nitingothwal-SEO/claude-skills

Competitor Online Marketing Reverse Engineering

Goal: explain HOW a competitor acquires, converts and retains customers, and where their money goes — then turn that into decisions for the user's own site. Evidence first, conclusions second, actions last.

1. Inputs (ask once, in a single AskUserQuestion if anything is missing)

• Competitor domain(s) — 1 to 5. If none given, discover them (step 2).
• User's own domain (for gap comparisons).
• Target market: country/city + language (drives Ubersuggest location, ad library region, AI prompts).
• Depth: Quick (phases 1–5 + synthesis, ~1 competitor) or Full (all phases, up to 5 competitors). Default: Full for 1–2 competitors, Quick for 3+.
• Business model hint if unclear: B2B service, B2C ecommerce, SaaS, local business.

If the user is not present, take the most reasonable reading, state it at the top of the report, and continue.

2. Competitor selection (only if not supplied, or to sanity-check)

Build three lists and pick the 3–5 most relevant:

• Business competitors — same offer, same buyer (WebSearch " ", "best companies ").
• Search competitors — Ubersuggest competitors on the user's domain + who ranks top 10 for 5–10 money keywords (serp_analysis).
• Attention competitors — brands cited in AI answers and running ads in the niche.

3. Data collection by phase

Set up a task list (one task per phase + synthesis + verification). Save raw pulls to files in the working directory (e.g. data/<domain>_top_pages.json) — compute from files, never estimate figures in prose.

Load Ubersuggest tools via ToolSearch in one batch before starting, e.g. select:mcp__Ubersuggest__domain_overview,mcp__Ubersuggest__domain_top_pages,mcp__Ubersuggest__domain_keywords,mcp__Ubersuggest__domain_top_countries,mcp__Ubersuggest__backlinks_overview,mcp__Ubersuggest__linking_domains,mcp__Ubersuggest__anchor_texts,mcp__Ubersuggest__competitors,mcp__Ubersuggest__serp_analysis,mcp__Ubersuggest__match_keywords,mcp__Ubersuggest__pagespeed_audit,mcp__Ubersuggest__brand_visibility_overview. Check user_limits first if doing many domains.

Phase 1 — Positioning & messaging

• WebFetch homepage, pricing, about, top service pages. Record: headline, subheadline, primary CTA, target audience, every claim and every proof point (logos, numbers, testimonials, awards).
• Pricing model: public/hidden, tiers, trial/demo, anchoring. Hidden pricing ⇒ sales-led funnel.
• History: Wayback Machine snapshots of homepage + pricing (≈ 1, 2, 3 years ago) — note positioning shifts.
• Voice of customer: reviews on G2/Clutch/Trustpilot/Google/Reddit via WebSearch. Extract top 3 praises (table stakes) and top 3 complaints (positioning openings). Paraphrase; never paste review text.

Phase 2 — Traffic mix & growth engines

• domain_overview, domain_top_countries, domain_top_pages.
• Identify the 10–20 pages driving most traffic = real growth engines. Classify each (service, location, blog, tool, comparison, glossary).
• Channel split (direct/organic/paid/referral/social): if no tool gives it, say so and infer cautiously from paid keywords, ad library activity and social footprint. Label all third-party numbers as directional estimates.

Phase 3 — SEO

• Keyword footprint: domain_keywords; split by intent and branded vs non-branded (compute in code).
• Keyword gap vs user's domain: keywords competitor ranks 1–10 that user doesn't; prioritise commercial/transactional.
• Content strategy: fetch sitemap.xml (WebFetch) → page-type inventory, publishing velocity from lastmod, clusters, programmatic patterns (e.g. service × city templates).
• Technical: schema types present, URL structure, internal-link priority pages, hreflang, pagespeed_audit on homepage + top money page.
• Backlinks: backlinks_overview, linking_domains, anchor_texts. Note referring-domain trend, top linked pages (linkable assets), anchor profile (exact-match heavy = aggressive), link source types. Link intersect when 2+ competitors: domains linking to ≥2 competitors but not the user.

Phase 4 — AI search visibility (GEO/AEO)

• Build 15–30 buyer prompts for the market (pricing, best-of, how-to, comparison).
• Use brand_visibility_overview / brand_prompts if configured; otherwise record what is verifiable via WebSearch and note manual checks the user should run in ChatGPT, Perplexity, Gemini, AI Overviews.
• Note cited-page patterns (definitions, tables, price ranges, stats, FAQs, freshness) and off-site sources feeding LLMs (Wikipedia/Wikidata, Reddit, YouTube, review sites, listicles).
• Check robots.txt for AI crawler rules and presence of /llms.txt.

