Kensink Labs
★ Framework14 sub-systemsReference
FRAMEWORK · 14 SUB-SYSTEMS

The full audit framework we run on every engagement.

Every check, every target state, every red flag we’ve seen blow up in production. This is the working reference. Bookmark it, send it to your dev team, audit yourself against it before we even start.

Book an audit →← Back to overview
§ 5.101 / 14

Crawlability & Indexation

Can search engines and AI crawlers reach every page that should be indexed, and nothing that shouldn’t?

Checks
  • robots.txt present, valid, not blocking critical paths
  • sitemap.xml referenced in robots.txt and contains only indexable URLs
  • Single canonical domain: all four variants 301 → one canonical
  • Self-referencing canonical tags on indexable pages
  • noindex directives audited, present only on pages we want excluded
  • Indexation coverage (submitted vs indexed): any gap explained, not accidental
  • Soft 404s: zero tolerated on important pages
  • Orphan pages identified, then either linked or deleted
  • Crawl budget concentrated on valuable URLs (log-file analysis)
  • Rendered HTML === raw HTML for content (JavaScript rendering parity)
Red flags
  • Paginated archive URLs + faceted navigation eating crawl budget
  • Staging / preview URLs indexed by accident
  • noindex left over from a migration
  • Crawlers blocked at the WAF or CDN (common with Cloudflare defaults)
§ 5.202 / 14

Site Architecture & Internal Linking

Does the structure communicate topical authority and let humans + machines discover important content in ≤3 clicks?

Checks
  • Click depth ≤ 3 from homepage to any important page
  • Zero orphan pages
  • Zero broken internal links
  • Max 1-hop redirect chains
  • Descriptive anchor text, no ‘click here’ on important links
  • Breadcrumbs present + marked up with BreadcrumbList schema
  • Crawlable pagination, no pure infinite scroll
  • Evidence of coherent pillar-cluster topic structure
  • Mega-menu link dilution avoided, no pages with 300+ internal links
  • Pillar pages receive many contextual links from clusters
§ 5.303 / 14

Performance & Core Web Vitals

Do pages load fast enough that humans stay and crawlers allocate budget generously?

Checks
  • LCP ≤ 2.5s (field data, not lab)
  • INP ≤ 200ms
  • CLS ≤ 0.1
  • TTFB ≤ 400ms
  • TBT ≤ 200ms
  • Field data (CrUX / GSC) used as source of truth, with lab data only for diagnosis
Red flags
  • Unoptimized hero images (not WebP/AVIF, not responsive, no width/height)
  • Render-blocking third-party scripts (chat widgets, GTM, analytics)
  • Font swap causing CLS
  • Ads or embeds injected after load pushing content down
  • Bloated JS bundles from drag-and-drop page builders
§ 5.404 / 14

Security, HTTPS & Status Codes

Is the site served securely, consistently, and without silent errors?

Checks
  • SSL valid, covers all subdomains, not about to expire
  • All HTTP → HTTPS via 301
  • No mixed content warnings
  • HSTS enabled (Strict-Transport-Security header)
  • Content-Security-Policy header present and sensible
  • No 5xx errors in status-code crawl
  • No accidental X-Robots-Tag: noindex in production headers
  • Security headers graded A or A+ at securityheaders.com
  • WAF / bot rules don’t block legitimate AI crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended)
§ 5.505 / 14

On-Page Technical & Semantic HTML

Does every page’s markup clearly communicate what it is and what it’s about?

Checks
  • Title tag: unique, 50–60 chars, primary keyword near front
  • Meta description: unique, 140–160 chars, CTA intent included
  • Exactly one H1 per page, describes the page topic
  • H2–H6: logical hierarchy, no skipped levels
  • Semantic HTML5 used correctly: article, section, nav, main, aside, header, footer
  • Descriptive alt attributes on all non-decorative images
  • lang attribute set on
  • Open Graph + Twitter Card complete on shareable pages
  • hreflang (if multi-locale): reciprocal, self-referencing, valid codes
  • Canonical: self-referencing on indexable pages
Red flags
  • div-soup styled to look like headings, invisible to screen readers and LLM parsers (the most common failure on modern React/Vue sites)
§ 5.606 / 14

Structured Data & AI-Readiness

Can LLMs, answer engines, and rich-result pipelines parse our site into structured facts they can cite correctly?

Checks
  • Homepage: Organization + WebSite (with SearchAction)
  • Articles: Article / BlogPosting + Author (Person) + BreadcrumbList
  • Products: Product + Offer + AggregateRating + Review
  • Services: Service + Organization
  • FAQs (real ones only): FAQPage
  • HTML fully rendered server-side or statically generated
  • llms.txt published at root
  • robots.txt explicitly allows GPTBot, ClaudeBot, PerplexityBot, Google-Extended, CCBot (unless strategically blocked)
  • WAF rules don’t shadowban the above
  • Structured data validates cleanly in Google Rich Results Test + Schema.org validator
  • Entity consistency: org name, founding date, founders, product names match across schema + copy
  • Key facts in answer-ready formats: definitions near the top, short paragraphs, H2 questions
  • Monthly citation spot-check across Perplexity, ChatGPT, Claude, Gemini
§ 5.707 / 14

Mobile & Responsive

Does the mobile experience match or exceed desktop, since Google indexes mobile-first?

