Methodology

Ehtebar aggregates news from 14 Afghan and international outlets in English, Dari, and Pashto. Each story is translated, grouped with related coverage from other outlets, and analyzed for reliability and editorial tone. When multiple sources cover the same event, we cross-reference their claims to show where they agree and where they diverge. This page explains how that process works and where the limits are.

The pipeline regularly scrapes new articles from all tracked outlets and runs them through three AI stages:

1. Translation and tone assessment. Each article is translated into English (regardless of its original language) and individually assessed for editorial tone. This step normalizes content across languages so that Dari, Pashto, and English sources can be compared on equal footing.
2. Evidence analysis. Articles covering the same event are grouped together using semantic similarity. The AI then compares them across sources: identifying key claims, noting which outlets support or dispute each claim, and flagging where reports agree or contradict each other. Reliability, importance, and category are assigned at this stage.
3. Article generation. A final article is written in neutral English, drawing from all sources in the group. A Dari headline is generated alongside the English one. The output is a balanced synthesis — not a copy of any single source.

When a story was previously published and new outlets pick it up in a later cycle, the existing article is updated: new sources are added, and the reliability label is reassessed based on the broader coverage.

Each article includes a short reasoning line that explains why a particular reliability label was assigned. When a story's reliability changes — because new outlets pick it up or sources are added — the update is tracked and the label is reassessed automatically.

When multiple outlets report on the same event, we don't just count them — we compare what they're saying. Each multi-source article includes an evidence breakdown:

• Key claims — the central factual assertions in the story, with which outlets support or dispute each one.
• Agreements — points where sources converge and tell the same story.
• Disagreements — points where sources contradict each other, report different figures, or attribute events differently.

This analysis is visible on every article page and is one of the primary inputs to the reliability label. A story where all sources agree on key facts is treated differently from one where they diverge.

For certain stories — particularly those with limited sourcing or unusual claims — the pipeline runs an external web search to check whether the story has been reported outside of the outlets we track. When corroborating results are found, they are recorded alongside the article as additional context.

When the tone is framed or sensationalized, specific loaded phrases are cited from the source texts so you can see what triggered the label.

What tone does not measure: Political terminology choices (such as “Taliban” versus “Islamic Emirate”) are not flagged. These are political labels, not rhetorical techniques. Direct quotes from officials using strong language are also not flagged if the article itself reports them neutrally. Editorial selection (which facts a source chose to cover) is inherent to journalism and is not penalized.

Each article is assigned to one of six categories: Politics, Security, Economy, Society, Culture, or International. It also receives an importance score from 1 to 10 reflecting newsworthiness — how significant the event is, how many people it affects, and how widely it's being covered. These are used to determine article prominence on the homepage.

We don't produce original reporting. Every story links directly to the primary source article, and when multiple outlets cover the same event, we list them all. You can always click through to read the original in full.

We currently track 14 outlets across three languages:

• English — Pajhwok, Khaama Press, AP News, Al Jazeera
• Dari / Persian — ToloNews, Ariana News, Hasht-e Subh, Amu TV, Afghanistan International, BBC Persian, Bakhtar News
• Pashto — Hurriyat, Omid Radio

Sources include a mix of independent Afghan outlets, state-affiliated media, and international news organizations. We aim to represent a range of editorial perspectives so that cross-source comparisons are meaningful.

Reliability labels are an indicator, not a guarantee. They reflect source-level patterns and cross-referencing — they are not a verdict on whether any specific claim is true or false.

We cover a curated set of outlets, not the entire Afghan media landscape. Stories that only appear in sources we don't track won't show up here. The entire pipeline is AI-driven, which means translations, summaries, and labels may occasionally contain errors. AI-generated labels are subject to human review — our team spot-checks assessments and can override any rating in the CMS. If you spot something that looks off, please let us know.