Newstools for publishers
Verification and editorial quality
Claims, evidence, and uncertainty in plain view: the final call stays with the reviewer.
Let's talk about your project
Evidence, not verdicts
Every claim arrives with its sources and its uncertainty; the reviewer decides.
Quality gate before the CMS
Whatever fails the agreed criteria is held, with the reason recorded.
Recurring failures measured
Every result is logged to fix the process, not just the piece.
From the piece to a decision with evidence
A person decides at this step
- Piece in productionA draft, written in-house or with AI assistance, enters the workflow before CMS export.
- ExtractionAgents extract the claims and entities from the piece.
- EvidenceThey search semantically indexed knowledge, compare against sources, and assess citations; they mark uncertainty.
- ThresholdsWhat clears the agreed criteria moves on; what is doubtful is held with the reason.
- Reviewer's decisionReviews claims, evidence, and uncertainty and decides: pass, correct, or send back.
- RecordEvery result is saved; recurring failures feed the improvement of the process.
No automated verdict on sensitive decisions.
What it solves
When part of your content is produced with AI assistance, the risk is not only error: it is not knowing where each claim came from. Manual verification does not scale to the pace of publication, and doubtful cases slip through.
We integrate verifiable checks into the workflow: claim extraction, source comparison, citation assessment, and traceability of every step. Thresholds and doubtful cases are surfaced for editorial review instead of being resolved in silence.
How we work
We start by agreeing with the newsroom what gets checked and against which thresholds: claims, sources, citations, style, and your own editorial policies. Those criteria are versioned and applied the same way to every piece.
In the workflow, agents extract claims and entities, search semantically indexed knowledge, and return evidence for review, with uncertainty marked. What clears the thresholds moves on; what does not, or what remains in doubt, reaches the reviewer with sources and evidence in plain view.
We can add quality gates before CMS export, log every result, and measure recurring failures to fix the process, not just the piece.
What we deliver
- Verification workflows that connect ingestion, evidence retrieval, and analysis, integrated into your production.
- Quality gates before the CMS, with the agreed criteria and the reason for every hold recorded.
- A review panel where sources, uncertainty, and decisions stay visible to the reviewer and to whoever looks at it later.
- Logged results and metrics on recurring failures to improve the process.
No automated verdict replaces an editor's check on sensitive decisions. AI-generated text signals, when used, are indicative only.
Who it is for
Newsrooms that bring AI into production and need to show that every piece went through checks; fact-checking teams that want to spend their time on doubtful cases rather than trivial ones; and publishing groups that want one quality standard across several titles.
What we have already built
Factlayer Verify brings together the assisted verification components: it extracts claims and entities, searches semantically indexed knowledge, and returns evidence for review. Its workflows connect ingestion, retrieval, and analysis, and keep sources, uncertainty, and decisions visible to the reviewer. It is the base on which we build each newsroom's checks.