How the Watchdog works

The Watchdog gives each outlet a score from 0 to 10 on a handful of topics. This page explains, plainly, how those scores are produced, what they do and do not mean, and where the method falls short.

What the score means

Every score runs from 0 to 10. It is directional, not a generic reading of bias against neutrality. Each topic measures one specific direction of framing, and only that direction:

0None detected

10Severe

  • Gazaanti-Palestinian framing

    Measures anti-Palestinian framing in Gaza coverage. 0 means none detected, 10 means severe.

  • Lebanonanti-Lebanese framing

    Measures anti-Lebanese framing in Lebanon coverage. 0 means none detected, 10 means severe.

  • Irananti-Iran framing

    Measures anti-Iran framing in Iran coverage. 0 means none detected, 10 means severe.

  • West Bankanti-Palestinian framing

    Measures anti-Palestinian framing in West Bank coverage. 0 means none detected, 10 means severe.

  • UK Politicsvilification of communities

    Ranks outlets by the share of their articles about a community (Muslims, migrants, trans people, Travellers, Jewish or Black people) that frame it, in the outlet's own voice, as a threat, as criminals, as a burden or as an invasion. Reporting what someone else said does not count.

  • Climateclimate misinformation

    Measures the techniques an article uses to misinform about climate: false claims stated as fact, false balance, hidden emissions effects and missing context. Outlets are ranked by the share of their articles with a false or dismissive claim; missing context is shown beside it and does not change the order. Each article is scored from 0 (none detected) to 10 (severe).

This is the most important thing to understand about the number. An outlet whose framing runs the other way scores near 0, because the scale only ever counts framing in the single direction each topic is built to detect. It never goes negative. A low score records the absence of the measured framing, not a reward for leaning the opposite way, and a 0 is not a certificate of neutrality. The score is not a general trust rating.

The outlet panel

The Watchdog scores a fixed, curated panel of 73 outlets. The same panel is scored the same way on every topic, so a score on one outlet is directly comparable with any other. The panel spans five source types:

Western mainstream39 outlets

  • BBC News
  • CNN
  • The New York Times
  • The Guardian
  • The Independent
  • The Spectator
  • Times Higher Education
  • Reuters
  • Associated Press
  • The Telegraph
  • The Washington Post
  • The Times (London)
  • NPR
  • Bloomberg
  • Sky News
  • ITV News
  • Channel 4 News
  • MSNBC
  • ABC News (US)
  • NBC News
  • Deutsche Welle (English)
  • France 24 (English)
  • Le Monde (English)
  • The i Paper
  • Evening Standard
  • UnHerd
  • HuffPost UK
  • LBC
  • Politico Europe
  • The Observer
  • The Jewish Chronicle
  • City AM
  • The Scotsman
  • CBS News
  • Newsweek
  • Los Angeles Times
  • Euronews
  • The Irish Times
  • CBC News

Tabloids9 outlets

  • Metro
  • Daily Star
  • Daily Mail
  • New York Post
  • The Sun
  • Daily Mirror
  • Daily Express
  • Fox News
  • GB News

Israeli6 outlets

  • Haaretz (English)
  • Times of Israel
  • The Jerusalem Post
  • Arutz Sheva / Israel National News
  • Ynet News (English)
  • i24News

West Asian11 outlets

  • Al Jazeera English
  • Al Arabiya English
  • Anadolu Agency (English)
  • Arab News
  • Middle East Eye
  • The National
  • Al-Monitor
  • Press TV
  • Tehran Times
  • IRNA English
  • Daily Star Lebanon

Western alternative8 outlets

  • The Intercept
  • Mondoweiss
  • Electronic Intifada
  • Drop Site News
  • Antiwar.com
  • Common Dreams
  • Truthout
  • Jacobin

The panel does not change from topic to topic. That is deliberate: a fixed panel is what makes the leaderboard comparable across topics and over time.

How an article is scored

Each article we collect on a topic is read by a large language model, Anthropic's Claude. For every check in the topic's rubric the model returns a 0 to 10 score and, crucially, must ground that judgement in a verbatim quote from the article itself. We call these the receipts.

