Most news outlets ask you to trust them. We ask you to scrutinize us.

That is not a slogan. It is a design decision. Every article Reed publishes has a transparency panel showing exactly how our articles are created. Which sources we used, which steps we took, what we know and can confirm, but also what we do not know. This article describes how that process works.

From source to article

Reed monitors over 1,000 news sources. When something happens in the world, dozens of outlets often write about the same event. Sometimes with the same facts, sometimes with different ones, sometimes with contradictory information.

Our process looks like this:

First, we collect reporting from all sources covering an event, then we cluster the articles. We identify which articles are about the same thing. Next, we extract facts, locations, timestamps, and people involved from the collected material. We do research -- background information about people, organizations, and places that are mentioned. Then we write our own article based on the collected material. Finally, we review our own text against the source material, independently of the writing process.

That last step is the most important: we fact-check ourselves.

The two-source principle

Reed does not publish an article based on a single source. We require at least two independent sources before publishing. With one exception: official government statements, since they constitute primary sources in themselves.

This means we are sometimes slower than others. An event that only one newspaper has reported on will not be published by us until more sources confirm it. That is a natural consequence of how we work -- we would rather be right than fast.

Fact-checking with RAG review

When an article is written, it goes through an automated review that operates independently of the writing process. The system extracts the central claims in the article and then searches for support in the source material.

This is done through what is called RAG review (Retrieval-Augmented Generation), a process where the system independently searches through the actual source material to find support, or lack of support, for each individual claim. It is not the same AI that wrote the text that reviews it, but an independent process that matches claims against the original reporting.

Each claim is assessed as confirmed, unconfirmed, or disputed.

If the review finds problems, there are three possible outcomes: the article is approved, the article is rewritten with corrections, or the article is escalated for manual review. An article may be rewritten at most twice. If it does not pass after that, it goes to manual review.

This means we never publish anything that has not passed our own fact-checking.

Five confidence levels

Not all information sources are equally reliable -- everyone who reads news knows this. But it is rarely shown. Reed shows it.

Every article has a confidence level on a five-point scale:

• Confirmed -- Multiple independent sources, verified information.
• High -- Official sources, consistent reporting.
• Medium -- A primary source, reasonable but not independently verified.
• Low -- A single source, conflicting information exists.
• Unconfirmed -- Circulating but not verified.

We believe it is better to publish a story with a clearly marked low confidence level than not to publish it at all -- provided that the reader truly knows what they are getting.

What we don't know

This is perhaps the most important section in every Reed article, and the most unusual.

Every article has a section called "What we don't know." It lists the questions we could not answer (if any exist), the information we lack, the perspectives missing from the reporting.

It was a deliberate design decision. Journalism is not just about what happened. It is about what we can actually be certain of, and being honest about the limits of our knowledge.

AI with open cards

Reed uses AI in every step of its editorial process. From identifying events in source material to writing articles, extracting quotes, and checking facts in the finished text. We know this raises questions -- can you trust AI-generated journalism?

Our answer is that you should not have to trust us. You should be able to scrutinize. Every article shows which steps were taken and which sources form the basis. If something is wrong, you should be able to see where the error occurred.

Most newsrooms that use AI to some extent (which is most of them) do not tell their readers about it. We do the opposite. Not because we have to, but because we believe transparency is the only way to earn trust.

How we handle errors

We will make mistakes. Every newsroom does. The difference is what happens next.

Every article has a changelog showing how the text has evolved over time. If we correct a factual error, it is visible. If the confidence level changes, it is visible. If new information is added, that is visible too.

We do not erase mistakes. We show them.

What we don't do

We don't take political sides. Reed does not try to convince you of anything. When a story is reported differently by different outlets, we try to identify what is common -- what is likely true -- and be clear about what is disputed.

We don't sell your attention. Reed has no ads. Your behavior in the app is used to give you more relevant news -- not to show you more relevant ads.

We don't hide how we work. If you wonder how a specific article was created, open the transparency panel. Everything is there.

Scrutinize us

We are not asking for your trust. We are asking you to look at how we work and judge for yourself.

Open an article. Open the transparency panel. Look at the sources. Read what we know -- and what we don't know. Compare with the original sources if you want.

That is the whole point of Reed.