Competitor Audit — Website Teardown & Scoring

Audit a competitor's website on 5 dimensions. Produce a grade (A+ to F), a competitive verdict per dimension (Threat / Opportunity / Parity), and prioritized recommendations for the user's own site.

This skill is self-contained and runs end-to-end on its own.

Input

• Competitor URL (required). If none provided, ask for one.
• Your own site URL (optional). If provided, every dimension is judged comparatively and the report frames threats and opportunities relative to it. If absent, run a standalone audit and skip the comparative verdicts (mark them "N/A — no reference site").
• Optional: industry context (SaaS, fintech, ecommerce, media, services). If not given, detect it from the homepage.

Step 1 — Decide the page set

A competitor audit is never just the landing page. Before fetching, settle on a page set:

1. Homepage — always.
2. Pricing page — always if it exists. If pricing is gated ("contact sales"), note it and audit what is visible.
3. One product / feature / offer page — pick the most prominent one from the nav.
4. About / company page — for trust signals and positioning.

Find these by fetching the homepage first and reading its navigation. List the 3-4 URLs you will audit before going further.

Step 2 — Capture the pages

For each page in the set:

1. If Chrome DevTools MCP is available (preferred):
   • Navigate to the URL. Take a screenshot at desktop (1440px) and mobile (390px).
   • Take a DOM snapshot. Use evaluate_script to count CTAs, form fields, trust badges, pricing tiers, structured-data blocks.
   • Run a Lighthouse audit on the homepage for real performance/accessibility/SEO data.
2. Fallback — WebFetch for text, copy, and SEO signals.

JS-heavy sites (React/Next SPAs — most modern SaaS) render empty or partial under WebFetch. Use Chrome DevTools for anything that depends on the rendered DOM: messaging, CTAs, pricing tables, schema. Use WebFetch only for static text and meta tags.

Never fabricate a Lighthouse score. If you cannot run Lighthouse, assess speed qualitatively and say so explicitly.

Step 3 — Score the 5 dimensions

Each dimension is scored 0-100 using its anchors below. The weighted aggregate gives the overall score.

For every issue you log, use this finding format. Each finding must reference something actually observed on a page — never generic advice.

Issue — what is wrong (quote the headline, name the page). Impact — why it costs them, or why it threatens the user. Evidence — the exact element: a quoted line, a screenshot reference, a DOM count. Fix — a specific, actionable change.

1. Positioning & Messaging (weight: 20%)

• Is the value proposition clear in under 5 seconds?
• Is the target audience named, or is it "for everyone"?
• Is there a category frame and a clear differentiator vs alternatives?
• Is the message consistent across homepage, product page, and about?
• 90+: Specific value prop + named audience + clear differentiator, consistent across pages.
• 70-89: Clear but generic positioning. "The modern way to X."
• 50-69: Feature-led, no audience, no differentiation.
• <50: No discernible value prop; visitor cannot tell what it is or who it is for.

2. Conversion / CRO (weight: 25%)

Condensed CRO logic — assess these sub-points and score the dimension as a whole:

• Headline: answers "what is this and why care?" — specific over vague.
• CTA: one primary CTA above the fold, action-specific text, visually distinct, repeated at scroll points.
• Social proof: named testimonials, logos, real metrics — placed near CTAs.
• Trust signals: HTTPS, real company info, guarantee / free trial / no-card-needed.
• Form friction: how many fields to start; one-click or single-email is best.
• Mobile: layout holds at 390px, CTA is a 44px+ thumb-reachable target.
• Speed: fast load feel; use the Lighthouse performance score if captured.
• 90+: Single sharp CTA above fold, strong proof near it, low friction, fast, flawless mobile.
• 70-89: Converts but with generic CTA copy, weak proof, or minor friction.
• 50-69: CTA buried or competing, thin proof, 4+ form fields, cramped mobile.
• <50: No clear CTA, no proof, heavy friction, broken mobile.

3. SEO & GEO (weight: 20%)

Classic SEO plus GEO (visibility inside generative engines — ChatGPT, Perplexity, Google AI overviews).

• Technical: indexable, crawlable, Core Web Vitals (LCP <2.5s, INP <200ms, CLS <0.1).
• On-page: title 50-60 chars, meta description 150-160 chars, single H1, keyword in the first 100 words.
• Content & E-E-A-T: depth, named authors, evidence, recency.
• Structured data: schema.org / JSON-LD present and valid (Organization, Product, FAQ).
• GEO: citable answer-shaped content, clear entity definitions, llms.txt, content that an AI can quote verbatim.
• 90+: Clean technical base, strong on-page, valid schema, content built to be cited by AI engines.
• 70-89: Solid classic SEO, little to no GEO consideration.
• 50-69: Gaps in technical or on-page basics; no schema.
• <50: Indexation problems, missing meta, no structure.

