Sources & methodology

What we publish

For each story we show a headline, a short discovery excerpt, and — once a story has been converted — an original summary written in our own words, always with clear attribution to the source publisher and a prominent link to the original article. We do not republish source articles in full. Where a summary has not yet been produced, the story shows only a short excerpt and a link to the source, and is not indexed by search engines.

How we select sources

Stories are aggregated from public news feeds and a partner news API spanning six continents. We prioritise established outlets with a masthead and a corrections process. Origin note: newsGlobal began as a Nepal-focused project and grew into a worldwide feed; our coverage today is global, with Nepal as one region among many.

What we exclude

• Non-English-only items with no usable English rendering.
• Sources on our exclusion list for reliability or licensing reasons.
• Content we cannot attribute to a named source with a working link back.

How we use AI

Summaries are drafted by a language model running locally and are constrained to the facts of the source — the model is instructed not to introduce claims, names, numbers or quotes that are not in the original. Every summary then passes a deterministic quality gate (length, source attribution, and an anti-copying check) before it can be published. No model decides whether a story may be published; publication rights are enforced by fixed rules, not by AI. Classification into sections is model-assisted and audited against a labelled test set.

Corrections

If a summary is inaccurate, a source is misattributed, or a publisher wants a story removed, email [email protected]. We act on takedown and correction requests promptly. See also our about page.