Publication standards

Editorial standards

Read Companion Curator’s rules for sourcing AI companion claims, handling public reports, correcting changes, using automation, and excluding harmful tools.

Editorial decisions stay separate from commissions

Products without affiliate programs remain eligible for coverage when they matter to a buyer’s decision. A merchant cannot buy a verdict, score, category win, favorable comparison position, or omission of a material caveat. Commercial eligibility and editorial evidence are evaluated separately, and compensation never changes the Finder’s matching rules or alphabetical ordering.

Every factual claim needs a source state

Every public factual field must cite an eligible source and verification date or be labeled unknown. Publisher claims, app-store records, community reports, editorial observations, spot checks, and longitudinal protocol results remain distinct. A source can document that a feature or policy is described without proving implementation quality, user outcomes, successful account actions, legal compliance, safety, or privacy.

Conflicting current sources remain visible as conflicts. Missing evidence remains unresolved. Prices are shown in their documented currency, billing channel, and commitment shape without inventing savings, checkout totals, entitlement parity, taxes, or a best-value winner.

Publication is a series of separate gates

A local draft, a tested page, an editorially approved page, an indexable route, a deployed artifact, and a monetized destination are different states. Passing automated checks does not grant human editorial approval. Publishing a source-checked page does not activate an affiliate link, prove merchant acceptance, or turn a documented claim into first-hand testing.

Public experience is context, not proof

Community and creator reports are selected for decision relevance and disagreement, not to manufacture consensus. Companion Curator discloses sample scope, publisher-family concentration, platform limits, anonymity, commercial signals, and other material concerns. Reports are paraphrased rather than republished, and a convenience sample is never presented as representative research.

Corrections preserve the evidence trail

Material product or policy changes create a new dated record rather than silently replacing history. A current page may be refreshed, narrowed, returned to draft, or removed from search eligibility when its supporting evidence expires or conflicts. The visible assessment date describes the evidence review boundary; it is not a promise that every remote source remains unchanged.

AI-assisted work remains accountable

Automation may help organize research, identify gaps, draft code, and run checks. It cannot manufacture tests, screenshots, receipts, user experiences, quotations, consensus, credentials, legal opinions, merchant acceptance, or author identities. Hands-on claims require real artifacts and disclosed scope. High-risk safety, privacy, and publication decisions remain separate accountable review lanes.

Hard exclusions protect consent and reader trust

Companion Curator does not promote nudifiers, nonconsensual deepfake tools, real-person sexualization, illegal content, age-ambiguous merchandising, deceptive affiliate routing, or merchants that fail required safeguards. These exclusions are explained because the boundary matters: consensual fictional adult-capable companions can be covered without directing readers toward tools built around a real person’s image or an absent person’s consent.

Readers can inspect the method

See the research methodology for evidence classes, freshness, and test requirements; the affiliate disclosure for the commercial boundary; and privacy and safety for how Companion Curator handles browser-local tool inputs and site measurement.