SEO Competitor Pages Skill for Claude

> Example output: examples/seo-competitor-pages-linear-vs-jira-20260514/COMPARISON.md

Competitor Comparison & Alternatives Pages

Produce conversion-tuned landing pages targeting comparative-intent keywords ("X vs Y", "alternatives to X", "best X for Y"). The deliverable is a paste-ready page draft with feature matrix, balanced verdict, schema markup, and a CTA flow that converts comparison-stage traffic.

Page types this skill produces

• "X vs Y" head-to-head — direct comparison between two products/services. Target keyword: [Product A] vs [Product B].
• "Alternatives to X" — listicle-format page positioning the user's product as one of N alternatives to a category leader. Target keyword: [Competitor] alternatives or alternatives to [Competitor].
• "Best X for Y" — segmented best-of page targeting a use case or audience. Target keyword: best [category] for [use case].

Prerequisites

• SE Ranking MCP server connected.
• Claude's WebFetch tool available.
• User provides: (a) the user's brand/product (the page's hero), (b) target competitor(s) — at least one, optionally up to 5 for an alternatives page, (c) page type (auto-detected from the keyword if user doesn't specify), (d) target country (default us).

Process

• Validate & determine page type
• From the user's input, detect: vs / alternatives / best-of.
• If page type unclear, ask the user. Don't guess silently.

• Pull competitor context DATA_getDomainCompetitors
• For the user's domain, list top organic competitors by common_keywords overlap.
• Validate that the user's named competitor is in the list (or close).

• Pull keyword data per brand DATA_getDomainKeywords
• For the user's domain and each named competitor, pull top 100 organic keywords.
• Identify: keywords each brand owns exclusively, keywords both rank for, gaps.

• Pull comparative SERPs DATA_getSerpResults and DATA_getKeywordQuestions
• For "X vs Y" / "alternatives to X" / "best X for Y" target keyword(s):
• Top 10 organic results — who else ranks for this comparative keyword?
• PAA questions (these become FAQ section content).
• Featured snippet (if present, capture the answer pattern).

• Fetch existing comparison pages WebFetch (always) + mcp__firecrawl-mcp__firecrawl_scrape (when available)
• WebFetch first (free): pull the top 3 SERP winners' markdown. Extract H2 outline, feature-matrix dimensions, verdict pattern, CTA placement.
• Firecrawl second (3 Firecrawl credits — 1 per winner) — recovers what WebFetch markdown can't show:
• Schema types from <script type="application/ld+json"> blocks (Product ×N, BreadcrumbList, FAQPage, Review, AggregateRating).
• og:title / og:description / og:image / twitter:card from metadata.
• <title> and meta description lengths from real HTML.
• If Firecrawl unavailable: WebFetch portion runs unchanged. The "schema types" line in evidence/04-existing-pages-teardown.md reads (skipped — Firecrawl required for JSON-LD). Schema generation in step 8 falls back to a default Product + BreadcrumbList + FAQPage template instead of mirroring whatever the winners use.
• This anchors the draft in observed-rewarded-pattern.

5b. Bulk competitor scrape mcp__firecrawl-mcp__firecrawl_scrape (optional, opt-in)

• When user passes --bulk-scrape <urls> or supplies a list of competitor URLs to compare directly (beyond the SERP top-3), Firecrawl-scrape each URL.
• For each URL, extract: <title>, og:*, twitter:*, JSON-LD @types, hero-image presence, pricing-block detection (regex on prose for $N/mo, €N/mo, etc.), CTA count, comparison-table presence, free-tier-mention boolean.
• Output competitor-elements.csv — one row per competitor URL × these signals.
• Cost: 1 Firecrawl credit per URL. Surface estimate before running; refuse >50 URLs without --confirm-cost.

• Pull keyword comparison data DATA_getDomainKeywordsComparison (if available for the brands)
• Side-by-side keyword overlap.

• Build feature matrix
• Dimensions inferred from the top SERP winners (e.g., "Pricing", "Free tier", "Integrations", "Support tiers", "Best for").
• Cells: ✓ / ✗ / partial / "TBD — confirm with PM" placeholders for fields you can't auto-infer.
• Where SE Ranking data informs a cell (e.g., "ranks for X enterprise keywords"), pull the number.