The quotes are stored alongside the score. They are what you see when you click any score on the site: the exact words that led to the judgement, so you can check the working rather than take it on trust.

The individual check scores are then combined into the article's overall 0 to 10 score using fixed weights. The weights do not change between articles. They are listed in full below.

Most articles reach the archive through our automated collection. NewsCord staff also select articles for review by hand, some already in the archive and some the collection missed, with a short private note that points the model at the passages they think matter. The note is never published and only tells the model where to look. The model applies the same rubric and the same weights, records a finding only where the article's own words meet a check's rule, and quotes the article for every finding, so a pointer the text does not bear out produces nothing. When an outlet's page cannot be read automatically, staff supply the article text themselves. When the article has a named author, up to ten of that journalist's other recent articles for the same outlet on the topic, already in the archive and not yet scored, are scored too, automatically, with no note. When the scoring instructions for a topic change, the articles staff selected are scored again the same way, with no note. These scores count on the leaderboard like any other, and each one says on its own page how staff were involved.

Climate works differently in two ways. First, its false-claim checks compare what an article states with a register of catalogued false claims about climate, each with what the evidence shows and the source that says so, such as the IPCC, the Met Office or the Climate Change Committee. The model never decides for itself whether a claim is true. It only says whether a sentence makes one of the claims in the register, and every flagged claim shows that evidence and source on the article's page. Second, Climate is the one topic that scores opinion columns, and on a column only false claims of fact count. Arguing that a policy costs too much is opinion. Stating that the planet has stopped warming is not.

Reader reports

Anyone can send us an article they think gets it wrong, mark the words and say what is wrong with them. We read the page ourselves, and text a reader pastes is never scored. If we cover the outlet, the model scores the article to the same standard as every other one, with the same rubric and the same weights, and it never sees the reader's marks. The reader then sees their marks set beside what the Watchdog found on its own.

Staff read the reports. When staff approve one, the article is scored again with the reader's notes as pointers, the same way staff point the model at passages on the articles they select, and that score carries the staff line on its page. An article a reader brought in that we did not already hold counts on the leaderboard and the reporter boards only once staff approve the report. Until then its score is on its own page and nowhere else.

The flag board counts what readers marked. A mark counts there only when its words appear word for word in the outlet's text as we read it, one reader counts once per article, and staff remove abuse. Each outlet is ranked by flags per 100 articles we scored from it, so an outlet does not rank higher just because more people read it. The board is reported by readers and is not checked by NewsCord.

The checks can also be refined. Guidance for a topic is learned from staff reviews and from readers' marks that a check had missed and the Watchdog then proved. It sits under the rubric and never changes the weights or a check's rules. Every change is tested on articles before it goes live, including readers' marks it was never shown. A change taught by readers' marks goes live by itself only when that test shows it counts more of what the checks had missed, and no more of the marks that did not hold up. The owner can switch automatic changes off, and can retire any change.

The checks, per topic

Each topic has its own rubric. Every check carries a fixed weight, shown here as a percentage of the article score. These are the live weights the scorer uses. Each check is listed by its plain name, with the name it carries in our data after it. Open a topic to see its full breakdown.

Gaza13 checks. Their weights add up to 100%.

Measures anti-Palestinian framing in Gaza coverage. 0 means none detected, 10 means severe.

  1. Hides who did the killing, listed in our data as Aggressor framing

    When Israel kills, is it named as the killer, or does passive voice hide who did it? Violence against Israelis is set aside: hiding Hamas's agency is not the bias this measures.

    16% of the article score

  2. Misleading headline, listed in our data as Headline analysis

    Does the headline hide who killed Palestinians, reduce them to a number, leave out deaths the article itself reports, or claim more than the article shows?

    14% of the article score

  3. Leaves Palestinians out, listed in our data as Palestinian erasure

    Does the article put Israeli victims and captives on the page while Palestinian victims are absent, and never state the scale of Israel's killing in Gaza?

