DIGITAL ASSET RISK

Crypto products can result in total loss. A rating is not proof of solvency, safety or local availability.

Read the safety standard

Evidence architecture

Sources are weighted by what they can actually prove.

A source can be authentic and still be inadequate for a particular claim. Our hierarchy binds each conclusion to evidence of the right type.

Version 1.0Effective 18 August 2026English

Level A — Primary records

Regulatory registers, court records, company filings and completed independent audit materials are the starting point for legal identity, license scope, enforcement and certain financial claims.

Level B — Operational documents

Current user agreements, custody terms, fee schedules, status pages, security reports and provider documentation establish contractual rules and public product behavior.

Level C — First-hand testing

Standardized, timestamped tests can establish observed costs, product flows, support paths, accessibility and execution at a stated time and place. They do not prove future performance.

Level D — Company claims

Marketing pages, press releases and official blogs establish that a company made a claim. Unless independently verified, our language remains “the company states” or “the website claims.”

Level U — Unverified

Claims without a traceable source, identifiable entity or appropriate evidence do not influence a score. When the gap concerns identity, custody or legal status, it may stop the score.

Community evidence

Visitor reviews are experience reports. Moderation approval does not establish their accuracy, and they cannot verify legal, financial or security claims.

Retrieval and change control

Each source carries a retrieval date. Fees, terms, entity scope, regulatory status, product versions and incident responses are re-checked against the linked primary record when a file changes; an old capture never silently becomes a current fact.

Scenario record

First-hand claims require a declared market, account class, product version, device, asset, amount, timestamp, steps and retained output. Until those fields and results exist, the scenario is not counted as evidence even when provider documentation describes the expected flow.

Citation and machine access

Core content is rendered as semantic server-side HTML. Canonical URLs, structured data, the research index, RSS and sitemap mirror the visible record. llms.txt is only a curated navigation aid; it is not special content for Google, a ranking claim or a substitute for unique people-first analysis.

Questions or corrections

Challenge the record.

Send the exact page, disputed claim and best available source. Material corrections will be logged visibly.

Contact editorial research