Skip to main content
reviewscorer
レポート依頼
仕組み

レビュー評価方法。

reviewscorer evaluates online projects by checking public evidence, entity signals, source citations, and the quality of visible claims. Each verdict is sourced, dated, and revisable.

5 Principles4 Review stages0–100 Trust score

Anatomy of a review.

Every reviewscorer record carries the same parts. Here is one, broken open.

Sample · EdmundsReviewed 2 days ago
EdmundsE-commerce · edmunds.com
82/100Low risk
Verdict

Low-risk public automotive research site with strong visible content, support, and domain signals. Users should still verify seller-specific details before transacting.

4 sources citedLast revised
  1. The 0–100 trust score.

    One number summarising the evidence weight. Petrol when low-risk, copper when mixed, red when high concern. The score is the verdict in shorthand.

  2. Review signal.

    The page shows the current reader-facing signal: low risk, needs review, or high concern. Internal workflow states stay inside the editorial workspace.

  3. Editor's verdict.

    Plain-language judgment that reads aloud cleanly. No marketing language, no hedge phrases, no fake aggregate authority. What we found and what it means.

  4. Sources and date.

    Every claim cites a public source a reader can inspect. Every record carries a date stamp and stays revisable as new evidence appears.

Five principles.

The standard we apply to every review we publish.

  1. Public evidence onlyEvery claim cites a verifiable public source a reader can inspect. Where evidence is private or unverifiable, we do not publish.
  2. Editorial gatekeepingOnly records approved for reader use appear on the website. Internal investigations stay out of search, navigation, public APIs, and AI-readable discovery files.
  3. Honest uncertaintyWhere evidence is thin, we say so plainly on the page. We avoid unsupported trust claims, fake aggregate authority, and urgency-driven language.
  4. Revised on signalReviews are updated as new evidence appears, not on a schedule. A corrected record is preferred over a stale one.
  5. IndependenceReviews are initiated by our editorial team or by user request. We do not accept payment for favourable verdicts and maintain a clear editorial/commercial separation.
  6. Machine-readablePublic pages publish structured data that matches visible content. We optimise for crawlability and answer-ready summaries, not for ranking promises.

Trust scoring.

How a 0–100 trust score maps to risk and rating.

65–100Great
Low risk

Strong evidence of operational transparency, verified identity, and credible conduct. Two or fewer unresolved minor signals tolerated.

35–64Average
Med risk

Mixed signals. Some evidence is present but key pillars are unverified, contradictory, or rely on operator self-disclosure. Further investigation recommended.

0–34Poor
High risk

Multiple unresolved signals, anonymous operators, fabricated references, or patterns consistent with prior failed projects.

Editorial lifecycle.

How a review moves from intake to a reader-facing record.

Review workflow
IntakeResearchPublished recordNeeds more evidence
4.2 dmedian time

Intake

Editor opens a casefile and records the domain, category, source candidates, and initial questions.

Research

Evidence is gathered, contradictions are resolved, and source references are checked before any reader-facing claim is made.

Published record

The verdict is sourced, dated, revisable, and available for readers, search engines, and answer engines.

Needs more evidence

If evidence is thin or contradictory, the record remains in editorial review until it can be written honestly.

External standards we monitor.

Public documentation we cross-check during technical review.