    12% of the article score

  4. Casts doubt on Palestinian sources, listed in our data as Palestinian source qualification

    Are Palestinian sources hedged ('Hamas-run ministry says') while Israeli and Western officials are quoted without qualification?

    8% of the article score

  5. Turns the dead into numbers, listed in our data as Casualty humanisation

    Are Palestinian deaths reported as human lives, with names and stories, or as abstract numbers, and are Israeli victims named while Palestinians are not?

    8% of the article score

  6. Leaves out the law, listed in our data as Legal framing

    How the article treats international law: war crimes, genocide allegations and legal authority.

    8% of the article score

  7. Labels used for one side only, listed in our data as Victim affiliation labels

    Are the dead, or the institution they belonged to, filed under a militant affiliation in the outlet's own voice, with no equivalent label on Israeli forces?

    8% of the article score

  8. Takes army claims on trust, listed in our data as Military claim verification

    Are military claims that the dead were legitimate targets carried without the verification caveat the article applies to Palestinian sources?

    6% of the article score

  9. Gives one side the space, listed in our data as Perspective allocation

    Whose voices are quoted or paraphrased, and does one side get materially more space?

    6% of the article score

  10. Doubts one side’s death toll, listed in our data as Casualty attribution symmetry

    Does each side's death toll carry the same burden of sourcing, or is one side's toll stated as plain fact while the other's is attributed and tagged?

    4% of the article score

  11. Hostages and prisoners described differently, listed in our data as Hostage / prisoner framing

    Are Israeli hostages and Palestinian detainees described with the same care, or asymmetrically?

    4% of the article score

  12. Misuses the Hamas label, listed in our data as Hamas affiliation context

    Are 'Hamas-affiliated' labels applied accurately, or used to justify harm to civilians?

    2% of the article score

  13. Who it blames for the ceasefire, listed in our data as Ceasefire framing

    Who is blamed for blocking or breaking a ceasefire, and how negotiation leverage is framed.

    4% of the article score
Lebanon12 checks. Their weights add up to 100%.

Measures anti-Lebanese framing in Lebanon coverage. 0 means none detected, 10 means severe.

  1. Hides who attacked, listed in our data as Aggressor framing

    Is it clear who carried out strikes and killings, or does passive voice hide the actor?

    16% of the article score

  2. Misleading headline, listed in our data as Headline analysis

    Does the headline editorialise, hide agency, leave out deaths the article itself reports, or claim more than the article shows?

    18% of the article score

  3. How it labels Hezbollah, listed in our data as Hezbollah framing

    How the journalist's own voice labels Hezbollah, from delegitimising 'terrorists' to legitimising 'resistance'.

    8% of the article score

  4. Turns the dead into numbers, listed in our data as Casualty humanisation

    Are Lebanese deaths reported as human lives, with names and stories, or as abstract numbers?

    8% of the article score

  5. Labels the dead as militants, listed in our data as Victim affiliation labels

    Are the Lebanese dead, or the institution they belonged to, filed under a militant affiliation in the outlet's own voice, with no equivalent label on the other party?

    8% of the article score

  6. Takes army claims on trust, listed in our data as Military claim verification

    Are military claims that the dead were Hezbollah members or legitimate targets carried without the verification caveat applied to Lebanese sources?

    8% of the article score

  7. Leaves out the law, listed in our data as Legal framing

    How the article treats sovereignty, international law, war crimes and the UN Charter.

    7% of the article score

  8. Who it blames for the ceasefire, listed in our data as Ceasefire framing

    Who is blamed for breaching the ceasefire, and are Hezbollah violations foregrounded while Israeli ones are downplayed?

    7% of the article score

  9. Calls everyone a militant, listed in our data as Civilian / militant distinction

    Does casualty reporting separate civilians from fighters, or lump everyone together as 'militants'?

    4% of the article score

  10. Calls Hezbollah an Iranian proxy, listed in our data as Iranian proxy framing

    Is Hezbollah reduced to an 'Iranian proxy', or treated as a Lebanese political actor with proper context?