Checks
  • Viewport meta tag present and correct
  • Tap targets ≥ 48×48px with adequate spacing
  • Text legible without zoom (≥ 16px body)
  • No horizontal scroll on any viewport ≥ 320px
  • Mobile serves identical content to desktop: no truncated articles, no hidden sections
  • Forms work on mobile keyboards (inputmode, autocomplete attrs)
  • Mobile CWV within range of desktop
  • Tested on real devices, not just Chrome DevTools
§ 5.808 / 14

Accessibility (WCAG 2.2 AA)

Can users with disabilities accomplish everything sighted/hearing/mouse users can?

Checks
  • Perceivable: alt text, captions, contrast ≥ 4.5:1 body / 3:1 large, text resize to 200% without loss
  • Operable: keyboard navigation, visible focus, skip-to-content link, no keyboard traps
  • Understandable: consistent navigation, clear errors, form labels correctly associated
  • Robust: valid HTML, ARIA used correctly and sparingly, works across assistive tech
  • Automated scan (axe, WAVE, Lighthouse) + manual keyboard and screen-reader testing
Red flags
  • Relying only on automated tools. They catch ~40% of issues. Manual keyboard + screen-reader testing is non-negotiable.
§ 5.909 / 14

Content Audit by Type

Different content types fail differently. Auditing them with the same rubric is malpractice.

Checks
  • 13 content-type rubrics applied per URL (homepage, pillar, cluster, case study, service, product, pricing, blog index, landing, docs, about, contact, legal)
  • Per-type structured data audited
  • Per-type quality signals scored (E-E-A-T, last-updated, author bylines)
  • Inventory of which type each URL is (many sites can’t answer this)
§ 5.1010 / 14

UX & Conversion Audit

At every decision point in the funnel, are we reducing friction and reinforcing intent?

Checks
  • Information scent: can users predict what’s behind each link?
  • CTA clarity: one dominant action per page?
  • Form friction: minimum fields, progressive disclosure?
  • Trust signals placed at the right decision points?
  • Error states clear and recoverable?
  • Empty states useful, not dead ends?
  • Loading states visible and honest?
  • Dark patterns absent (third-party reviewer with ethics brief)?
  • Session replay sample (20–50 sessions, mix desktop/mobile)
  • Heatmaps on top 5 conversion pages
§ 5.1111 / 14

Backlink & Off-Page Audit

Who links to us, and do those links help or hurt us?

Checks
  • Total referring domains + DR/DA distribution
  • Top 20 linking domains (manual quality review)
  • Anchor text distribution (over-optimized anchors are a risk signal)
  • Toxic / spammy backlinks: disavow only if necessary + manual action exists
  • Link velocity, with sudden spikes investigated
  • Brand mentions unlinked, treated as outreach opportunities
  • Competitor backlink gap analysis
§ 5.1212 / 14

Analytics, Event & Observability

Can we see what’s happening, attribute it to a source, and detect when something breaks?

Checks
  • GA4 (or alt) installed on 100% of pages, verified via crawl
  • Event schema documented (20–40 named events, typed properties, no PII)
  • Key conversion events tested via DebugView
  • Attribution model agreed and stored with every lead
  • UTM parameters captured and persisted into CRM + analytics
  • Uptime monitoring with 99.9% SLO minimum
  • Client-side JS error monitoring (Sentry or equivalent)
  • Form submission success rate monitored with daily threshold alert
  • Core Web Vitals tracked over time (RUM or PageSpeed API cron)
  • Privacy / consent compliance: banner, DPA, PII handling
Red flags
  • If a lead form silently fails for 48 hours, how do you find out? If the answer is ‘a salesperson notices the drought,’ fix that before anything else.
§ 5.1313 / 14

Competitive & Content Gap

Where are competitors winning that we’re not, and is it worth catching up?

Checks
  • Top 3–5 competitors identified (business-defined, not tool-defined)
  • Their top-100 organic keywords vs yours (gap report)
  • Their content velocity (posts / month)
  • Their backlink sources you don’t share
  • Their schema coverage vs yours
  • Their AI citation presence (same brand-query spot check)
  • Topic clusters they own that you don’t
  • Their site speed + UX benchmarks vs yours
§ 5.1414 / 14

Log File Analysis

What are crawlers actually doing, not what we hope they’re doing?

Checks
  • Googlebot hit volume per day (trend)
  • Top 20 most-crawled URLs should be the most valuable
  • URLs crawled repeatedly that return errors
  • URLs never crawled that should be
  • Crawl budget wasted on parameter URLs, filters, infinite paginations
  • Bot distribution: Googlebot vs Bingbot vs AI crawlers (GPTBot, ClaudeBot, PerplexityBot)
  • Unusual user agents flagged for investigation
Red flags
  • Most mid-market sites skip this. Every enterprise site must do it.
Want this scored against your site?

Send the URL. We’ll run the framework and come back with a ranked, shippable backlog.

Book your audit →
[DEEPER · RELATED REFERENCES]

Go deeper on the sections that matter most to you.

REFERENCE ↗

13 content-type rubrics

How we audit homepages, pillars, products, landing pages, docs, and 8 more, each against its own rubric.

REFERENCE ↗

AI-readiness deep dive

Schema, SSR parity, bot allowlists, llms.txt, citation spot-checks across Perplexity, ChatGPT, Claude, Gemini.

REFERENCE ↗

5-day process + prerequisites

What access we need, what we capture on day one, and the exact five-day cadence to a scored backlog.