Intro
AI visibility work loses value when nothing is comparable.
Different agencies use different prompts. Different consultants frame competitors differently. Different reports apply different fact-check rules. The output looks like work, but a brand cannot compare two audits — let alone two retests of the same audit.
The Capston QA standards exist to solve that.
This page describes the rules that apply to every Capston Core engagement, every certified partner deliverable, and every internal audit. Without these rules, the score, the evidence layer, and the methodology do not produce comparable results.
The five QA pillars
1. Prompt rules. A prompt set is co-designed at baseline and locked. Prompts cannot be added, removed, or reworded between baseline and verification. New prompts open a new cycle.
2. Competitor rules. The competitor set is defined upfront and held constant. Adding a competitor mid-cycle invalidates the comparison and requires a documented note in the evidence layer.
3. Evidence rules. Every claim in a report ties back to a specific capture. No claim, no capture, no inclusion. The evidence layer is the source of truth — not slides or summaries.
4. Claim rules. Factual claims about a brand or its visibility are flagged as confirmed, partial, or contested. No invented numbers. No projected lift. No revenue figures without an underlying source.
5. Report rules. Reports follow a fixed structure: state, gap, action, expected lift, success criterion. Format is the same across clients and partners. Custom dashboards exist; custom methodology does not.
What QA catches
Five categories of failure that QA actively prevents.
- Prompt drift — the prompt set silently changing between captures
- Cherry-picking — selecting captures that flatter the narrative
- Source inflation — counting low-trust sources as if they were trade press
- Fact softening — describing wrong facts as “partial” to avoid hard conversations
- Out-of-scope creep — bundling content production into the methodology
When a QA review fails, the work goes back to the team — Capston Core or partner — until it passes.
Why QA is non-negotiable for partners
A partner can execute. A partner can localise. A partner can sell.
But the method has to stay Capston-grade. If a certified partner runs an audit that does not meet QA, the audit cannot carry the Capston Core attribution. The partner can deliver it as their own work, but not as a Capston Core score.
This is what protects clients from inconsistent quality across markets, and what protects the broader Capston Core brand from drift.
How this fits into Capston Core
QA standards govern the Capston Core methodology, the AI visibility scoring system, and the AI answer evidence layer. They are also the curriculum core of the certified partner program.
→ Back to Capston Core
FAQ
Who runs QA reviews?
Capston Core. Reviews are independent from the team that produced the work — internal or partner.
Can a client override QA?
A client can ask for an additional view or a custom export. A client cannot relax the methodology rules — that is what protects the comparison.
How often does QA evolve?
Standards are reviewed quarterly. Changes are versioned, documented, and applied to new engagements; existing engagements stay on the version they signed off on.
Do partners have access to the QA criteria?
Yes. Certified partners receive the full criteria as part of certification. Transparency is the point.
Final CTA block
Run AI visibility work that holds up.
Talk to QA
Read about certification