What we can test, and what we cannot claim.

First-hand evidence

A flow test needs a boundary

For our own service, an editor can open a defined route, record the browser and viewport, note whether sign-in appears, and inspect the words and amounts displayed at checkout. The observation belongs to that date and state. A successful path does not prove every device or account receives the same offer.

Screenshots should support a specific statement and avoid exposing participant data. When an interface changes, we update or remove the claim. We do not infer moderation response time, match supply or payment reliability from a marketing page.

• Record device, browser, date and starting URL.
• Separate visible interface text from interpretation.
• Retest unstable price, permission and access claims.
• Keep construction pages out of search until reviewed.

Third-party evidence

The publisher speaks for its own product

When describing the unrelated SoMatch social app, we prefer its official website and verified store listings. We identify the review date and avoid importing its reviews, screenshots or branding as our own. If the official sources disagree, we state the difference rather than picking the more convenient claim.

A comparison should help the reader choose. It can say the app appears better suited to a community or party goal while our browser route serves a focused one-to-one goal. Commercial interest must be visible because our comparison links back to our product.

Observed

A dated screen or route an editor actually checked.

Publisher claim

A statement on an official page, attributed to that publisher.

Inference

A reasoned interpretation, labelled and kept narrower than the evidence.

Corrections

A clear mistake should be easy to report

Meaningful corrections should state what changed and update the review date. Small grammar fixes do not need a public history, but a wrong price, identity statement, permission description or safety instruction does. Support contact must reach a monitored receiver before search publication.

AI can assist with outlines and language. It cannot be the source for a price, product test or user outcome. A human remains responsible for checking duplication, accuracy, tone and whether a page deserves to exist.