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.
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.
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.
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.
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, 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.