Source scoring methodology

The four criteria

Transparency

Who stands behind the source. Disclosure itself weighs most; the rest are verifiable traces: a named beneficial owner, a legal entity, a media-registry record, disclosed funding.

Component	Points
Ownership disclosure	55
Beneficial owner named	15
Legal entity on record	10
In the media registry	10
Funding disclosed	10

Editorial practice

Editorial behaviour: bylines, news–opinion separation, corrections practice, sourcing. Today this is a curated assessment from our research; where it does not exist yet, the criterion is marked unscored rather than zero.

Component	Points
Curated editorial assessment	100

Signal reliability

The technical honesty of the stream: regular output, full texts, stable collection. Components that do not apply to a source (say, full-text share for a social-only source) redistribute their weight instead of zeroing the score.

Component	Points
Activity: articles past 7 days	40
Full-text share	30
Fetch health past 7 days	30

What one source is

A source is a newsroom, not a feed. When an outlet publishes the same story to its site, its Telegram channel and its YouTube channel, that is one source with three channels: one score, one voice in an event's coverage. So «2 sources» under an event means two different newsrooms, not two feeds of one. Every channel is listed on the source's own page, and the register shows the count beside the code.

How we count media groups

Since 29 September 2026 one question sets the media groups: who pays for the newsroom. Before that, Viche sorted newsrooms into seven groups that mixed three different questions — who pays, whether the source is a public authority, and what format it writes in. Those are now separate fields, and the seven groups became six groups by who pays.
A group belongs to the newsroom, not to a single report: all its channels and reports belong to the same group.
The media groups do not form a scale: none stands «left» or «right» of another. So all six are always shown in one and the same order.
The strip under an event is six equal segments, one per media group. A segment is filled when a newsroom of that group is among the collected reports; how many reports it filed does not change the strip. «Covered by 3 of 6 media groups» is the number of filled segments. The shares — how many sources and reports each group accounts for — are shown only in the table on the event's page.
An empty segment means the collected reports include no newsroom of that group; it does not prove a deliberate silence.
A newsroom we have not classified yet counts toward no group: it is named on its own — «1 of 6 newsrooms not yet assigned to a group». «Undisclosed» is a finding about the outlet: it does not say who pays for it. «Not classified yet» is our gap: we have not checked yet.
Earlier versions of Viche called the media groups «blocs», and there were seven of them.

Criteria

These are the rules by which we put each newsroom in a group. They only read what is already known about the outlet and add no new claim about its ownership. They are structural criteria, not a judgement of quality.

Who pays

Every newsroom belongs to one of six groups by who pays for it:

Budget: — A state or local budget: public authorities, state and municipal media.
Public money: — Public service broadcasters: public funds and editorial independence secured by their institutions.
Media business: — A commercial owner with no ties to a large non-media business group or to politics.
Big capital: — The ultimate owner is a large non-media business group or political capital.
Readers and grants: — A non-profit newsroom, or money mostly from grants, donations and subscriptions.
Undisclosed: — The outlet discloses neither its legal entity, nor who is responsible, nor who pays.

In «Big capital» we also mark where the capital comes from: oligarchic ties — a large non-media business group; political ties — a party, a sitting politician or their family. Outside that group we mark no owner's ties.

What it does

The second field is what the source does with the news. It does not change the strip; it shows in the share table on an event's page and on the source's page:

Actor: — the source is itself a public authority or a command: a primary source, not a medium.
Newsroom: — a daily news feed, or original news and investigations.
Commentary: — analysis, commentary or fact-checking without a daily news feed.

The evidence standard

No ownership claim without a source: the register of legal entities, the media register, the outlet's own imprint or «About us» page, media-ownership monitoring, or an investigation by a newsroom that answers for its text.
A trace nobody has confirmed (a domain-owner snapshot on a third-party site, an anonymous post) does not go into the profile.
«Undisclosed» comes only after a check: we went through the outlet's site and the registers and found no legal entity, no one responsible and no one who pays. A public, named person in charge without a disclosed ownership structure is not «Undisclosed».

Border rules, September 2026 version

Decisions for the cases the rules above do not settle (29 September 2026, after 341 outlets were checked):

Former politicians. «Big capital» with political ties only if the owner held an elected mandate in the last five years or holds a political or state office now. An older mandate is «Media business»; the history goes into the decision's reasoning.
Proof of a tie. A political tie counts only when a register, an official form or a signed publication names the owner (or a member of their family) as that politician. A matching name or a cautious «probably linked» goes into the reasoning but does not set the group.
Foreign state money. «Public money» when the law guarantees editorial independence (the BBC, Radio Liberty, Voice of America). «Budget» for a foreign government's own outlet without that guarantee. An NGO or diaspora paper with a foreign state's grants is «Readers and grants».
Mixed money. Between «Media business» and «Readers and grants» the owner's legal form decides: an NGO or a foundation is «Readers and grants»; a company or a private person is «Media business» unless the outlet publishes a revenue split in which grants and donations are more than half. A subscription-only newsletter, and a volunteer project on donations with no advertising and no company, are «Readers and grants».