    4% of the article score

  11. Gives one side the space, listed in our data as Perspective allocation

    Do Israeli and Lebanese voices get materially unequal column inches?

    4% of the article score

  12. Doubts one side’s death toll, listed in our data as Casualty attribution symmetry

    Are Lebanese deaths attributed and tagged while Israeli deaths are stated as fact, in the same article?

    8% of the article score
Iran12 checks. Their weights add up to 100%.

Measures anti-Iran framing in Iran coverage. 0 means none detected, 10 means severe.

  1. Hides who attacked, listed in our data as Aggressor framing

    Who is shown as initiating strikes, sabotage and escalation across Israel, the US and Iran, or does passive voice hide it?

    18% of the article score

  2. Misleading headline, listed in our data as Headline analysis

    Does the headline editorialise, hide agency, leave out deaths the article itself reports, or claim more than the article shows?

    14% of the article score

  3. Calls the government ‘the regime’, listed in our data as Regime terminology

    Is Iran's government delegitimised as 'the regime' in the journalist's own voice, or described in neutral terms?

    8% of the article score

  4. Assumes a nuclear weapons programme, listed in our data as Nuclear framing

    Is the nuclear programme presupposed to be a weapons programme, or described neutrally with context?

    8% of the article score

  5. Labels the dead as military targets, listed in our data as Victim affiliation labels

    Are the Iranian dead, or the institution they belonged to, filed under a military or regime affiliation in the outlet's own voice, with no equivalent label on the attackers?

    8% of the article score

  6. Takes the attackers’ claims on trust, listed in our data as Military claim verification

    Are Israeli or US claims that those killed or sites struck were military targets carried without the verification caveat applied to Iranian sources?

    8% of the article score

  7. Who it blames for the ceasefire, listed in our data as Ceasefire framing

    Who is blamed for obstructing or breaching a ceasefire, and are Iranian violations foregrounded over US or Israeli ones?

    8% of the article score

  8. Calls allied groups proxies, listed in our data as Proxy framing

    Are Iran-aligned groups reduced to 'proxies', or treated as actors with their own local political context?

    6% of the article score

  9. Leaves out what sanctions do to people, listed in our data as Sanctions framing

    Context, not scored: are sanctions framed as a clean policy response, or is their impact on Iranian civilians acknowledged?

    0% of the article score

  10. Leaves out the law, listed in our data as Legal framing

    How the article treats international law: sovereign attack, the UN Charter, the NPT and IAEA standards.

    8% of the article score

  11. Doubts one side’s death toll, listed in our data as Casualty attribution symmetry

    Are Iranian deaths attributed and tagged while Israeli or US deaths are stated as fact, in the same article?

    6% of the article score

  12. Turns the dead into numbers, listed in our data as Casualty humanisation

    Are Iranian deaths treated as human losses (named, aged, with a profession or family) as much as Israeli or US deaths?

    8% of the article score
West Bank11 checks. Their weights add up to 100%.

Measures anti-Palestinian framing in West Bank coverage. 0 means none detected, 10 means severe.

  1. Misleading headline, listed in our data as Headline analysis

    Does the headline editorialise, hide agency, leave out deaths the article itself reports, or claim more than the article shows?

    18% of the article score

  2. Hides who did it, listed in our data as Aggressor framing

    Is it clear who carried out raids, killings and demolitions, or does passive voice hide the actor?

    16% of the article score

  3. Turns settler attacks into clashes, listed in our data as Settler violence framing

    Are settler attacks reported with settlers as the named agents, or dissolved into 'clashes' and 'confrontations' that mutualise a one-directional attack?

    8% of the article score

  4. Repeats the permit excuse for demolitions, listed in our data as Demolition and displacement framing

    Do home demolitions carry Israel's 'lacked permits' justification without the permit regime or displacement context that tests it?

    8% of the article score

  5. Turns the dead into numbers, listed in our data as Casualty humanisation

    Are Palestinian deaths reported as human lives, with names and stories, or as abstract numbers?