Phase 5 — Paid advertising

Use the browser (read the chrome-browser skill first; Claude in Chrome is preferred) to view public libraries — read only, never click ads or submit anything:

• Google Ads Transparency Center (adstransparency.google.com) — search advertiser, set region.
• Meta Ad Library (facebook.com/ads/library) — active ads, start dates, variants.
• LinkedIn Ad Library (B2B), TikTok Creative Center (B2C). For each notable ad log: platform, start date, format, hook, offer, CTA, landing page. Flag ads running 60+ days as likely profitable. Note creative testing (same offer, multiple hooks) and dedicated vs generic landing pages. Paid keywords from Ubersuggest where available. If no browser is available, say which libraries the user should check and continue.

Phase 6 — Funnel & CRO

From fetched pages: conversion paths (form, chat, phone, WhatsApp, booking, calculator), lead magnets, form field count and qualifying questions, trust element placement, popups/sticky CTAs. Recommend (do not perform) a mystery-shop enquiry by the user: response time, contact channel, script, follow-up cadence. Never submit forms or enquiries yourself.

Phase 7 — Tech stack & tracking

From page source (WebFetch / browser): CMS, analytics, tag manager, CRM, email platform, chat, A/B testing, heatmaps, pixels (Meta, LinkedIn Insight, TikTok, Google Ads remarketing, Clarity, Hotjar). Interpret: e.g. HubSpot ⇒ automation-driven nurture; VWO/Optimizely ⇒ testing culture.

Phase 8 — Email & automation

You cannot subscribe. Document visible opt-ins and lead magnets; give the user a 4–6 week tracking checklist (welcome sequence, cadence, subject lines, content mix, segmentation) to run with a dedicated inbox.

Phase 9 — Social & distribution

Per platform: presence, posting frequency, formats, top-performing themes (from what is publicly visible via WebSearch/WebFetch/browser), founder/employee personal brands, YouTube rankings, communities/webinars. Engagement rate = engagement ÷ followers when numbers are visible.

Phase 10 — PR, partnerships & hiring signals

WebSearch news coverage, awards, sponsorships, partner/integration pages, affiliate programme, and current job posts (careers page, LinkedIn) — roles being hired signal next 3–6 months of investment.

4. Synthesis (mandatory — this is the deliverable's core)

1. Marketing model — one sentence per competitor: how they acquire → convert → nurture.
2. Channel matrix — rows = channels (positioning, SEO, AI visibility, paid search, paid social, content, social, email, PR/links, CRO), columns = competitors + user; cells Weak/Average/Strong with one line of evidence each.
3. Opportunity list — every gap scored Impact (1–5), Effort (1–5), Time-to-result (weeks). Sort by impact ÷ effort.
4. Copy / Beat / Skip — table stakes to copy, weaknesses to beat, budget/brand-dependent plays to skip.
5. 90-day plan — month-by-month actions with owner role and KPI.
6. Monitoring setup — Google Alerts, monthly ad library + top pages check, competitor rank tracking, quarterly AI prompt re-run, change tracking on homepage/pricing.

5. Output

• Main report: a Docs artifact (list artifact types first; fall back to a self-contained HTML artifact, or .docx only if the user asks for Word). Structure: Executive summary (5 bullets) → Competitor snapshots → Phase findings → Channel matrix → Opportunities → Copy/Beat/Skip → 90-day plan → Monitoring → Data gaps & manual checks → Sources.
• Optional companion: Sheets artifact with tabs Competitors, Keyword Gap, Top Pages, Backlink Intersect, Ads Log, AI Visibility, Opportunity Scoring — offer it; build if requested or if Full depth with 3+ competitors.
• Language: match the user's language for prose (English or Hinglish); keep tool names, metrics and headings in English.

6. Rules

• Public information and genuine interactions only. No fake identities, no logging in, no scraping behind logins, no form submissions, no clicking competitor ads (costs them money).
• Never copy competitor content, ad copy or reviews verbatim; paraphrase and quote at most one short phrase per source.
• Label every third-party traffic/spend number as an estimate; compare competitors with the same tool.
• Correlation ≠ causation: draw conclusions from patterns across several pages/competitors, not one page.
• Separate evidence from inference in the report (e.g. "Observed:" vs "Likely:").
• List every phase you couldn't fully cover (tool limits, no browser, paywalled data) under "Data gaps & manual checks" with the exact manual step.
• Final check before delivering: verify numbers against saved data files, re-open the report, confirm every recommendation traces back to a finding.