How this site works
Each page takes one disputed question about the Israeli–Palestinian conflict and sets out, next to each other, the case each side makes, with its sources. The site gives no verdict. It is meant for readers who want to work out what they think.
What is covered
The questions are chosen by the editor, from those that come up again and again in arguments about the conflict. They are grouped in chapters, ordered by the period in which each dispute begins.
The site does not follow breaking news. A question is added once there is enough reliable material to set out both cases properly.
The site uses the usual name, the Israeli–Palestinian conflict; the order of the words is convention, not precedence.
It is written for anyone curious to read each side’s case, and the documented facts behind it.
How each side is written
Both sides are written in the same plain voice, and at comparable length where the material allows, so that neither is helped by how it is put. In a debate, how well a point lands often depends on who is speaking; here the aim is to leave only the point.
Each side is built from what its serious advocates actually argue, and every side links to a source where that argument is made. The test is whether someone who holds the view would recognise it as their own position.
Which side appears first changes from page to page. Across the site, neither is shown first more than one time more often than the other. The home page follows each page’s order, so the side shown first changes from one question to the next.
Each side has its own colour: gold for Israel’s side, purple for the Palestinian side. Neither is a flag colour. Blue, green and red carry national associations in this subject, and the site avoids colours that would make the columns read as teams.
Words and names
On this subject, the names of things are often part of the argument: “West Bank” or “Judea and Samaria”, “occupied” or “disputed”, “settlement” or “community”.
In its own voice, the site uses the term most widely used internationally, and names the alternative where it matters. Within each side’s case, that side’s own words are used, as its advocates use them.
Where a page depends on a disputed term, a note on terms says who uses which word. It describes usage and does not say which word is right.
Balance
Both sides get the same treatment: the same voice, the same checks, the same standard for marks. That does not mean the two columns must look alike.
Where the evidence on one side of a point is thinner, the page says so. Nothing is added to make the two sides look even, and neither side’s argument is weakened to make the page tidier.
Where a court or a UN body has reached a conclusion on a page’s question, the page’s summary says so, names who disputes it, and says whether any court has ruled.
Organisations that campaign for one side are named as advocacy groups in the text, whichever side they support. Organisations whose work is to report on conduct, such as human rights organisations and United Nations bodies, are named for what they are, whichever side their findings concern.
Who does the research
Each page starts from a short, neutral brief written before any research and confirmed by the editor. The same brief is given to AI systems from three companies: ChatGPT (OpenAI), Grok (xAI) and Claude (Anthropic). Their findings are treated as leads, not as sources, and agreement between them is not treated as proof of anything.
The draft is then reviewed by the same systems, asked to look for one-sided wording, overstatement, and claims the sources do not support. Their objections are weighed one by one, and the reasons for taking or leaving each are kept with the research.
The editor reads every page before it is published, in each language separately. Nothing is published, marked, or changed without the editor’s approval. If any part of a page is edited after approval, it has to be approved again before the change appears.
English and French
Each page is written in English first, then translated into French, and the French is reviewed separately. Where a document has an official French version, such as a United Nations text, that version is quoted. Otherwise the quotation is translated and marked as a translation.
How sources are checked
Every factual claim cites at least one source. For each source, a short passage that supports the claim is recorded, and the source is opened to check that the passage is really there. Where that cannot be done automatically, the editor checks it by hand. A quotation that cannot be checked is not published.
Sources are listed at the foot of each page with a link, and a link to an archived copy where one exists. Several passages from one document share one number.
On the list of sources, each news outlet, advocacy group and human rights organisation has one line saying who it is: where it is based, who owns or funds it, who founded it, and, for an advocacy group, its mission in its own words. Each fact in these lines was checked against a source, as on the pages. Where a fact could not be found, the line says it is not established; where an organisation’s funding is disputed, the line gives the claim and the organisation’s answer. The site does not rate its sources.
What the labels and marks mean
Every statement carries one of three labels. A factual claim states what the record shows: an event, a date, a figure, a vote, or what a treaty, a resolution or a court decided. A position reports what a named speaker says, and the label names the speaker: a government, a party, an armed group, an organisation, a scholar or a witness. An interpretation is a side’s reading of the facts, and the site does not rule on interpretations.
A position shows that someone said something, not that it is true. Governments, armed groups and campaigning organisations on both sides state their positions to persuade; the label shows whose words they are. Where an organisation or an investigator reaches a judgement, such as that a crime was committed, the page shows it as that body’s position; what the body counted or observed is a factual claim. What a court decided is a factual claim.
Where a claim’s only source is a party to the conflict (a government, its army, an armed group, or a body one of them runs) and no independent source confirms or contradicts it, the statement says so. That note tells you whose word the claim rests on; it does not say the claim is wrong.
A factual claim, or what a position states as fact, can carry one of two marks. “Marked as factually wrong” means the sources contradict it. “Marked as misleading” means it is accurate as worded but leaves out context that changes its meaning: what the sources, its own included, say about the very thing it asserts, not the answer to a different question.
A mark applies to one statement, never to a side as a whole. It requires at least two sources that are not advocacy for either side, and each mark shows the sources it rests on. The same standard applies to both sides. Pages described as a disagreement about interpretation carry no marks at all.
How many marks fall on each side depends on which questions are covered and on what each side states as fact. It is not a score of which side is right. A statement without a mark has not been endorsed by the site: it means only that the test for a mark was not met.
Corrections and changes
Each page shows when it was last reviewed and, once it has changed since publication, a history of the changes made to it.
Anyone can report an error or suggest a source, from the link at the end of each page or from the contact page. A correction is checked against the sources in the same way as the page itself: if it holds, the page is changed, approved again by the editor, and the change is listed in its history.
Limits
This is a small project with one editor. AI research can miss things or get them wrong, and so can the editor. A page shows the strongest version of each case that was found, which is not the same as everything that could be said.
Each page sets out two cases, but neither Israelis nor Palestinians speak with one voice. A side here is the case its serious advocates make, not the view of every person on that side.
The AI systems that help with the research learned mostly from English-language material and may share its blind spots, and reviewing one another does not remove a bias they hold in common. That is why the editor reads everything, and why corrections from readers on both sides matter.