Fact-Checking Standards

Every published article is checked at four distinct points in the editorial process (writer, assigning editor, fact-check, final review — per our editorial process). The fact-check stage specifically verifies:

• Every direct quote, against transcript or recording.
• Every numeric claim, re-pulled from the primary source.
• Every on-chain assertion (TVL, transaction counts, token flows, contract addresses), confirmed against the chain.
• Every cited document (filings, audits, statements), opened and read by someone other than the original writer.
• Names, titles, dates, and identifying details of every person.
• Implicit claims — assertions in framing or headline that depend on facts not stated explicitly.

We trust sources in roughly this order:

1. On-chain data — the chain itself. Block explorers (Etherscan, Solscan), node queries, indexer APIs. Highest authority for transactions, balances, and contract state.
2. Primary documents — SEC filings, court records, audit reports, official statements, signed-and-sent letters. Open the PDF; don’t paraphrase the press release.
3. Direct sources — named, on-the-record interviews with people who have first-hand knowledge.
4. Verified secondary reporting — reporting from outlets we trust (Reuters, Bloomberg, FT, Coindesk, The Block, Decrypt) where the original source is identifiable.
5. Background sources — per our sourcing policy, used when essential and corroborated.

Press releases, exchange announcements, and project blog posts are not authoritative on their own. Treat them as claims to verify, not facts to repeat.

Some claims trigger heightened scrutiny by default:

• TVL, market cap, daily volume — multiple sources disagree by design. We disclose source and timestamp for every published number.
• “According to a person familiar”— the editor must know who. The phrase is not a license to run a single source claim.
• Hack & exploit numbers — loss estimates evolve. We label early estimates as preliminary, update with later confirmed figures, and link both in the article.
• Token launches, airdrops, vesting cliffs — verified against on-chain contract state, not project announcements alone.
• Regulatory action — verified via the primary docket (SEC EDGAR, court records), not via reposts or screenshots.

AI-assisted drafts (per our AI policy) are fact-checked by a human with the same rigor as a human first draft — arguably more, because models hallucinate. AI is never a substitute for the fact-check stage.

Articles that pass the full four-stage process display a fact-checked flag in the header. Breaking-news pieces that publish before the full check is complete are explicitly labelled as such and updated with the flag once verification is finished. Pieces that fail re-check after publication are corrected immediately and a correction note is appended.

When an article is materially wrong — a key fact misread, a quote misattributed, an on-chain claim that does not survive re-check — we:

1. Append a timestamped Correction block to the article footer (see /corrections).
2. If the headline is wrong, the headline is rewritten in place.
3. If the article cannot stand, it is fully retracted with a replacement note explaining what was wrong and what we now believe the truth to be. The original URL serves the retraction.
4. Retracted articles do not silently disappear. The URL remains live, permanent, and indexable.

We do not delete published work to hide errors. We correct it.

Spot a factual error? Email [email protected]. Cite the article URL, the specific claim, and your source for the correct fact. Credible reports get acknowledged within 24 hours and either corrected or addressed in a published response.