Evidence and repeatability
Freeze inputs, task version, allowed tools, environment, expected output, criteria, exclusions, retries, cost, and raw observations.
Review status
Both published runs name Matthias Ramahi as benchmark owner and openly state that no second independent human reviewer exists.
Provider gate
- Benchmark and output rights documented
- Equal frozen task and rubric
- Provider, data, retry, infrastructure, and review budget fixed
- Reviewer and adjudication rule named
- Failures and excluded runs remain visible
Ownership conflict
SEO AI Agent, Crawl Foundry, and seo-mcp.de share the same operator. Common ownership is disclosed beside relevant links and can never act as independent corroboration.