# Adversarial review

Band: emerging — Real practice, thin evidence. Mostly vendor-stated or reported rather than measured.
Group: Verification and authority
Implemented by: https://clickai.dev/skills/advisory-board

Running several reviewers with deliberately different lenses, or several investigators told to disprove each other, rather than one reviewer working sequentially. The rationale is anchoring: once one theory is explored, later investigation bends toward it, and a single reviewer gravitates to one class of issue at a time. The mechanism is sound and cheap to try. The supporting evidence is vendor-stated rather than benchmarked, and nobody should pretend otherwise.

## See also

- [Echo risk](https://clickai.dev/legend/echo-risk)
- [Couldn't-verify](https://clickai.dev/legend/couldnt-verify)
- [LLM-as-judge](https://clickai.dev/legend/llm-as-judge)

---

Full legend: https://clickai.dev/legend.md