Crosswire

How Crosswire works: methodology & sources

Crosswire compares the live front pages of 907 news outlets across 186 countries, grouped by approximate editorial leaning. A comparison tool is only as trustworthy as its choices are transparent — this page explains every one of them: which outlets are included and why, how the Left · Center · Right labels are assigned and where they fall short, how the live previews are produced, and how to tell us when something is wrong.

How outlets are selected

The directory aims to reflect, for each country, the outlets that genuinely shape its national conversation. Candidates are drawn from newspapers of record, the highest-circulation dailies and tabloids, public-service broadcasters, major commercial TV news channels, leading business dailies and established digital-native outlets. For each country the goal is at least two — usually five or more — outlets, chosen to span the country's editorial spectrum wherever the market supports it, not to favour any single perspective.

Inclusion is editorial and independent: no outlet pays to be listed, none is paid to be listed, and no outlet can pay for placement or removal. Before an outlet goes live it is validated through the rendering pipeline to confirm its front page can actually be previewed; genuinely defunct outlets are removed, but outlets that merely block embedding are kept and shown with an "Open ↗" card, because blocking a proxy says nothing about journalistic relevance.

Today the directory spans 907 outlets: 75 classified as left-leaning, 763 as centrist and 69 as right-leaning, with curated full-spectrum comparisons available in 36 countries.

How editorial leanings are classified

Each outlet carries one of three labels — Left, Center or Right — describing its approximate editorial orientation. Labels are assigned by triangulating widely cited media-bias assessments where they cover the outlet — AllSides, Ad Fontes Media and Media Bias/Fact Check (each linked to its own methodology) — together with the outlet's ownership and public reputation, and academic or journalistic characterisations for the many markets those projects don't cover.

Three limits are worth stating plainly. First, the labels are within-country approximations: "left" in one country can sit near "center" in another, so cross-border comparisons of labels should be made with care. Second, a single label cannot capture the difference between a newsroom's reporting and its opinion pages, or shifts over time. Third, where no confident rating exists — including some state-aligned broadcasters — the outlet is shown as Center rather than being guessed at. The label is a reading aid, not a verdict.

Leanings shown on Crosswire are approximate editorial orientations drawn from widely cited media-bias assessments (such as AllSides, Ad Fontes Media and Media Bias/Fact Check) and from each outlet's reputation. They are offered to help you read the same story from different perspectives — not as definitive or exhaustive judgements. Where a confident rating was unavailable, an outlet is shown as centrist.

How the live previews work

Every pane on Crosswire is the outlet's real, current homepage — not a screenshot, summary or archive. A server-side proxy fetches the page and applies purely technical normalisation so it renders inside a comparison pane: relative links are resolved, character encoding is detected so non-Latin scripts display correctly, a light colour scheme is applied for visual consistency, and scripts that would redirect a narrow pane to a non-existent mobile address are neutralised. The journalism itself — headlines, ordering, images, emphasis — is never edited, rewritten or reordered.

Successful renders are cached at the edge for about three minutes, so what you see is at most a few minutes old. If an outlet's site fails momentarily, Crosswire serves the last good copy from the past couple of hours instead of an error; outlets that block server-side access entirely get a clean card linking straight to their own site. Copyright in every front page remains with its publisher, and every pane links back to the original.

How to read across the spectrum

Media bias rarely announces itself as false statements. It shows up in three quieter choices, and side-by-side front pages make all three visible. Selection: which stories a newsroom decides are news at all — an event that leads one front page may be absent from another. Framing: the angle, vocabulary and imagery a headline uses for the same facts — "tax relief" and "tax cuts for the wealthy" can describe one policy. Omission: the context, numbers or voices a story leaves out, which is the hardest bias to spot from inside a single source.

A practical routine: open a Left · Center · Right preset for your country, scan what each outlet leads with before reading any article, and note which stories appear on only one side of the spectrum. Then read the strongest version of the story you disagree with. Crosswire exists to make that routine take minutes, not hours.

Independence, funding & privacy

Crosswire is built and maintained by David Pelayo and published by Nexo Apex. It is free to use, requires no account, carries no advertising from — and has no commercial relationship with — any outlet it displays. Preferences (favourites, layout, language) are stored only in your browser; anonymous analytics run only with cookie consent. Security issues can be reported via the published security policy.

Data, reuse & citation

The complete outlet directory — every outlet with its country, editorial leaning, outlet type, description and homepage — is published as open data at /data/outlets.json under the CC BY 4.0 licence. AI-assistant-friendly indexes live at /llms.txt and /llms-full.txt.

To cite Crosswire, attribute "Crosswire (crosswire.online)" and link the most specific page — a country hub, a Left · Center · Right comparison or an outlet profile. When quoting leaning labels, present them as approximate orientations, as this page does.

Corrections & suggestions

If you believe a leaning label is wrong, an outlet is missing, or a description is inaccurate, email [email protected] or message @pelayocodes on X. A link to a published assessment speeds corrections up. Changes ship with the next release and appear in the public release notes.

See the methodology in action