Evidence standards
Every finding declares how it was established. An inference is never presented as a direct observation. These are the only six states a finding can carry.
Written and reviewed by Vitor Barbosa · Editor
Responsible for the editorial standard of this site and for the methodology published on it. Reviews every page that makes a technical claim before it is published.
- Published
| State | What it means | What it does not mean |
|---|---|---|
| Observed | Collected directly from the URL or the resource — a response header, a status code, a tag in the markup. | It does not mean a search engine sees the same thing. It means we saw it, at the recorded time, from the recorded location. |
| Verified by source | Returned by an external source you authorised. Nothing is connected today — the billing FAQ and /methodology both say we connect to no external source, and this row would contradict them if it did not say so here. | It does not mean the data is complete. Connected sources sample and aggregate, and we report their limits with the number. |
| Derived | Computed deterministically from evidence we already hold. Given the same inputs it produces the same output, every time. | It does not mean it was measured directly. It means the calculation is reproducible and its inputs are listed. |
| Inferred | A probabilistic judgement, usually about content or intent, with the reasoning attached. | It does not mean it is fact. It carries a confidence level and it is never styled to look like an observation. |
| Self-reported | Something you told us — your priority markets, your target queries, your stack. | It does not mean we confirmed it. We use it as context, and we say so. |
| Unavailable | We could not measure it. A firewall challenge, a timeout, a missing permission, insufficient sample. | It does not mean a pass, and it never counts as one. A blocked crawl is our limitation, not a defect in your site. |
Why unavailable is not a pass
A report that quietly converts what it could not measure into a clean result is worse than useless, because it produces confidence that is not earned. Every category shows how many checks applied, how many ran, and how many could not be evaluated. If the denominator is small, you can see that it is small.