How It Works

1. Source Selection & Monitoring

ClearFeed monitors RSS feeds from dozens of established news organizations across categories including politics, business, technology, science, health, entertainment, and sports. Sources are selected based on editorial standards, factual reporting track record, and breadth of coverage.

Our automated pipeline checks these feeds every 30 minutes for new articles. When a new article is detected, it enters our processing queue for summarization.

2. Article Extraction

For each article, our system retrieves the full text from the original source. We extract the article body, author, publication date, and any associated images. The original source URL is always preserved so readers can access the full article directly.

3. AI Summarization

Each article is processed by a large language model with specific instructions to produce a factual, neutral summary. The AI is directed to:

• State facts directly — report what happened, who was involved, and what the consequences are, without editorializing or adding opinion.
• Preserve nuance — include relevant context, caveats, and competing perspectives mentioned in the original article.
• Be concise — deliver the essential information in a few sentences that can be read in under a minute.
• Avoid sensationalism — use measured, factual language regardless of the tone of the original article.

Every summary includes a direct link to the original article. ClearFeed is designed to complement original journalism, not replace it.

4. Headline Accuracy Score

One of ClearFeed's distinctive features is the headline accuracy score. For each article, the AI compares the original headline against the actual content of the article and assigns a score from 1 to 10 based on how faithfully the headline represents the reporting.

The score considers factors such as:

• Does the headline accurately reflect the main finding or event?
• Does it omit important qualifiers or context?
• Does it use emotionally charged language not supported by the article?
• Does it exaggerate the certainty or scope of what was reported?

A score of 8–10 means the headline is a fair, accurate representation. A score of 4–7 indicates some mismatch or mild sensationalism. Below 4 suggests the headline is significantly misleading relative to the actual reporting.

This score is an AI assessment and should be treated as an informational tool, not a definitive judgment.

5. Framing Notes

When the AI detects notable framing in an article — such as an emphasis on one perspective, missing context, or language that could influence reader perception — it generates a brief framing note. These notes help readers understand how the story is being presented and what alternative angles might exist.

Framing notes are not accusations of bias. They are observations about editorial choices that readers may want to consider when forming their own understanding of a story.

6. Trending Detection

When multiple independent sources publish articles about the same story within a short time window, ClearFeed identifies this as a trending topic. Trending stories are scored based on the number of sources covering the event and the recency of coverage.

This cross-source clustering helps surface the stories that matter most right now, based on actual editorial attention rather than social media engagement or algorithmic amplification.

7. Continuous Updates

The entire pipeline runs continuously throughout the day. New articles are typically processed and available on ClearFeed within minutes of publication. The feed is sorted by relevance (trending score) by default, with an option to sort by most recent.

Transparency & Limitations

ClearFeed is transparent about what it is and what it is not:

• We are not a news organization. We do not have reporters, editors, or original sources. All content originates from third-party publishers.
• AI-generated summaries may occasionally contain errors or misinterpretations. We always link to the original source so readers can verify.
• Headline accuracy scores and framing notes are AI assessments, not definitive fact-checks.
• Our source list is curated but not exhaustive. We continuously evaluate and expand our coverage.

Questions about our methodology? Contact us.