store-analyzer

Store Analyzer

The most comprehensive public-data Shopify store audit. Built on what 50+ real store reviewers actually check (Shopify Community research, March 2026) plus current Google, Bing, and Shopify guidance.

The skill reasons through eight questions in order:

1. Would a first-time visitor trust this store enough to buy? (Trust & Credibility)
2. Does the store convert browsers into buyers? (CRO & Conversion)
3. Is the store fast enough to keep visitors? (Page Speed)
4. Can search engines crawl and understand the store? (Technical SEO)
5. Do product pages rank and convert from search? (Product-Page SEO)
6. Can Google create rich ecommerce results from the pages? (Structured Data)
7. Can an AI agent answer common shopping questions from the content? (AEO)
8. Is the content strong enough to be cited, not just indexed? (GEO)

Eight Audit Modules

Module	What it measures	Reference
Trust & Credibility	Logo, favicon, about page, contact completeness, branded email, footer, payment icons, policy transparency, social links	trust-ux-checks.md §1-4
CRO & Conversion	Hero CTA, cart type, cross-sells, shipping bar, size charts, trust badges near ATC, express checkout, review widget, email capture, urgency elements	cro-checks.md §1-5
Page Speed	Core Web Vitals (LCP, CLS, FCP), mobile performance score, third-party script count, image loading strategy	trust-ux-checks.md §5
Technical SEO	Crawlability, robots, sitemap, canonicals, internal linking, crawl efficiency	seo-checks.md §3-4, 8-9
Product-Page SEO	Titles, meta descriptions, headings, content depth, image coverage	seo-checks.md §1-2, 5-7
Structured Data & Merchant Readiness	Product/Offer schema, merchant listing fields, review markup, Organization, identifiers	seo-checks.md §1 + catalog-checks.md
AEO Readiness	FAQ content, policy/shipping/returns clarity, product specs, Q&A format, answer blocks	aeo-checks.md
GEO Citation Readiness	AI bot access, unique content, brand identity, trust signals, content extractability	geo-checks.md

Conversion & trust (40% of audit weight) matter most — these are what real reviewers flag first. Then SEO fundamentals (25%), structured data (15%), and AI readiness (20%).

Bundled Resources

• references/cro-checks.md — Conversion checks: hero CTA, cart experience, product page elements, urgency, email capture
• references/trust-ux-checks.md — Trust & UX checks: branding, contact, footer, policies, page speed/Core Web Vitals
• references/seo-checks.md — SEO audit checks, heuristics, and Shopify-specific technical issues
• references/geo-checks.md — AI visibility checks: bot access, entity signals, citation-worthiness
• references/aeo-checks.md — Answer engine checks: snippets, FAQ schema, voice search, PAA
• references/catalog-checks.md — Product and collection data quality checks from /products.json
• references/troubleshooting.md — Edge cases: blocked endpoints, missing data, how to phrase partial findings
• assets/report-template.md — Exact output format. Follow it precisely.
• references/testing.md + evals/evals.json — Trigger tuning and regression checks
• Validate saved reports: python ${CLAUDE_SKILL_DIR}/scripts/check_report.py --input path/to/report.md --mode full|focused|review

Critical Rules

• Use only public Shopify endpoints and public page HTML. Never attempt authentication.
• Never make authenticated Shopify changes from this skill. store-analyzer is analysis-only.
• Report exact counts. Do not round, estimate, or use "approximately."
• Label major findings as Catalog-wide, Sampled, or Inference.
• Never generalize from sampled HTML to the entire site without saying so.
• If products.json is blocked, empty, or unusable, say so clearly and stop the catalog audit instead of guessing.
• Treat missing server-rendered JSON-LD carefully — apps can inject it client-side.
• Never claim Google penalties, guaranteed rankings, or traffic loss without direct evidence.

