Limitations statement

What an Alt Matrix does not prove.

This exact text appears on every report, video end-card and PDF. No plan or setting removes it.

  1. Personas are simulations. They enforce parameters such as patience, reading level, dexterity and input mode. They are not people and do not represent any group.
  2. Findings are observed behaviour. They describe what specific agents did on a specific date with specific model versions. Agents are non-deterministic; results can vary between runs.
  3. No compliance statement. Nothing in this report means a property is accessible, compliant (WCAG, ADA, EAA or other), secure or “agent-ready”.
  4. Consumer-agent classes are approximations. Where a column is labelled as a consumer-agent class, it is a behavioural approximation using a vendor model under alt.qa’s identity, not the named product.
  5. Payments stop. Outside a sandbox every run stops before payment, so post-payment steps are not tested.
  6. Precision is stated. Confidence levels (Confirmed / Likely / Reported) and per-persona precision (confirmed ÷ flagged, with n and 95% CI) are shown. Human confirmation status is shown per finding.
  7. AI-generated content. Narration, persona monologues and videos are generated by AI.

Related: methodology · published precision · claims policy