How NewsLI works

What NewsLI is

NewsLI is an independent, AI-driven news-analysis service — not a primary news outlet. We aggregate reporting from Israel's leading outlets across the political spectrum, together with world coverage, translate it into English, Russian and Hebrew, and analyse it for bias, propaganda and logical integrity. In Israeli terms we are a Kli Ezer — a helper tool. We do not replace the truth; we give you a logical filter for more conscious reading.

How the analysis works

1. Ingestion. Every few minutes we pull new items from public RSS feeds and a few Telegram-fronted publishers across the Israeli spectrum.
2. Translation & first pass. Each item is faithfully translated into three languages and screened for propaganda, clickbait, sophistry, emotional manipulation, tone, and gaps between a headline and its body.
3. Clustering. Coverage of the same event from different outlets is grouped, so you can compare how each source framed it.
4. Deep audit. On demand, an article is assessed across multiple pillars — emotional and identity pressure, language warping, cognitive traps, factual density, sourcing transparency, omission, temporal framing, bias alignment, conflict of interest and publication motive.
5. External check. Where possible, claims are cross-checked against independent outside sources.
6. Context. People, organisations and places are linked to a Wikidata-backed registry, and a media-literacy glossary explains the techniques we flag.

What the verdict means

Every analysed article gets a plain-language verdict, not a number to decode:

• Sound / low risk — little manipulation detected.
• Mixed / moderate — some pressure or framing worth noting.
• Manipulative / high risk — heavy emotional, rhetorical or omission pressure.

Under the hood the deep audit does score manipulation across the pillars above, but we translate that into a verdict rather than show a raw figure — a false-precise "73/100" invites more trust than any automated read deserves. It is a logical-integrity signal, not a verdict on whether a story is true: a factually correct report can still read as manipulative, and vice-versa.

Our use of AI, and what we don't touch

Summaries, translations and scores are produced or assisted by large language models. Two rules keep us honest:

• Card headlines are faithful translations of the original source — we do not rewrite a publisher's headline, because doing so would hide the very framing we are trying to measure.
• The scores are our own automated assessment, clearly attributed to NewsLI and separate from the source's words.

AI can misread context or hallucinate. Treat every analysis as a prompt to think, not a final authority.

Sources & independence

We read across the political spectrum on purpose, and we always link back to the original article so you can read it in full at the source. NewsLI acts as an intermediary analysis layer; original headlines, text and images belong to their respective publishers and are used for analysis, criticism and reporting.

How to verify anything you read here

Everything on NewsLI is informational. For anything that matters, follow the source link to the primary outlet and confirm it there. If you believe something is inaccurate or infringes your rights, contact us via our public channels and we operate a notice-and-takedown procedure.