How to appeal a classification

An outlet can appeal its group or any other field of its classification. Write to [email protected] with the outlet's name, the group you contest and your evidence. We check it by the same standard and record the decision — change or keep — with its date and reason; every profile's history is kept.

How we measure

Besides the media groups, story and source pages show how news spreads: whose report we collected earliest, who picked it up, where they write from and for whom. We count this from when and where reports were published and from what we know about each newsroom. It measures behaviour; it does not rate quality or truthfulness.

Speed

The first newsroom is the one whose report on the story came out earliest among the reports we collected. That does not prove it was the first to learn of or report the story at all: an early report can describe the run-up to an event rather than the event itself. Every other newsroom is either «in the wave» or «caught up».

The wave is the two hours after the first report. We measured this window on the stories of the last 30 days: in half of them the second newsroom arrives within 96 minutes of the first, and a third of all later newsrooms arrive within two hours.

Next to the band is the lag: how many minutes, hours or days after the first report the newsroom's first report came out.

Distance

Distance shows how far the newsroom stands from the event:

«the source itself»: — an authority or a command reporting the news itself;
«local»: — a regional newsroom from the same oblast as the event;
«national desk»: — a national newsroom, or a regional one from another oblast;
«interpreters»: — analysis or commentary without a daily news feed;
«from abroad»: — a newsroom abroad, or one that writes for the world.

When we do not know where the event happened or whom the newsroom writes for, we show no distance.

Channel

The channel is where a report came out: a website, Telegram, TV and radio together with video channels, an agency, a weekly or an analytical outlet, a social network. «Telegram only» means every report on the story we collected came out on Telegram: no website, broadcast or agency.

Geography

«Who the outlets write for» counts the newsrooms that covered the story by whom they write for: a community, a region, the whole country or the world. It is the newsroom's intended audience, not how many people read the report.

«No collected reports from» lists large national newsrooms with no report on the story among the reports we collected. We take the 30 national newsrooms that covered the most stories over the last 30 days, and show the list only when at least three newsrooms covered the story. Absence does not prove intent: a newsroom may not have got to it, may not have thought it mattered, or our collection may have missed its report.

«National newsroom coverage», on an oblast's page: the share of the stories placed in the oblast over the last 30 days that reached at least one national newsroom.

A newsroom's role

For every report we check whether it repeats an earlier report on the same story from another newsroom: forwards are visible directly, rewrites by how close the texts are. For each week a newsroom gets one of four roles:

Primary source: — The source of the news itself: an authority or a command, not a media outlet.
Reporter: — Mostly writes its own: fewer than half of its compared reports were rewritten or forwarded.
Rewriter: — Mostly retells others' reports: half or more of its compared reports were rewrites.
Relay: — Mostly forwards others' posts: half or more of its compared reports were forwards.

«Derivative reports» on a source's page is the share of rewrites of other newsrooms' reports among its compared reports over the last 8 weeks. Forwards are not counted: a forward names its source, and shows in the «Relay» role. The role and the shares describe how a newsroom works, not whether it is right.

How we count the rubrics

The league, the archetypes, a source's card, a story's family tree, the media weather and the media diet are the same measures of speed, rewrites and tone, shown another way. They describe how newsrooms work and in what tone they write, not which of them is right.

League of first sources

The league counts whole weeks in Kyiv time, Monday to Sunday. It includes the public sources we count toward coverage.

The place is by the number of events in which the source's report was the earliest we collected. That is not necessarily the first report of the event: we see only what we collected. It is an absolute count, not a share, so a source that publishes more is more often the earliest; the place is not a quality rating. The same number, the same place. Under the source's name is how many of its reports we collected over the week.

«Repeats» is how many reports by other newsrooms over the week were similar to, or forwarded from, the source's earlier reports. We see similarity by how close a text is to an earlier report on the same event, and a forward directly.

«Similar» is the share of the source's compared reports over the week whose text is very close to an earlier report by another newsroom. Forwards do not count: a forwarded post names its author itself. The league does not yet show how many reports were compared.

Below the table are the sources with a high share of reports similar to earlier ones: the great majority of the reports compared over the week, when there were a good number of reports that week. The exact thresholds are stated on the league's page beside this list; they are provisional, and we will review them once more weeks are in. Textual similarity does not prove rewriting: reports based on the same press release or agency story are similar too, and for commentary and analysis retelling others' reports is ordinary work.

Archetypes

Each report a source published over the week that we compared gets one of three archetypes. Courier: a report that came out first or in the wave, within two hours of the first report, and repeats no one. Detective: its own too, but later. Echo: a rewrite or a forward of another newsroom's earlier report.

A source's page shows the archetypes' shares over the last complete week. Reports that were compared with nothing are not counted.

Three more archetypes describe the source itself, not the week: Herald, an authority or a command reporting the news itself; Philosopher, commentary or analysis without a daily feed; Ghost, a source that does not disclose who pays for it.

Source card

The card shows five figures over the last 8 complete weeks, each in its own unit and with no total. Under each is the share of the other sources that count toward coverage and have the figure whose value is lower; a source with the same value is not counted as lower. It is a comparison, not a quality ranking.

