Observed evidence
Public signals captured by ECIndex or another reproducible inspection. Include the source and observation date.
ECINDEX ACADEMY · OPEN PLAYBOOK v0.1Use AI to build
Use AI to build
a better store.
Practical, evidence-led guidance for building storefronts that people, search engines and AI systems can understand and trust.
What this playbook is.
A practical operating guide for owners, managers, developers and advisers. It is not a guarantee of ECIndex ranking, search visibility, AI citation or commercial results.
01 · EVIDENCE MODELKeep facts and
Keep facts and
suggestions separate.
A useful AI workflow tells the reader what was observed, what the merchant supplied, what the system inferred, and what still requires verification.
Merchant facts
Policies, operations or commercial details supplied by an authorized representative. Label their provenance.
AI recommendations
Proposed actions based on available evidence. Treat them as hypotheses until a qualified person reviews them.
Validation results
Checks completed after implementation, including what changed, who approved it and whether the intended outcome occurred.
02 · EIGHT PRACTICESWhere AI can
Where AI can
create practical value.
Start with a narrow operational problem, give the system reliable context, and require a measurable acceptance check.
Clean catalogue records
Standardize names, attributes, variants, units, identifiers and descriptions without changing factual product claims.
Product schema lesson →Search and AI readiness
Create page-specific metadata, structured data and useful category language grounded in visible storefront content.
Metadata guidance →Assist, then escalate
Draft answers from approved policies, identify uncertainty, and route exceptions or sensitive cases to a person.
Improve product presentation
Find missing attributes, inconsistent imagery and weak comparisons while preserving the merchant’s voice.
Summarize exceptions
Help teams classify delays, returns and stock discrepancies without exposing unnecessary customer data.
Turn audits into action
Translate measured speed and usability findings into prioritized tickets with owners and acceptance checks.
Run SpeedAudit ↗Draft responsibly
Create campaign variants from verified offers and brand rules. Never invent scarcity, testimonials or performance claims.
Record human approval
Keep prompts, sources, outputs, approvals and material changes available for review.
03 · SAFE WORKFLOWInspect before
Inspect before
you generate.
Define the outcome
State the business problem, affected page or process, and the measure that would show improvement.
Collect bounded evidence
Use the current storefront, ECIndex profile, policies and approved product data. Avoid private customer or payment information.
Ask for diagnosis before content
Require the AI to identify gaps, uncertainty and priorities before it drafts a solution.
Review with a responsible owner
Check factual accuracy, legal or policy implications, platform constraints and brand fit.
Publish narrowly and validate
Release a controlled change, test the live result, and preserve a rollback path.
Measure and document
Compare the new observation with the baseline and record what worked, failed or remains uncertain.
04 · STARTING PROMPT
Ask for an evidence-led store review.
This prompt is deliberately structured to prevent an AI assistant from presenting assumptions as verified findings.
Review this storefront using the ECIndex evidence model. Store URL: [URL] Primary goal: [GOAL] Approved context: [FACTS OR LINKS] Separate your response into: 1. Observed public evidence 2. Merchant-supplied facts 3. Uncertainty and missing evidence 4. Prioritized recommendations 5. Human review required 6. Validation checks after publishing Do not invent reviews, ratings, prices, availability, certifications, policies, performance data or customer claims.
05 · NON-NEGOTIABLESProtect trust
Protect trust
while moving faster.
No invented ratings, testimonials, certificates, prices, stock, delivery promises or product properties.
Do not place customer, order, payment or credential information into general-purpose AI tools.
Use clear internal records and public disclosure where AI materially shapes content or decisions.
A named person owns approval, publishing, monitoring and rollback.
Validate recommendations against current provider, advertising and search requirements.
Improvements may strengthen observable readiness; they do not purchase or guarantee an ECIndex position.
06 · PORTABLE AI SKILLApply the same standard
Apply the same standard
inside an AI workspace.
The ECIndex Store Improvement Skill packages the evidence model, safety boundaries and output structure for compatible AI assistants. It does not connect to private store systems or authorize changes.