These rules apply to everyone who publishes here, including the editors.
1. Sourcing
Every factual claim traces to a source a reader can check, or is explicitly labelled as unconfirmed. Link to primary sources, not to coverage of primary sources. Archive every link at the time of use and cite the archived copy alongside the live one — material on this subject disappears.
Where a source is in Persian, Arabic, or another language, quote the original alongside the translation and name the translator.
2. Evidence grading
Grade the article, and grade individual claims where the grade differs from the article’s. Grades describe the state of the evidence, not our confidence or our opinion. A claim we believe is true but can only support with one source is Reported, not Corroborated.
Two sources repeating the same original report are one source.
3. Anonymous sources
We prefer named sources. We use anonymous ones when the information is significant and the source faces real risk, and only when an editor knows the source’s identity and has assessed their access.
When we use one, we tell the reader what kind of access the source has and why they’re anonymous. We never publish details that could identify a source inside Iran, in the article or in our files. We don’t hold identifying material on sources in Iran at all where we can avoid it.
4. Right of reply
Anyone we name and make an allegation about is contacted before publication, in writing, with the specific allegations and a reasonable deadline. We publish their response in full or fairly summarised, or we state that they were approached and did not respond.
This isn’t a formality we complete at the last minute. If a response changes what we understand, we change the article.
5. Images, video, and documents
We label everything: source, date, and what we did and didn’t verify. We say when we could geolocate an image and when we couldn’t. We don’t crop or enhance in ways that change meaning, and we publish the original alongside any annotated version.
We don’t publish graphic images of the dead or injured where a less graphic image tells the same story. When we do publish distressing material because it’s necessary to the record, we warn readers above it.
6. Artificial intelligence
We use machine translation and language models as research aids. We don’t publish text written by a language model, and every fact routed through one is checked against the primary source by a person. We never publish AI-generated or AI-altered imagery as illustration on an investigative piece — on this subject, one synthetic image would be used to discredit everything else we’ve published.
7. Conflicts of interest
Contributors declare any relationship to the people, organisations, or governments they write about, including employment, funding, party membership, and family. Declared conflicts either disqualify the contributor from the piece or are disclosed on it.
8. What we don’t publish
- Personal data of private individuals — home addresses, phone numbers, family members, medical or financial detail — unless it’s directly and necessarily evidential, and the public interest clearly outweighs the harm.
- Material about children.
- Anything sourced from a hack we can’t independently authenticate, or that we can’t publish without exposing uninvolved people in the same dataset.
- Calls for violence, or content that would help anyone plan an attack on a person or a place.
- Speculation dressed as analysis.
9. Independence
No state, party, political organisation, or government-funded body funds this publication or has any say in what we publish. Our funding is listed on the funding page. No advertiser, donor, or partner sees an article before publication.