    8% of the article score

  6. Labels the dead or detained as militants, listed in our data as Victim affiliation labels

    Are the Palestinian dead or detained filed under a militant or 'wanted' label in the outlet's own voice, with no equivalent label on the other party?

    8% of the article score

  7. Takes army claims on trust, listed in our data as Military claim verification

    Are Israeli claims that those killed or detained were militants carried without the verification caveat applied to Palestinian accounts?

    8% of the article score

  8. Doubts one side’s death toll, listed in our data as Casualty attribution symmetry

    Are Palestinian deaths attributed and tagged while Israeli deaths are stated as fact, in the same article?

    8% of the article score

  9. Uses the language of annexation, listed in our data as Occupation terminology

    Does the outlet, in its own voice, adopt annexationist terms ('Judea and Samaria', 'disputed territories') or render settlements as ordinary 'neighbourhoods'?

    7% of the article score

  10. Leaves out the law, listed in our data as Legal framing

    How the article treats the occupation's illegality, the ICJ advisory opinion, settlement illegality and annexation.

    7% of the article score

  11. Gives one side the space, listed in our data as Perspective allocation

    Do Israeli official and settler voices get materially more column inches than Palestinian voices?

    4% of the article score
UK Politics5 checks. Their weights add up to 100%.

Ranks outlets by the share of their articles about a community (Muslims, migrants, trans people, Travellers, Jewish or Black people) that frame it, in the outlet's own voice, as a threat, as criminals, as a burden or as an invasion. Reporting what someone else said does not count.

  1. Vilifies a community, listed in our data as Community vilification

    In its own words, does the outlet frame a named community (Muslims, migrants, trans people, Travellers, Jewish or Black people) as a threat, a criminal group, a burden or an invasion? Reporting what someone else said does not count. A community is who people are, never what they did, and never a state body or a job.

    64% of the article score

  2. Misleading headline, listed in our data as Headline analysis

    Does the headline claim more than its own article shows: an allegation stated as fact, words in quotation marks nobody in the article said, or a past fact put in the present?

    36% of the article score

  3. How it treats politicians, listed in our data as Politician treatment

    Context, not scored: which politicians the outlet defers to and which it scrutinises, in its own voice. Not checked on opinion pieces.

    0% of the article score

  4. Extreme labels for one side only, listed in our data as Populist labelling asymmetry

    Context, not scored: are extremity labels ('populist', 'hard-right', 'far-left') applied to one bloc and withheld from another?

    0% of the article score

  5. Quotes one party far more, listed in our data as Quote space asymmetry

    Context, not scored: is direct-quote space materially unbalanced across the parties?

    0% of the article score
Climate11 checks. Their weights add up to 100%.

Measures the techniques an article uses to misinform about climate: false claims stated as fact, false balance, hidden emissions effects and missing context. Outlets are ranked by the share of their articles with a false or dismissive claim; missing context is shown beside it and does not change the order. Each article is scored from 0 (none detected) to 10 (severe).

  1. False claim about the science, listed in our data as False science claims

    In its own voice, does the article state a catalogued false claim about climate science, such as "no warming since 1998", "it's the sun" or "scientists are divided"? Columns are scored on this.

    16% of the article score

  2. Leaves a false claim standing, listed in our data as Unchallenged false claims

    When a quoted source, a minister included, makes a catalogued false claim, does the article challenge it anywhere, or leave it standing as one side of a debate?

    12% of the article score

  3. Leaves out the emissions, listed in our data as Emissions effect hidden

    When a government or company decision changes emissions (a new oil or gas field, an airport expansion, a rolled-back target), does the article say what it does to emissions or targets? Are green claims checked?

    12% of the article score

  4. False claim about policy, listed in our data as False policy claims

    Does the article state a catalogued false claim about climate policy or technology as fact, such as "green levies are the main reason bills are high" or "heat pumps don't work in the cold"? Columns are scored on this.

    10% of the article score

  5. Misleading headline, listed in our data as Headline misinformation

    Does the headline assert a catalogued false claim, use a dismissive label, or present a climate policy only as a cost?

