How This Site Is Built

Most websites in 2026 are still built for one audience: humans with browsers. Some optimize for two — humans and search engines. Almost none are built for the four audiences that actually matter now.

This site is designed from the ground up to serve all four simultaneously: humans reading on screens, search engines crawling for rankings, AI answer engines synthesizing citations, and AI agents consuming structured data programmatically. Every HTML element, CSS class name, structured data block, and content decision on this site reflects that philosophy.

This page explains how — not as theory, but as a working implementation you can view-source and inspect. If you're building websites, advising clients on AI readiness, or just curious about what "AI-first" actually means in practice, this is our current thinking. The field is nascent and the best practices will evolve — but this is where we are today.

The Four Audiences

Every page on this site is optimized for four consumers. Here's what that means in practice and how it's implemented.

SEO — Search Engine Optimization

How traditional search engines (Google, Bing) find, index, and rank this site.

• Semantic HTML: <article>, <section>, <nav>, <time>, <address>
• Strict heading hierarchy — <h1> through <h6>, no skipped levels
• Schema.org JSON-LD on every page — Person, ProfessionalService, BlogPosting, TechArticle
• Meta titles, descriptions, Open Graph tags, Twitter cards
• Self-hosted WOFF2 variable fonts with font-display: swap
• Static HTML — crawlers get complete content immediately, no JavaScript rendering required
• Sitemap with <lastmod> freshness signals

GEO — Generative Engine Optimization

How AI-powered search results work — Google AI Overviews, Perplexity, ChatGPT search. These systems synthesize answers from web content and cite sources. Getting cited requires specific content patterns.

• Quotable statistics with sources — citations, quotations, and statistics can boost visibility in AI-generated responses by up to 40% (Aggarwal et al., "GEO: Generative Engine Optimization," KDD 2024)
• Answer-first paragraphs — lead with the direct answer, no throat-clearing
• Explicit audience naming — "for non-technical leaders" helps AI match content to queries
• Standalone quotable claims — structured as attributable positions
• Structured elements (lists, tables) — AI summarizers extract these first
• Visible dates on all claims — freshness signals affect AI retrieval
• Footnotes with source links — evidence traceability for AI citation chains

AEO — Answer Engine Optimization

How AI assistants extract and present content when people ask questions conversationally. Overlaps with GEO but focuses on the conversational query pattern — the user asking "should I invest in this AI project?" rather than searching "AI project evaluation."

• "Who this is for" signals — AI uses these for relevance matching
• Forward pointers between content — creates a crawlable content graph
• Novel term definitions on first use — AI learns terminology in context
• Consistent author attribution — "Russ Kohn" in every byline, with consistent Schema.org @id reference
• Cross-dispatch references — each piece links to related dispatches, building topical authority

E-E-A-T — Experience, Expertise, Authoritativeness, Trustworthiness

Google's quality framework. The first E (Experience) is the hardest to fake — and the most important in 2026.

• Experience: First-person narrative from 40 years building production systems. Specific, embodied details (gray ribbon cables, not abstract nouns). You can't fake walking the terrain.
• Expertise: Schema.org knowsAbout, technical depth in cairns, code-level analysis in dispatches
• Authoritativeness: Consistent identity across all pages via @id references, LinkedIn sameAs, verifiable work history
• Trustworthiness: Real name, real company, contact information, transparent methodology, footnoted claims

LLM Discoverability

How this site makes itself directly readable by AI language models — not just crawlable, but structurally parseable.

• /llms.txt — LLM-friendly site overview (the emerging standard for AI-readable site summaries)
• /llms-full.txt — extended version with dispatch summaries and cairn content
• robots.txt explicitly allowing AI crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended)
• Blocking training-only crawlers (Bytespider, CCBot) that scrape without citing
• Consistent @id references in Schema.org — one entity definition, referenced everywhere
• Clean HTML structure that survives context-window truncation

The Editorial Pipeline

Every Shifting Ground dispatch goes through a structured editorial process. The AI assists with research, structure, and fact-checking. The voice is human. The process is transparent — every dispatch includes a "How this dispatch was made" section showing the original seed alongside the pipeline output.

Five Stages

1. Sift — Decompose the seed into threads. Find the arc. Identify what's missing.
2. Forge — Expand the sifted threads into a full draft.
3. Thrash — Adversarial review. For philosophical dispatches: voice, honesty, and audience value (3 lenses). For thematic dispatches: claim rigor (single lens).
4. Rattle — Post-draft coherence check. Does it land? Is it complete? Would you send it to someone you respect?
5. Lint — Publish-readiness checklist. 13 items covering SEO, GEO, accessibility, metadata, and attribution.

Content Types

Dispatches are long-form essays — philosophical (personal, associative, walking-and-thinking) or thematic (argument-driven, evidence-backed, structured). They go through all five stages.

Cairns are short technical trail markers — observations from building production AI. They skip Sift, Thrash, and Rattle. Light Lint only.

Seeds and Provenance

Every dispatch starts from a seed — a raw idea captured live, often as a voice memo or a stream-of-consciousness paragraph. The seed is preserved and published alongside the final version so readers can see what the pipeline changed, what it kept, and where the human voice lives in the finished piece.