How Histor works

Histor is a digital newspaper that uses automation to monitor hundreds of sources, cross-check them and write news stories. The automation serves one unbreakable rule: we never publish as fact anything we cannot verify. When speed and accuracy conflict, we choose accuracy.

1. Where information comes from

We read public, lawful sources: official bodies (gazettes, central banks, seismological services, ministries), news agencies, news outlets with identifiable newsrooms, universities and scientific journals. We only use their headlines, summaries and data to cross-check, we always link to the original and we never copy their articles.

We respect robots.txt, terms of use and each publisher’s text-and-data-mining reservation. We never bypass paywalls or blocks.

Social networks (X, Reddit, Telegram, TikTok…) are leads, not evidence: they tell us something may be happening, but they are never enough to publish.

2. How we verify

Each event gathers every piece that mentions it. For each claim (who, what, when, where, how many, who said it) we look for the sources that support it and those that contradict it.

We count INDEPENDENT sources, not links: if ten outlets repeat what one agency said, that is one source. Two outlets from the same group, or copied texts, do not count as independent either.

We compare dates, places, names and figures; we detect old stories resurfacing, recycled photos and quotes nobody said or that are attributed to the wrong person.

From all this we compute an internal confidence level. Only what passes a demanding threshold and has at least two reliable independent sources is published as fact; a single source, even an official one, is never enough.

3. Confirmed, attributed or unconfirmed

When something is confirmed, we report it as fact and say who confirms it.

With a single source we do not publish: we wait until another independent source reports it. When several sources report it but do not agree on everything, we publish it as “developing” and attribute each detail: “according to X…”. We never turn it into fact until it is confirmed.

What we cannot confirm appears in its own section: “What is not yet confirmed”. If sources contradict each other, we say so.

Every story includes a “How we verified this story” box with what is confirmed, who confirms it and what is still pending.

4. Who writes

Stories are produced by automated news desks (for example, “Histor Newsroom”). They are not people and we do not pretend they are. If a journalist ever signs a story, it will be a real, verifiable person.

When we use AI models to write, they may only use the sheet of verified facts. A validator then checks sentence by sentence that every name, number, date and quote exists in the sources. If anything does not match, the text is discarded.

5. Checks before publishing

Every story passes eleven checks: facts, sources, duplicates, dates, names, quotes, location, language, SEO, legal and a final editorial check. If a critical check fails, it is not published.

Headlines must reflect the content: no sensationalist or misleading headlines, not even in A/B tests.

6. After publication

A story does not end when it is published. We re-check it periodically: if new facts emerge, we update it and show “Updated at…” with the change history.

If we get something wrong, we correct it openly with a note explaining what changed. If a story turns out to be false, we retract it but keep the text visible, marked as retracted, for transparency.

7. Viral is not true

We measure separately how fast a story is growing (its “momentum”) and how much evidence it has. Momentum never increases confidence: a viral story without evidence is not published, however widely it circulates.

8. Images

We only use licensed or permitted images with attribution. We never present AI-generated images as photographs of an event. File or illustrative images are labelled as such.

9. Your help

If you spot an error, tell us via “Report an error”. Every report becomes a review task.