    10% of the article score

  6. Cost with no context, listed in our data as Cost without context

    Does the article present a climate policy only as a cost, with no savings, cost of inaction or real driver of the cost anywhere?

    10% of the article score

  7. Leaves out climate change, listed in our data as Missing climate link

    Does an article reporting extreme heat, wildfire, drought, extreme rainfall or a major storm mention climate change anywhere?

    10% of the article score

  8. Plays down the harm, listed in our data as Impact minimisation

    When the article itself reports deaths, health warnings or damage, does it still frame the event as a treat or as harmless?

    6% of the article score

  9. Hides who a lobby group is, listed in our data as Misdescribed sources

    Are groups that campaign against climate policy, or fossil fuel trade bodies, quoted without telling the reader what they campaign for or represent?

    6% of the article score

  10. Dismissive language

    In its own voice, does the article use dismissive labels ("eco-zealots", "alarmists", "climate cult") or distancing devices ("so-called climate emergency")?

    4% of the article score

  11. Leaves out what protesters want, listed in our data as Protest without cause

    Does coverage of a climate protest say what the protesters want, or only report disruption and arrests?

    4% of the article score

From articles to the leaderboard

An outlet's score on a topic is the median of its article scores over a rolling 90-day window. We use the median, not the mean, so a single unusual article does not swing the figure.

Climate and UK Politics are ranked differently. Climate is ranked by the share of an outlet's articles in the window with a false or dismissive claim (a catalogued false claim stated, left unchallenged or put in the headline, or a dismissive label in the outlet's own voice). Most climate articles have no finding at all, so the median is 0 for almost every outlet and would hide the ones that publish false claims. Missing context (a heatwave reported with no climate link, a policy shown only as a cost, an emissions effect left out) is shown beside each outlet and does not change the order: it is a fair criticism, but it is not a false claim. Each climate article is still scored from 0 to 10, and its page shows every finding.

UK Politics is ranked by the share of an outlet's articles about a community that frame it, in the outlet's own words, as a threat, as criminals, as a burden or as an invasion. The communities are Muslims, migrants and asylum seekers, trans people, Gypsy, Roma and Traveller people, Jewish people and Black people. Quoting or describing what someone else said does not count, and people defined by what they did, such as convicts or gangs, are not a community. Ranked by the median article score, every outlet sat between 1.2 and 3.2, because the other UK checks fire at much the same rate everywhere. An outlet needs 10 articles about a community in the window to be ranked. Each UK article is still scored from 0 to 10, and its page shows every finding.

Alongside the median we show a consistency range: the middle 50% of an outlet's article scores, from the 25th to the 75th percentile (the interquartile range). A narrow range means the outlet frames the topic much the same way across its coverage; a wide range means its framing comes and goes.

An outlet is only ranked once it has published enough articles on the topic for the median to mean something. Below that threshold it shows as "not enough data" rather than being handed a score it has not earned. The minimums are:

Gaza
8 articles
Lebanon
5 articles
Iran
5 articles
West Bank
5 articles
UK Politics
10 articles
Climate
5 articles

The full panel has only been scored continuously since 22 June 2026, so the 90-day windows are still filling. Expect ranges to tighten and more outlets to cross the ranking threshold as coverage accumulates.

Limitations

The method has real limits. We would rather state them than have you assume they are not there.

  1. It measures one direction of framing per topic, by design. It is not a general trust score and not a verdict on whether an outlet is accurate, honest, or good.

  2. It is generated by an automated model, which can misjudge an individual article, especially opinion pieces, explainers, and very short items. There is no published accuracy figure and no human re-scores every article. The safeguard is transparency: every score links to the quotes it was based on, so you can check any judgement yourself.

  3. Coverage depends on what each outlet publishes and on what we are able to collect and access. An outlet that publishes little on a topic will have a thinner, less settled score.

  4. The score is a measure of the text as published. It is not a claim about any individual journalist's intent, and it does not attribute motive to any writer or editor.

Back to the leaderboard