Research-Backed Principles (sources: Google Search Central, Bing Webmaster, Shopify AI docs)

These override any conflicting guidance in reference files:

1. AI citation readiness = indexability + snippet eligibility + content completeness. Google says "you don't need new machine-readable files or AI text files" for AI features. Do NOT recommend creating llms.txt.
2. FAQPage schema does NOT drive AI citations. Google deprecated FAQ rich results (Aug 2023) for non-gov/health sites. Score FAQ content quality, not schema presence.
3. Title/meta description character counts are not actionable. Google rewrites titles and truncates snippets dynamically. Only flag missing, mass-duplicated, or misleading titles.
4. Shopify default technicals are pass/fail. Robots.txt, sitemap, canonicals, SSL are auto-generated. Only flag if broken or customised harmfully.
5. Performance micro-optimisations are never findings. Missing fetchpriority, image dimensions, preload hints — informational only.
6. Organization schema is conditional. Not in Google's/Bing's core AI requirements. Only flag for nationally recognised brands with entity disambiguation needs.
7. HowTo schema is deprecated and product-gated. Google deprecated HowTo rich results (Sep 2023). Skip for clothing/shoes/accessories.
8. Comparison tables are context-gated. Only relevant for multi-brand stores. Single-brand DTC stores do not need "A vs B" content.
9. OG/Twitter tags, meta keywords, rel=prev/next are NOT AI citation factors. Do not weight for AI readiness.

Finding Quality Bar

Not every failed check is a finding. You are auditing real businesses. A finding earns inclusion ONLY if:

1. Provable revenue or traffic impact

Can you connect this to lost revenue, lost traffic, or lost competitive position with a specific mechanism? "Best practice" is not enough.

• YES: "All 884 collection pages have zero descriptions — no content for Google to rank for category searches"
• YES: "Product titles truncated at 55 chars in every SERP listing — reduced CTR"
• NO: "Missing llms.txt" — no store has one, no AI engine requires it
• NO: "No HowTo schema" — irrelevant unless the store sells instructional products
• NO: "Missing fetchpriority on hero image" — micro-optimization, mention in passing only

2. Proportional to store maturity

Assess the store from the data. A 500-product brand with reviews, proper Product schema, and a blog is doing many things right — the report should reflect that.

• Established store (500+ products, reviews, schema present): 4-6 findings max. Ask: "Would the Head of Growth act on this?" If no, cut it.
• Growing store (50-500 products): Mix of strategic issues and quick wins. 5-7 findings.
• New/small store (< 50 products): Foundational issues matter more. Up to 8-10 findings.

3. Not checkbox noise

These are NOT findings for established stores. Do not report them:

• Missing llms.txt (Google says no new AI text files needed — NEVER recommend creating one)
• Speakable schema (beta, news-only)
• HowTo schema on non-instructional products (deprecated by Google Sep 2023)
• FAQPage schema absence (deprecated for non-gov/health sites Aug 2023 — score content, not markup)
• No comparison tables on single-brand DTC stores
• Brand name "inconsistency" between domain and display name (neemans.com vs "Neeman's")
• Missing Organization/sameAs links (unless brand has national recognition)
• No Knowledge Panel (most DTC brands don't have one)
• Missing fetchpriority/image dimensions/preload hints (micro-optimisations, never a finding)
• Title/meta description character-count violations (Google rewrites freely)
• OG/Twitter tag completeness (social preview, not AI citation factor)
• Meta keywords tag (Google explicitly ignores)
• rel=prev/next pagination (Google no longer uses)
• Blog freshness thresholds (unless blog is a core traffic channel)
• Default Shopify robots.txt/sitemap/SSL being "present" (these are auto-generated, not achievements)

Merge related issues

• Meta descriptions + thin titles + H1 issues = ONE finding about SERP presentation
• No FAQ schema + no question headings + no voice readiness = ONE finding about answer-engine gap
• Missing Organization schema + no sameAs + no Knowledge Panel = ONE finding about entity identity

Output Rules

Report structure

Follow assets/report-template.md EXACTLY. The allowed sections are:

1. STORE ANALYSIS — {domain}
2. BOTTOM LINE — biggest issue + quick win
3. SNAPSHOT — store facts + speed + trust signals in one block
4. FINDINGS — flat list ordered by business impact, NOT grouped by dimension
5. SAMPLE FIXES — 2-3 real examples with Problem/Fix/Result
6. CONVERSION GAPS — missing elements real shoppers expect
7. AI-FACING GAPS — questions AI agents cannot answer from this store's content

Nothing else. No EXECUTIVE SUMMARY, no DIMENSION SUMMARY, no SCORECARD, no SEO/GEO/AEO/CRO FINDINGS headers, no TOP ACTIONS, no WHAT'S WORKING checkmark lists, no scores, no sign-off paragraphs, no "This is a diagnostic benchmark."

