Shadowfetch

Standards and accountability

Trust is a process readers can inspect

Shadowfetch exists to show meaningful differences in framing without losing the verified facts underneath them. These standards apply to every desk, every lane, and every format.

Verification before velocity

Reporters identify the source of every material factual claim and distinguish direct evidence, attributed reporting, analysis, and inference. High-impact claims require independent corroboration whenever practical. Anonymous sourcing is reserved for information the public needs and receives an editor review of the source’s access, motive, and risk.

A story that cannot meet the verification bar is narrowed, delayed, or held. Competitive pressure is not evidence.

Balance without false equivalence

Left, Center, and Right are labels for observable framing and source patterns, not scores for truthfulness or moral worth. We classify the specific coverage, not the reader or the person being covered. A label never substitutes for checking the underlying claim.

Balance means representing the strongest relevant evidence and good-faith arguments across a dispute. It does not require equal weight for a demonstrably false claim. Our public methodology explains the classification signals and limitations.

Common Ground has a higher floor

A Common Ground item must be independently supported across all three coverage lanes and written as a checkable statement. Shared repetition is not enough when every outlet traces back to the same unverified source. We record the agreeing outlets and remove or correct the item if its evidentiary basis changes.

News, analysis, opinion, and advertising stay distinct

Opinion is labeled at the headline and page level. Analysis may interpret evidence but must expose the reasoning and cannot present speculation as a reported fact. Sponsored or advertising material is labeled and visually separated. Advertisers, contributors, and commercial partners do not select stories, sources, lane labels, or corrections.

AI may assist; people remain accountable

Automation may help transcribe, translate, search documents, compare drafts, generate accessibility text, or surface possible gaps. It may not invent reporting, sources, quotations, images presented as documentary evidence, or factual confidence. A responsible editor reviews publishable work and remains accountable for every claim.

Corrections are visible and specific

We correct factual errors promptly and append a dated note saying what changed. Material errors are disclosed near the top of the story. Clarifications, updates, and editor’s notes use their own labels because they mean different things. We do not silently rewrite the record to make an error disappear.

Send a correction request to support@shadowfetch.com with the story URL, the disputed passage, and supporting evidence. Lane-label appeals follow the same review path. Read the full classification methodology or our privacy policy.