Speed: in what share of the stories the source reported it came first or in the wave, within two hours of the first report.

Originality: what share of the source's compared reports were neither rewrites nor forwards.

Transparency: the transparency criterion of the practice score, in points of 100: ownership disclosed (55 points, partly: 30), beneficial owner (15), legal entity (10), a media registry number (10), at least one named funding source (10).

Framing tone: the mean tone of the source's reports from −1 to +1: 0 is neutral, lower is gloomier, higher is brighter. Tone does not say whether the source is right, nor rate its quality.

Post views: the total views of the source's posts in collected events, on Telegram, YouTube and X (impressions there). It is not a count of people. We compare only with sources whose posts have view counts; the others show a dash: we do not know, not zero.

A story's family tree

The family tree shows a story's reports in order: the earliest we collected first, then forwards and reports similar to earlier ones, and how long after they came out. The earliest collected report is not necessarily the first report of the event.

We see a forward directly: Telegram names the channel a post was forwarded from. «Similar to an earlier report» is a report whose text is very close to an earlier report on the same story by another newsroom; it sits under the report it is closest to. Similarity is not proof of rewriting: reports based on the same press release, agency story or another shared source we do not collect are similar too.

«No matched predecessor, or not compared» means we found nothing the report repeats: it may have been written independently, taken from a source we do not collect, or never compared. It is not proof of originality. The tree shows up to the story's first 200 reports; when there are more, the page says so.

Media weather

The weather shows tone, not who is right: how gloomily or brightly the media groups wrote about the day's stories.

For each group we take the day's reports with a tone, average each source's tone and then average across sources, so each source weighs the same. Tone runs from −1, very gloomy, to 1, very bright.

The thresholds come from the data: over the preceding period we rank every day of every group that had enough reports, and split them into thirds. Storm is a day in the gloomiest third, clear in the brightest, cloudy in between. When a group had too few reports in a day, there is no forecast. Under the forecast are its day's thresholds and the period they come from. The line beside it is the group's mean tone over 7 days.

Media diet

The media diet in «Picture» shows what the stories you opened were made of. We know which story you opened, not which report you read, so each story is one plate and all weigh the same.

Each report of a story goes to one part of the plate: Vegetables, a report by the newsroom that reported first or by the authority itself; Spices, commentary or analysis; Sugar, a rewrite or a forward; Protein, a compared report that repeats no one. Reports that were compared with nothing are not counted.

We count up to the last 500 stories opened in the chosen period. On a reader's page the plate shows only while they show their «Picture».

Four honesty rules

We do not score truth. The word “credibility” does not appear in this product: a source scoring 85 can be wrong, a source scoring 40 can be right.
The formula is public. Every source's every point decomposes into components on its page. If the arithmetic cannot be shown, the number is not shown.
Missing data is not bad practice. An unscored criterion is marked and excluded from the mean; it never counts as zero.
No movement arrows until score history exists. “Better/worse than last month” arrives only with stored monthly snapshots.

How we measure tone

Tone is how each media group framed the event: how it was written, not what happened. We estimate each report's tone automatically, from the words of its headline (for a post without one, from the start of its text), so every tone figure is an estimate, and for a single report it can be wrong.
For each media group we average each source's reports, then average across the group's sources, so each source weighs the same. The scale runs from −1 (negative framing) to +1 (positive framing). The spread shows how far apart the groups stand on that scale: 0 — framed alike, 2 — opposite.
When the spread is at least 0.3, the «Tone» tab shows each group's estimate on a scale marked −1, 0 and +1, with how many of its reports were estimated. Below that threshold the difference between groups is noise, not a finding: the tab says «No meaningful measured difference found», and the groups' estimates open only on request, with «Show tone estimates».
Tone is drawn in one neutral colour: its direction is shown by the position on the scale and the sign of the number, not by colour, so it cannot be mistaken for the media groups' colours.
«Good» or «bad» is an assessment of the event for Ukraine and its citizens, not a neutral fact. A model sets the label, and it can be wrong.

Two languages

An event's summary, its key facts and a storyline's title are Viche's own words: a language model writes them after reading every report of the event. The English is written by the same model in the same pass, from the same reports — it is not a machine translation of the Ukrainian, it is the same material composed in English. Outlets' own headlines are never translated: the source list shows what each newsroom published, in its own language, linked to the original. An event covered by a single source has no Viche text at all, so on an English page it stays in Ukrainian and is marked as Ukrainian.

The formula

The four criteria

Transparency

Coverage completenessnot scored yet

Editorial practice

Signal reliability

What one source is

How we count media groups

Criteria

Who pays

What it does

The evidence standard

Border rules, September 2026 version

How to appeal a classification

How we measure

Speed

Distance

Channel

Geography

A newsroom's role

How we count the rubrics

League of first sources

Archetypes

Source card

A story's family tree

Media weather

Media diet

Bands

Four honesty rules

How we measure tone

Two languages

Where the data comes from

How we count the rubrics

League of first sources

Archetypes

Source card

A story's family tree

Media weather

Media diet

How we measure

Speed

Distance

Channel

Geography

A newsroom's role