Format rules

• Full audit: 70-160 lines max. If you exceeded 160 lines, you wrote too much.
• Each finding: exactly 4 lines — Scope, Proof, Fix, Impact. One line each.
• Proof = ONE line with exact data (numbers, not adjectives).
• Fix = ONE concrete action in ONE line.
• Impact = ONE line saying what improves.
• SAMPLE FIXES: 2-3 items, each with Problem/Fix/Result (one line each).
• AI-FACING GAPS: 4-6 specific questions AI agents can't answer from this store today.
• NO checkmark lists (✓/✗) in FINDINGS. The SNAPSHOT trust line is the only exception.
• NO scores or numerical ratings (no X/100, no X/Y) in FINDINGS. SNAPSHOT speed metrics are raw data, not scores.
• NO "What this means" explanations. The 4-line format is self-explanatory.

Workflow

1. Confirm the task mode

• Full audit: all eight modules.
• Focused audit: user wants a specific area (e.g., "check AI readiness" = GEO focus).
• Audit review: user provided an existing audit; verify its claims.

2. Normalize domain and confirm store is auditable

• Strip scheme, www., trailing slashes, path fragments, query strings.
• Fetch store metadata and products endpoint first.
• If public catalog data is blocked, explain the limitation and continue with sampled HTML only.

3. Collect catalog-wide data

Fetch using any available method (HTTP requests, browser, CLI curl):

1. https://{domain}/meta.json
2. https://{domain}/products.json?limit=250&page={n} — paginate until < 250 returned
3. https://{domain}/collections.json?limit=250
4. https://{domain}/pages.json
5. https://{domain}/robots.txt
6. https://{domain}/sitemap.xml
7. https://{domain}/llms.txt

4. Collect sampled HTML + CRO/trust/speed data

Fetch 30-50 representative pages. Extract ALL signals in one pass per page — SEO, CRO, and trust checks all run against the same HTML. Do not fetch any page twice.

Pages to fetch:

• 15-20 product pages (varied handles, price points, and collections)
• 8-10 collection pages (prioritize collections with most products)
• Homepage, about page, contact page
• 3-5 blog articles (if blog detected from sitemap)
• Policy pages (shipping, returns, FAQ)

Extract from every page: title, meta description, canonical, H1, JSON-LD blocks, OG/Twitter tags, internal link patterns, external scripts, image attributes, question headings, lists/tables.

Additionally extract per cro-checks.md and trust-ux-checks.md:

• Homepage: hero CTA buttons + text, hero heading, footer (payment icons, policy links, social links, newsletter form), email capture scripts, chat widget scripts
• Product pages: cart drawer/redirect indicators, trust badges near ATC, size chart links, shipping info blocks, return policy blocks, express checkout buttons, review widget + star ratings, image count, product badges
• Contact page: phone, address, email domain, contact form, chat widget
• About page: word count, real content vs placeholder

Page speed (separate API call):

• curl "https://www.googleapis.com/pagespeedonline/v5/runPagespeed?url=https://{domain}&category=performance&strategy=mobile"
• Extract: performance score, LCP, CLS, FCP. Count third-party <script> tags on homepage.
• If PSI returns non-200 or times out after 30 seconds, report "PageSpeed data unavailable" in SNAPSHOT and skip speed findings. Do not retry more than once.

5. Analyze across all eight modules

Use the data already collected in Steps 3-4. Do not fetch additional pages unless a specific check requires a page not yet sampled. Run checks from each reference file across all eight modules. Start with Trust & CRO — these are what real store reviewers flag first and what store owners care most about. The reference files are comprehensive checklists — not every failed check becomes a finding. Use the Finding Quality Bar to determine which issues earn a spot in the FINDINGS section.

