. What is published here is the description: who wrote it, under what terms, what it links to, what words it contains. The prose belongs to the people who wrote it, and the gate on this section fails the build if twelve consecutive words of any piece appear on any page we generate.
Every number walks back to bytes we hold. Each page was fetched once, frozen to a dated snapshot in this repository and hashed with SHA-256. The counts are re-derived from those bytes by a second program before anything ships. The register → · The method →
What a count here is, and is not.
- No sentiment, no stance, no scoring. A count of words is not a measure of a person.
- No claim that two authors agree. agrees_with is a banned verb in this graph.
- No claim about what an author believes, only about what this one piece contains.
- No ranking of the publications. The corpus is not a competition and we are not a judge.
What they reach for
Twenty published patterns, run over each article’s own prose. A theme means the piece contains those words. It does not mean the author cares about the subject, and its absence does not mean they do not — a nine-hundred-word piece has room for four ideas. The patterns are published in full so you can disagree with them precisely.
| Theme | Share of the corpus | Count | Pattern |
|---|---|---|---|
| Audience, attention and avoidance | 19 of 21 | \baudiences?\b|\breaders?\b|news avoidance|\battention\b|\ | |
| Truth and facts | 19 of 21 | \btruth\b|\bfacts?\b|factual|verif\w+|accuracy|accurate | |
| Trust and credibility | 17 of 21 | \btrust\w*|credibilit\w+|\bcredible\b|reliab\w+ | |
| Platforms and distribution | 15 of 21 | \bplatforms?\b|social media|search engines?|\bGoogle\b|\bM | |
| Investigative reporting | 15 of 21 | investigat\w+|accountab\w+|\bexpos\w+|watchdog|corruption | |
| Democracy and civic life | 14 of 21 | democra\w+|civic|elections?|\bvoters?\b|public interest | |
| Misinformation and propaganda | 13 of 21 | misinformation|disinformation|propaganda|fake news|falseho | |
| Ownership and independence | 12 of 21 | independen\w+|owner\w*|proprietor|editorial independence|s | |
| Artificial intelligence | 11 of 21 | \bAI\b|artificial intelligence|\bLLMs?\b|large language mo | |
| Money and the business model | 11 of 21 | revenue|business model|subscription|advertis\w+|paywall|fu | |
| Journalist safety and persecution | 10 of 21 | \bkilled\b|murder\w*|imprison\w+|\bjailed\b|detained|perse | |
| Collaboration and alliances | 10 of 21 | collaborat\w+|alliances?|partnerships?|\btogether\b|coalit | |
| War and conflict | 9 of 21 | \bwar\b|conflict|invasion|frontline|\bUkraine\b|\bGaza\b | |
| Public service and infrastructure | 8 of 21 | public service|infrastructure|\bcommons\b|public good | |
| Language and access | 7 of 21 | \blanguages?\b|multilingual|translat\w+|mother tongue|\bdi | |
| Press freedom | 6 of 21 | press freedom|freedom of the press|censorship|censor\w*|fr | |
| Climate | 5 of 21 | climate|global warming|\bemissions\b|environmental | |
| Local and community news | 5 of 21 | local (?:news|journalism|newspapers?|reporting)|community | |
| Copyright, licensing and scraping | 3 of 21 | copyright|licens\w+|intellectual property|\bscrap\w+|train | |
| Synthetic content and hallucination | 2 of 21 | hallucinat\w*|\bslop\b|deepfakes?|synthetic (?:media|conte |
Reached for together
- Audience, attention and avoidance + Truth and facts — 17 pieces
- Trust and credibility + Truth and facts — 17 pieces
- Audience, attention and avoidance + Trust and credibility — 15 pieces
- Audience, attention and avoidance + Investigative reporting — 14 pieces
- Audience, attention and avoidance + Platforms and distribution — 14 pieces
- Democracy and civic life + Truth and facts — 14 pieces
- Investigative reporting + Truth and facts — 14 pieces
- Platforms and distribution + Truth and facts — 14 pieces
Never in the same piece
Of the pairs where both themes appear somewhere in the corpus, these never appear in the same article:
- Synthetic content and hallucination + Money and the business model
- Synthetic content and hallucination + Collaboration and alliances
- Synthetic content and hallucination + Local and community news
- Synthetic content and hallucination + Ownership and independence
- Synthetic content and hallucination + Public service and infrastructure
- Climate + Local and community news
- Climate + War and conflict
- Copyright, licensing and scraping + Ownership and independence
- Copyright, licensing and scraping + Public service and infrastructure
- Copyright, licensing and scraping + War and conflict
- Language and access + Local and community news
What they give the reader
This is the count this publication exists to make, and here it is over the people making the case for trusting journalism, in the week they made it.
