Read the scope as well as the label
A testing label describes the work supporting a particular article. Look for the task, product/version, date, evidence and limitations. The owner gives final approval of material testing claims. A label does not extend one task into a verdict about every capability of a product.
Research / documentation based
This work relies on checked sources and scoped analysis. It can explain documented behavior or a procedure without claiming that SunloopMedia completed a direct product test. Source-reading and synthetic teaching exercises are described as such. Vendor documentation supports what the vendor documents, not an unperformed independent test.
Not tested
No supported direct-testing claim is made. A feature may be documented or a question may remain unresolved; neither becomes an observed result because it appears in a Review.
Hands-on and partially tested
Hands-on testing requires genuine recorded work and evidence for the stated task. Partially tested means direct work supports only part of an evaluation. Both need clear boundaries, including relevant conditions and what was not tested.
If an AI agent operated material work, that role is identified. It does not imply a human usability panel, a second independent reviewer or testing on devices that were not used. We do not fabricate prompts, results, benchmarks, sources or experience.
Retested and outdated / review required
Retested means recorded subsequent work has been connected to the earlier test. A changed screenshot, edited paragraph or new date alone is not a retest. Outdated / review required means information needs checking; the label must not imply that fresh verification has already happened.
What an observation supports
A successful task supports a conclusion about that task in the recorded conditions. It does not establish universal performance, privacy, security, compatibility or recovery. A copied folder reopening, for example, is different from a cross-application migration or a recovery from corrupted data.
Unknown pricing or availability should remain unknown rather than being guessed. Numeric review ratings are not currently used. Comparisons must follow the recorded evidence without inventing a winner, endorsement or stronger test claim.
Evidence and presentation
Sources, evidence state, methodology, disclosures and material updates should correspond to the article’s editorial record. Screenshots document the task they show; their captions should distinguish historical evidence from later illustrative captures. Conceptual diagrams and synthetic examples are labeled so they are not mistaken for a product traffic trace or measured result.
AI can assist analysis and production, but its output is not automatically evidence. Commercial benefits cannot determine a conclusion. See our Editorial Policy for responsibility and independence, and our Corrections Policy for reporting a material issue.