How we work

Unowned turns scattered reports and footage into a clear, checked record. These are the rules every item follows.

Where our information comes from

Unowned does not have reporters on the ground yet. Most of what we publish is checked against established news organisations and wire services, and against official statements. Posts and videos from people who were there appear in From the ground: we describe them, say who posted them and which outlets reported on them, and link to the original. A post on its own is never enough to label anything confirmed.

Many of the outlets we cite are owned by companies, governments or wealthy individuals. We show who reported what on every item so you can judge for yourself, rather than ask you to trust us.

Sourcing

1. Every item links to where it came from.
2. We link to original posts rather than re-uploading them, and say who posted them and when. We check for old or recycled footage before using any clip.
3. Under each item, a line says why it has its label, for example "Confirmed by independent reports from AP and Reuters".

Labels

• Confirmed Two or more independent sources report it. Sources are independent when they did their own reporting. When several outlets publish the same wire story (for example AFP or Reuters), that counts as one source. So do outlets owned by the same company, such as franceinfo and ICI (both Radio France).
• Reported One source so far. We are looking for a second.
• Disputed The accounts conflict. We give each side in its own terms, with its sources, and say what is still unknown.
• Alleged One side's claim that has not been independently confirmed, such as an accusation by a party, a group or the authorities.
• Checking Expected or unresolved, such as a court hearing or a planned march. We replace it when the outcome is known.

What "confirmed" does and doesn't mean

• When we write "police say 150 were detained" and label it confirmed, we have confirmed that police gave that figure, not that the figure is right. Official figures are always attributed, and where other counts differ we show them.
• Confirmed means independent sources agree. It is not a guarantee. If later reporting shows we were wrong, we change the label and say so on the corrections page.
• Labels can go down as well as up. An item moves from reported to confirmed when a second independent source appears, and back if we find the sources were not independent.

Fairness

• Every disputed event shows both accounts, in their own terms.
• We describe actions, not motives, and keep commentary out of the record.
• We don't name ordinary protesters or private individuals.

How AI is used

Official statements and social media

We also read what the people in our stories publish themselves: government press offices and ministries, and the movements' own websites and channels. We list only accounts we have confirmed belong to them (linked from their own website, or identified in news reports). We never embed or play their posts: we read them as sources. We also keep a track record for every account we quote: how many of its earlier reports were later confirmed by independent reporting, and how many turned out wrong. It is shown next to its reports, like knowing which reporters' scoops usually hold up; a weak record sends the account's posts to our editor first. To find the voices around each story the way a social feed does, we follow who our verified accounts mention, repost and tag, and who with real reach is posting about the story; our editor verifies each new account before we read it. Posts by professional reporters count like their outlet: one source, so reported. A strong record never makes a single post confirmed. When a group or authority announces a protest, strike or hearing for a coming date, we list it in advance as pending, and replace it with what happened. Text posts are used as written; videos (YouTube, Instagram, TikTok, X) are transcribed and translated to English; text in images and posters is read and translated. Then the result is checked like any other source. These items are labelled "Official statement" or "Their own channel", always say who said it, and are graded strictly: an announcement of their own action is reported; anything they say about others is alleged; photos and video they post go in "From the ground" as alleged, with what we could not verify; and nothing from a single statement or post is ever confirmed until independent news reports agree. Claims of deaths, injuries or abuse are never published automatically.

AI finds, translates, sorts, checks and drafts our updates. Every hour, our newsdesk reads reporting from established news outlets only and drafts updates; two further AI checks by different services, plus automatic checks that every quote, number and name appears in the cited articles, must all pass before an update is published, so stories stay current. Until our editor has re-checked an update, it says so under the item. Within hours (five times a day), a stronger AI editor re-checks every one of those updates against its sources, improving, correcting or removing anything that falls short. Anything sensitive is never published automatically: an accusation against a named person, details that could identify a minor or a private individual, claims of deaths, or anything resting on a single unverified post goes to that editor first, and is left out unless it meets our safety rules. Mistakes are corrected in public on the corrections page.

Corrections

We correct mistakes quickly and in public. Every change made after publishing is listed with its date. Report an error.