If structured data is injected client-side it may not show in WebFetch — verify via the rendered DOM or say "could not verify". Never invent it.

4. Offer & Pricing (weight: 20%)

• Is pricing transparent and easy to find, or hidden behind "contact sales"?
• Is the plan structure clear, with sensible anchoring and a guided default?
• Is what each plan includes legible at a glance?
• Is there a free trial, freemium tier, or guarantee that lowers the entry risk?
• Is the perceived value vs price defensible?
• 90+: Transparent pricing, clean tiers with a clear anchor and recommended plan, low-risk entry.
• 70-89: Pricing visible but tiers are confusing or value is unclear.
• 50-69: Pricing partly hidden, or packaging is hard to parse.
• <50: No visible pricing, no entry path, opaque packaging.

Gated pricing is a deliberate choice for some segments — score what is visible, do not over-penalize an enterprise model. Pricing changes often: screenshot it and date-stamp the audit as a point-in-time snapshot.

5. Design & UX (weight: 15%)

• Visual hierarchy: does the eye land on the right things?
• Typography and spacing: consistent, readable, deliberate.
• Component and brand consistency across pages.
• Navigation: predictable, shallow, labelled clearly.
• Responsive behaviour and loading/empty/error states.
• Accessibility: contrast (WCAG AA), touch targets, alt text.
• Dark patterns: flag manipulative UX (forced continuity, confirmshaming, hidden costs).
• 90+: Coherent, accessible, confident design that supports the message.
• 70-89: Clean but generic, or small inconsistencies.
• 50-69: Dated, inconsistent, or accessibility gaps.
• <50: Broken layout, illegible, or dark patterns present.

Step 4 — Calculate the overall score

Overall = (Positioning x 0.20) + (Conversion x 0.25) + (SEO_GEO x 0.20) +
          (Offer_Pricing x 0.20) + (Design_UX x 0.15)

Grade scale:

• A+ (95-100): Exceptional. A serious competitive threat.
• A (90-94): Strong across the board.
• B+ (85-89): Good, with one or two real gaps.
• B (80-84): Solid foundation, clear weak spots.
• C (70-79): Significant issues. Beatable.
• D (60-69): Major problems. Wide open.
• F (<60): Fundamentally weak.

If a dimension cannot be assessed, score it "N/A" and exclude it from the weighted average (re-normalize the remaining weights).

Step 5 — Competitive verdict per dimension

For each dimension, if the user's own site URL was provided, audit it on the same dimension and assign a verdict:

• Threat — the competitor is clearly stronger here. The user is exposed.
• Opportunity — the competitor is weak here. The user can win this ground.
• Parity — roughly equal; not a deciding factor.

If no reference site was provided, skip verdicts and mark them "N/A — no reference site". The rest of the audit still stands as a standalone teardown.

Step 6 — Output

Write the report as a markdown block.

## Competitor Audit: [Competitor name] — [URL]

**Date**: [YYYY-MM-DD]
**Industry**: [detected or specified]
**Pages audited**: [list]
**Reference site**: [user's site URL, or "none"]
**Overall Score**: XX/100 (Grade: X)

### Snapshot

| Dimension | Score | Weight | Verdict |
|-----------|-------|--------|---------|
| Positioning & Messaging | XX/100 | 20% | Threat / Opportunity / Parity |
| Conversion / CRO | XX/100 | 25% | ... |
| SEO & GEO | XX/100 | 20% | ... |
| Offer & Pricing | XX/100 | 20% | ... |
| Design & UX | XX/100 | 15% | ... |

### Dimension detail

[For each of the 5 dimensions: score, a one-line verdict rationale,
then the findings in Issue / Impact / Evidence / Fix format.
What they do well = threats to the user. Their weaknesses = the user's openings.]

### Strategic takeaways

**Top 3 threats** — where the competitor beats the user:
1. ...
2. ...
3. ...

**Top 3 opportunities** — gaps the user can exploit:
1. ...
2. ...
3. ...

### Prioritized recommendations (for the user's own site)

| # | Action | Effort | Expected impact |
|---|--------|--------|-----------------|
| 1 | ... | Low/Med/High | ... |
| 2 | ... | ... | ... |
| 3 | ... | ... | ... |

### Screenshots
[Reference the desktop and mobile captures taken, if any.]

Rules

• Every finding references something actually observed on a page. "Their hero says 'The pro bank that...'" is useful. "Improve the messaging" is not.
• Recommendations are framed for the user's own site, never as advice to the competitor.
• Do not fabricate Lighthouse scores, schema, or metrics. If you cannot verify, say "could not verify".
• Treat the audit as a point-in-time snapshot. Date-stamp it. Pricing and copy change.
• If a page is blocked (Cloudflare, anti-bot) or gated, mark the dimension partial and score only what is visible.
• Keep dimension detail tight: a few sharp findings beat a long generic list.