Evidence standard
How we test
AI can organize evidence; a person has to supply it. Here is what each label means and what a page must answer before it publishes.
Evidence labels
| Label | What it means |
|---|---|
| Owned and used by the owner | Product identity and daily use are confirmed by the desk owner, with a receipt where one exists. |
| Hands-on tested | A documented test was completed, possibly on a loan. Duration and conditions stated. |
| Researched alternative | Specs and suitability checked against current sources. Not personally used. |
| Pending verification | Model, claim, or variant is unresolved. Excluded from buying recommendations. |
The minimum review brief
Before a review publishes it answers: What problem did the owner have? Why did they choose this? How long have they used it? What has worked? What annoys them? What would they buy instead? Who should skip it? What must someone check before ordering? Would I pay for it again?
What the tests look like
- Displays: actual app layouts, text legibility, input switching, and connector constraints. Manufacturer specs are verified separately and linked.
- Desks: usable space measured, real load checked, and a repeatable wobble demonstration. The result is described; "rock solid" is not a finding.
- Calls: the same phrase recorded at the same distance and settings before and after a change, with light position and room conditions shown.
- Workflows: a defined task repeated and the observed result reported, without turning a short test into a universal productivity claim.
What is never claimed
Invented time savings. Ownership histories the owner does not have. Copied customer ratings. Medical promises about posture or pain. Gifted or sponsored gear is held to the same standard and labeled as such.
Every new first-person claim maps to an approved founder note, recording, or test record. No source means the page asks a question instead of inventing an answer.