StoreSteadyStoreSteady
Public methodology

StoreSteady AI Commerce QA Methodology

AI Commerce QA is evidence-backed testing of how AI systems, shopping channels, and crawlers interpret, cite, recommend, or misrepresent a merchant's public commerce surface.

Browse the finding taxonomy →

What this methodology covers

StoreSteady findings are generated from structured evidence, checked against a public methodology, constrained against unsupported claims, and tied to a validation path. The public standard explains the method without publishing private prompts, eval suites, thresholds, scoring weights, or repair logic.

Evidence before recommendation

A recommendation must point back to a public source, structured field, channel observation, or merchant-connected source.

Confidence before certainty

A finding should name how strong the evidence is before it asks the merchant to act.

Validation before closure

A fix is not complete until the source evidence can be retested.

No ranking, revenue, or legal claims without scoped evidence

StoreSteady does not promise rankings, revenue lift, legal compliance, or automatic crawler behavior.

Finding lifecycle

The public lifecycle is intentionally simple enough for merchants and crawlers to inspect.

Observation

StoreSteady captures a public page, structured field, shopping-channel response, AI answer, or supported probe.

Claim

The system turns the observation into a bounded statement that can be linked to evidence.

Conflict

The system checks whether sources disagree or whether an AI or channel response misstates the merchant surface.

Severity

The finding is categorized by merchant impact without publishing internal score weights.

Fix

The merchant gets the source surface to change first, not generic AI advice.

Validation

StoreSteady names how the same evidence should be retested after the fix.

The standard, in writing

Five short documents that make up the published methodology. Each links to a human-readable page; the machine-readable JSON and Markdown versions are in the right rail.

Rubric

The Rubric

How we decide whether a finding is solid enough to ship to you: evidence cited, scope clear, action specific, retest defined. Also available as Markdown.

Reference

What we look for

The families of issues we report on — product truth, policy claims, offer interpretability, AI answer gaps, schema, crawler access — with confidence labels. Also available as JSON.

Standards

How we keep the AI honest

The rules our own LLM follows when it writes findings or copy suggestions. Evidence first; no "studies show…" without a study. Also available as Markdown.

Reference

What counts as evidence

The sources we will cite (pages, markup, channel responses, AI answers, optional connected data) and the retest paths that close a finding.

Examples

Examples, side by side

What a generic "improve your AI visibility" tip looks like before and after we put it through this method.

What StoreSteady refuses to infer

No guaranteed rankings or AI visibility

StoreSteady can document evidence and observations, but it does not guarantee how a search engine, LLM, crawler, or shopping surface will rank or cite a store.

No revenue or conversion claims without evidence

A finding can explain likely commerce risk, but it cannot claim lift or loss unless a scoped measurement supports it.

No legal determinations from public scans

StoreSteady labels observed mismatches and missing evidence. It does not decide legal compliance unless a separate scoped review exists.