Twenty-three domains, and no two of the twenty-one pieces point at the same one.
At the level of the institution rather than the hostname there is exactly one overlap:
unesco.org.
There is no shared evidence base here. Twenty-one arguments about why the public should trust
journalism, and the reader is asked to take almost all of it on the author’s word.
That is not an accusation, and it is not unusual: an 800-word commissioned op-ed is not a piece of investigative reporting and nobody promised footnotes. It is the ordinary form, which is the point — this site’s first page has argued since 2025 that the ordinary form gives a reader nothing to walk. Here is the measurement, on the best-resourced writers in the industry.
Every piece, by the number of links it offers
Who is speaking
23 named authors across 21 pieces (2 carry more than one byline), at 26 organisations as their own role lines name them:
AFP, Association of Independent Regional Press Publishers of Ukraine, Brazilian Newspaper Association (ANJ), Coda Story, Covering Climate Now, Daraj.com, European Broadcasting Union, Fondation Hirondelle, Globe and Mail, Haitian Times, Informed, News Corp Australasia, Paraluman News, Project Kontinuum, Quint Digital Limited, The Atlas, The GroundTruth Project, The Quint, Truth Tellers, UNESCO, URL Media, Vocento, World Association of News Publishers (WAN-IFRA), World Editors Forum, ZEG Network, confidencial.digital.
How this maps to what we have published
The striking thing about reading twenty-one of these at once is not disagreement. It is that the diagnosis is shared, precisely, across continents and business models — and that almost nobody proposes a mechanism. The corpus says trust is eroding, truth is contested, AI is taking the work, the model is broken. It is a diagnosis this publication has been making since its first page. The difference is what comes next.
| What the corpus says | What we have argued, and where |
|---|---|
| Trust is the industry’s core asset and it is eroding. 17 of 21 pieces reach for trust and credibility. | Trust is not a brand. It is a walkable chain from a claim to the evidence under it. |
| Misinformation spreads faster than corrections. 13 of 21 pieces say so. | A correction that does not reach what it disproved is not a correction. This is the one mechanism nobody in the corpus proposes. |
| AI is taking journalism’s work without paying for it. 11 of 21 pieces raise AI; 3 mention copyright or licensing. | You cannot be paid for a fact whose provenance nobody can check, and you cannot license what you have not made machine-readable. |
| Readers cannot tell what is true. 19 of 21 pieces are about truth and facts. | Provenance is the product: the chain, not the byline, is what a reader checks. |
| Journalism needs a sustainable business model. 11 of 21 pieces. | The unit that can be sold is the verified fact and its chain, not the page view or the paywalled article. |
| The industry must work together across borders. 10 of 21 pieces. | Collaboration between newsrooms needs a shared, checkable substrate. This section is one: a corpus somebody else published, described so anyone can build on it. |
For an agent
Fetch data/analysis.json: theme frequencies with the pattern that produced each one, co-occurrence pairs, the per-article link counts and the cited-domain overlap, plus the explicit list of what this analysis refuses to infer. Cross-reference data/graph.json for the same facts as typed edges. Every count is re-derived from frozen bytes by build/gates.py. No sentiment, stance or agreement is modelled anywhere in this section, and agrees_with is a banned verb in the ontology — if you need those, derive them yourself and say that you did.