Sources in view

How we choose.
How we check.

We look beyond the dominant headlines for consequential stories, credit the people behind the evidence and explain what the sources can support.

Consequence beyond the headline cycle

We look for significant developments readers may have missed: what changed, who is affected and what deserves a closer look. Reporting from local and specialist newsrooms, public-interest organizations, researchers and primary documents can widen the picture. Inclusion is an editorial choice; it does not prove that other outlets overlooked a story.

Each daily edition draws from a verified, dated selection in Ori’s Outside the Lead master file. Ori researches the candidates, writes headlines and decks, and supplies the ratings. ChatGPT checks evidence, operates the site and preserves the publication record.

Ori selects the front page from the edition’s verified reporting. Topic and region filters cover the published pool. A story’s score tier and front-page position are separate decisions. Returning stories retain their URL and first-publication date. Saved editions preserve the original selection, ranks and scores.

Independent checking can confirm or challenge a master-file story. It does not add an unrelated story to the front page. A direct instruction from the owner to add a particular story is recorded separately.

How to read a story’s score

Every score carries its rubric version and maximum. Earlier editions retain their original 100-point assessments. Versions 1.9 and 2.0 use 1,000 points. Historical scores retain the precision of the assessment made at the time.

Criterion100-point editionsVersion 1.9Version 2.0
Impact30300300
Emergence / momentum20200200
Coverage gap20200200
Hook15150100
Reader interestUnscoredUnscored50
Evidence10100100
Reach55050

The numeric hook score measures the strength of a concrete, sourced detail. The sentence beneath a headline is the writer’s hook. Under version 2.0, a story with at least 150 impact points can receive up to 50 reader-interest points. The article identifies measured signals such as search interest, pageviews, posting activity or placement in home-country reporting. When signals were not measured, the article says so.

On the 1,000-point scale, Lead starts at 800, Featured at 650, and Brief at 500. On the 100-point scale the thresholds are 80, 65 and 50. Evidence can cap the published score below the component total. Score tiers guide editorial selection. The front page retains its own order.

Coverage evidence identifies who did the check. A publisher panel reports indexed matches. A producer panel records the outlets Ori observed, once per outlet, in the original dated source file. The story labels the publisher’s independent findings separately. An unavailable outlet has unknown coverage. In version 2.0, when panel coverage consists only of AP or Reuters wire copy, the wire-only range is 110–140. Historical assessments keep the limits and evidence standards in force when they were made. A tracker’s findings are attributed separately from the publication’s own panel check.

Provisional ratings identify assessments supplied by ChatGPT. The dated score is an editorial judgment. The article links to its sources so readers can assess the underlying evidence.

Make the original work easy to find

Each brief links directly to its cited reporting and documents. Source titles remain visible, with verified publisher and author credits where available. If attribution has not been verified, we show the source’s actual website address rather than inventing a publisher or byline.

A primary source is an original document, dataset or statement used for a claim. Reporting and analysis interpret or add to that material. These labels describe the source’s role; they do not guarantee accuracy or independence. Sources listed by Ori but not rechecked for publication are marked separately from sources whose content was read.

For single-source original reporting, the brief names the publisher, links to the original bylined report and identifies what the publisher could not independently establish. A qualifying brief retains its evidence cap. We credit original reporting and name the originating publisher when syndication has been verified. A journalist credited in the source list wrote the linked work, not our brief. Several links may trace back to the same reporting, and their number is not a count of independent confirmations.

Check the claim, not just the citation

The publishing process opens sources supporting the headline and central claims. It checks dates, denominators, comparisons, estimates versus observations, and whether a summary goes beyond the evidence. Source entries retain their actual publication dates.

Two articles repeating the same report do not establish independent confirmation of its underlying findings. Official statements establish what an institution said or did; they do not settle every disputed assertion. Material contradictions trigger narrower wording, further research or a hold.

Numbers in a headline or hook must be supported by the listed sources. Source-not-located notes and correction warnings cannot become headline claims. The automated checks enforce structure and evidence floors; they do not independently prove factual truth.

Concrete stakes, precise words

The goal is to draw readers through verified facts, numbers and implications, using a professional, matter-of-fact tone. Sensationalism and clickbait are excluded. The writing guide favors concise headlines with a concrete stake or number, an active verb and a supported consequence; decks carry the second fact and essential qualifications.

Ori’s supplied headlines, decks, “Why this story” blocks and X-post drafts are preserved. Proposed corrections are documented instead of being silently rewritten. A separate publication note can identify a material source qualification while leaving the supplied text visible. Allegations remain attributed. Approved licenses are not confirmed shipments; verified deaths are not necessarily all deaths; linked cases are not automatically separate outbreaks.

Automation is part of the byline

AI assists research, evidence review, writing and publication. Routine editions do not require human pre-publication approval. We do not claim on-the-ground reporting or human review that has not occurred.

Each story links to its evidence, distinguishes event and data dates from publication here, and records important limitations. Stories with an Emergence score of 6 or less display an Older story label and retain their actual dates. Source evidence, edition snapshots, prior factual revisions and run records preserve the audit trail.

All reporting and archives remain free. Free email subscriptions are separate from voluntary financial support; contributing does not change access or control editorial choices.

Correct the record visibly

Supported factual corrections receive a dated note explaining the material change. Retractions retain a notice at the article’s address. Previously published URLs remain available even when a story is no longer in the current master section or front page.

Technical recovery must preserve already published corrections. As distribution channels are introduced, material corrections must also be tracked across feeds, newsletters and social posts.

View corrections & updates

Know the limits

Source review reduces errors; it does not eliminate them. Search coverage can be incomplete, records can change, translation can lose context and automated systems can misread evidence. Attribution and direct source links let readers assess a story for themselves.

“Independent selection” describes the publication’s editorial choices. It does not mean that every source is independent of every other source. When evidence is insufficient for a central claim, that claim must not publish as fact.