← Methodology

Methodology · Story Quality

Story Quality & Publishing Gates

Not every article that enters Updat3 becomes a story. Here is how we decide what has enough verified content to show in the feed.

Text quality tiers

Every article gets a content score from 0–100 based on how much usable text we can extract. Articles score higher when we have access to the full body text, not just a headline or excerpt.

Headline only

< 200 chars of text · Content score 0–18

Not used for story summaries

Excerpt

200–2,500 chars of text · Content score 18–52

Used with caution; summary may be thin

Full text

> 2,500 chars of text · Content score 52–100

Full enrichment, richer context

Publishing gates

A story only appears in the feed when it passes all of the following checks:

TitleAt least 8 characters

Filters out blank or corrupt entries

Summary (bluf)At least 120 characters and not identical to the title

Ensures a real summary exists, not just a headline copy

Story articleAt least 900 characters

Ensures the Updat3 article is substantive, not a placeholder

Historical contextAt least 700 characters

Ensures background context was generated, not skipped

Key factsAt least 2 distinct facts

Prevents single-fact stub stories

Named entitiesAt least 4 identified people, places, or organisations

Filters stories too vague to be useful

Feed ranking

Stories that pass the publishing gates are ranked in the feed by a decay-weighted score:

score = decayFactor × 100 + log₂(sources) + importance × 0.5

Where decayFactor halves every 12 hours, sources is the number of linked articles, and importance is an optional editorial weight (0–10).

Stories with only one source older than 72 hours are hidden — they haven't attracted additional coverage and are likely too niche or unverified to surface reliably.

What Updat3 does not do

×We do not fact-check individual claims — we aggregate and summarise what sources report
×We do not editorially suppress stories based on their political lean — the publishing gate is content quality only
×We do not assign confidence scores visible to users — the quality check is binary (published or not)
×We do not manually curate the feed — ranking is algorithmic
← Bias ScoringAll methodology