Module priority order:

1. Trust & Credibility → trust-ux-checks.md §1-4
2. CRO & Conversion → cro-checks.md §1-5
3. Page Speed → trust-ux-checks.md §5
4. Technical SEO → seo-checks.md §3-4, 8-9
5. Product-Page SEO → seo-checks.md §1-2, 5-7
6. Structured Data → seo-checks.md §1 + catalog-checks.md
7. AEO Readiness → aeo-checks.md
8. GEO Citation → geo-checks.md

6. Select findings

Apply the Finding Quality Bar:

• Does it have provable revenue/traffic impact?
• Is it proportional to this store's maturity?
• Is it noise or signal?
• Can it be merged with a related issue?

For established stores: 4-6 findings. For struggling stores: up to 8-10. Order by business impact, not by dimension.

7. Write the report

Use assets/report-template.md. Include: BOTTOM LINE, SNAPSHOT, FINDINGS, SAMPLE FIXES, CONVERSION GAPS, AI-FACING GAPS.

8. Validate

• Use references/troubleshooting.md if data was partial.
• If saved to file, run python ${CLAUDE_SKILL_DIR}/scripts/check_report.py --input path/to/report.md --mode full|focused|review.

9. Hand off implementation

If the user wants fixes:

• Do NOT execute from store-analyzer
• Hand off to store-fixer with the prioritized fix list
• State that store-fixer requires explicit approval before any write and a rollback path

Audit Review Mode

When verifying an existing audit, classify each claim as:

• Supported and important
• Supported but secondary
• Real but overstated
• Unsupported from current evidence
• Missing important issues

Examples

Example 1: Audit this Shopify store: neemans.com Result: Full 8-module audit with 4-5 focused findings, conversion gaps, and AI-facing gaps.

Example 2: Check AI visibility for this store Result: GEO-focused audit — AI bot access, entity signals, content extractability, citation readiness.

Example 3: Review this SEO audit and tell me what matters Result: Verify claims against live data, classify by importance, flag missing GEO/AEO dimensions.

Example 4: Can ChatGPT find products from this store? Result: GEO-focused — robots.txt AI bot rules, content extractability, entity clarity.

Example 5: Check structured data on this Shopify store Result: Focused audit on Structured Data module — Product/Offer schema, merchant listing fields, review markup.

Example 6: Implement the fixes from this audit Result: Hand off to store-fixer. Do not execute changes from store-analyzer.

Writing Replies & Outreach From Audit Results

When the user asks to draft a community reply, DM, or outreach message based on an audit:

Rules

1. Facts only, no unsourced advice. State findings with exact numbers. Do not prescribe fixes unless citing Google Search Central, Google Merchant Center, or Shopify docs.
2. No AI/SEO jargon. Never use: "long-tail ranking", "structured data", "SERP presentation", "conversion gaps", "unlock". Write like a human who looked at the store.
3. Frame as buyer perspective. "A buyer can't find what phones fit" not "Your meta description is missing."
4. Never prescribe word counts. Don't say "write 300 words." Say what information is missing.
5. Open with a genuine compliment. Find one real thing done well.
6. AI visibility = one-liner, not the lead. Most store owners don't think about AI yet.
7. End with an offer, not a pitch. "I have the full report if useful" — not a tool pitch.

Template

Hey — [genuine compliment about one thing done well]. I ran an automated audit on your store and a few things stood out:

- **[Finding 1 as fact]** — [buyer-perspective impact in plain English]
- **[Finding 2 as fact]** — [plain English]
- **[Finding 3 as fact]** — [plain English]

I also tested whether AI assistants (ChatGPT, Perplexity) could answer basic questions about [product] — [one-line summary of gaps].

I have the full report if you'd find it useful. Happy to share.

When Not To Use This Skill

• The site is not a Shopify storefront.
• The user wants backlink analysis, competitor gap analysis, or local SEO.
• The task is purely copywriting with no audit component.
• The user needs authenticated data from Search Console, GA4, or Shopify admin.
• The user wants to FIX issues. Use store-fixer (requires approval + rollback).

Maintenance Notes

For trigger tuning, regression checks, and eval prompts, use:

• references/testing.md
• evals/evals.json

Keep this file focused on workflow and rules. Detailed checklists go in reference files.