Crosscheck: accurate, balanced, evidence-first.
Crosscheck exists to debunk unsubstantiated and viral social narratives by showing what the underlying evidence actually says. We aggregate the same event across many sources and languages, translate everything to English, and transparently surface where reporting agrees and where it contradicts itself. We do not take sides — evidence does.
Our editorial principles
- Same evidentiary standard for every side. If a claim is well-evidenced, it stands — regardless of which side it favors or who it criticizes.
- When any party admits fault, we report it straight. No spin, no softening, no attempt to explain it away. Facts, not interpretation.
- No blanket dismissals of institutions. Critics of UN bodies, NGOs, wire services, or state media get a hearing on the specific claim. We flag specific claims with specific evidence — not entire organizations with a sneer.
- Lower-reliability sources are flagged, not hidden. State media, anonymous Telegram channels, and partisan outlets across the spectrum are all visible — labeled with their reliability tier and reliability profile (ownership, funding, governance, correction record, known controversies).
- Narrative Flag fires on unsupported linkage, not on criticism. Well-evidenced criticism of any government or actor passes the check. Conspiracy tropes, scapegoating, and fabricated causal links — from the left or the right — get flagged with an evidence-based rebuttal.
Reliability tier system
Every source in the registry gets a tier. The tier sums up the source's editorial track record, not its viewpoint. A right-wing tabloid and a state-run agency can both be "partisan-sourced"; that's information, not a verdict.
International wire / public broadcaster with strong corrections record.
Established editorial desk with normal sourcing standards.
Editorial line aligned with a state or political actor — sourcing is real but framing leans.
Primarily commentary / opinion — treat as a viewpoint, not a primary source.
Social media / OSINT — claims need corroboration before being treated as fact.
Persistent disclaimer
Everything you see on Crosscheck is an aggregated claim from one or more sources, not Crosscheck's assertion of truth. We surface what each source says and how independently those claims are corroborated. Compare and judge for yourself.
How the pipeline works
- RSS feeds from ~55 outlets are fetched every 15 minutes.
- Non-English articles are translated and structured claims are extracted (who/what/where/when/casualties/attribution).
- An AI clustering step groups articles that describe the same underlying event.
- Each event gets a confidence label based on how many independent source categories corroborate it.
- Narrative Check and Narrative Flag use the same Lovable AI gateway with the editorial principles above hard-coded into every system prompt.