• Synthesise COMPARISON.md
• Hero (target keyword in H1, balanced positioning).
• TL;DR / verdict box in first 200 words.
• Feature matrix.
• Section per major dimension (each H2 = one dimension).
• PAA-derived FAQ (top 3–5 questions from step 4).
• Verdict / recommendation.
• CTA flow.
• Schema-ready JSON-LD: Product (×N) + BreadcrumbList + FAQPage (if real Q&A).

Output format

Create a folder seo-competitor-pages-{target-slug}-{YYYYMMDD}/ with:

seo-competitor-pages-{target-slug}-{YYYYMMDD}/
├── COMPARISON.md                     (the page draft — primary deliverable)
├── 05-feature-matrix.md              (inferred dimensions × brands — load-bearing reference for PMs/writers)
├── schema.jsonld                     (paste-ready Product + Breadcrumb + FAQ — load-bearing artefact for engineering)
├── 05b-competitor-elements.csv       (only if --bulk-scrape ran: competitor URL × on-page-element grid)
└── evidence/
    ├── 01-competitor-context.md      (DATA_getDomainCompetitors — raw step output)
    ├── 02-keyword-overlap.md         (DATA_getDomainKeywords for each brand — raw step output)
    ├── 03-comparative-serp.md        (top 10 + PAA for the target keyword — raw step output)
    └── 04-existing-pages-teardown.md (top-3 SERP winners' structure + schema/og — Firecrawl-recovered)

Top-level: COMPARISON.md + 05-feature-matrix.md + schema.jsonld. The 01–04 step files preserve raw API/scrape outputs in evidence/. 05b-competitor-elements.csv only appears when --bulk-scrape was passed.

COMPARISON.md for an "X vs Y" page follows this shape:

# {User's Brand} vs {Competitor}: 2026 Comparison

> Updated {YYYY-MM-DD}. Compare {Brand A} and {Brand B} on pricing, features, integrations, and best-fit use case.

## TL;DR
{One paragraph balanced verdict — when to choose A, when to choose B}

## At a glance

| Dimension | {Brand A} | {Brand B} |
|---|---|---|
| Starting price | {$X/mo} | {$Y/mo} |
| Free tier | {✓/✗} | {✓/✗} |
| Best for | {use case} | {use case} |
| Integrations | {n} | {n} |
| Support | {tier} | {tier} |
| ... | | |

## {Dimension 1 header — e.g., Pricing}
{Side-by-side detail, balanced. Avoid hyperbole.}

## {Dimension 2 header — e.g., Features}
{...}

## {Dimension 3 header — e.g., Integrations}
{...}

## {Dimension 4 header — e.g., Support}
{...}

## When to choose {Brand A}
- {scenario 1}
- {scenario 2}
- {scenario 3}

## When to choose {Brand B}
- {scenario 1}
- {scenario 2}
- {scenario 3}

## FAQ
**{PAA question 1}**
{Answer}

**{PAA question 2}**
{Answer}

**{PAA question 3}**
{Answer}

## Get started
{Brand A} CTA — {link}
{Brand B} CTA — {link if balanced; otherwise drop}

## Schema
See `schema.jsonld` — paste into `<head>`.

Tips

• Balance is conversion. Pages that pretend the user's product is always better lose trust and rankings. Honest assessments outperform partisan ones.
• Respect rate limit. Step 5 (fetching top 3 SERP winners) takes 3 WebFetch calls + earlier MCP queries.
• Cost: ~15–25 SE Ranking credits typical, +3 Firecrawl credits for the schema/og benchmark in step 5, +1 Firecrawl credit per URL in step 5b (opt-in only). Pass --no-firecrawl to skip both Firecrawl steps.
• Schema: use Product for both products in a vs page, plus BreadcrumbList, plus FAQPage if the FAQ section is real Q&A (not a manufactured one).
• For "alternatives to X" pages, position the user's product as one of N (typically 5–10), not as #1. Numbered listicles convert better than self-promotional alternatives pages.
• For "best X for Y" pages, segment by use case explicitly — "best for solo developers" vs "best for enterprise teams" — this lets you win multiple long-tail variants.
• Run seo-page on the published page after 30 days to track ranking trajectory.
• Pair with seo-content-audit to E-E-A-T-check the draft before publishing.
• The PAA-derived FAQ in step 4 is gold — those are the questions users are actually searching, and answering them in-page raises citation-readiness for AIO.
• Don't auto-publish. Hand the draft to a writer/PM for fact-checking and brand-voice tuning.