Field notes on AI & vertical SEO
How to run real SEO workflows with AI skills — structure, compliance, chaining, and what actually moves rankings in regulated verticals.
Use cases
Skill chaining with Claude CodeFrom empty domain to a ranking medical site
The complete 25-skill MedAuthority pipeline in Claude Code: market research through architecture, compliant content, local SEO, and analytics as one workflow.
Ranking a single-location clinic with the Local SEO Starter
Five MedAuthority skills chained in Claude Code to optimize a profile, answer reviews HIPAA-safe, post, and convert — one afternoon, start to finish.
Proving SEO works: the quarterly retention report
How the Agency Operations bundle turns a quarter of medical SEO into the report that renews a retainer — equity, attribution, audits, decay, and an SOP.
Building a multi-practice-area law firm site with all 26 skills
The full LawGrowth pipeline in Claude Code: jurisdiction-aware architecture, advertising-compliant content, and case-type attribution wired from launch.
A personal-injury firm's first 30 days of local visibility
Six legal-specific skills chained in Claude Code, with the advertising and confidentiality guardrails that make the legal chain genuinely different.
Defending a law firm's retainer with attribution and equity
Case-type-aware lead attribution and durable equity against the LSA treadmill, across ten LawGrowth skills that close on a repeatable SOP.
Field notes
Foundations & conceptsWhy “prompt packs” fail and skill stacks don't
Generic prompts produce generic output. Here's what structure, gating, and chaining change about working with AI.
EEAT for medical sites without the liability
Writing authoritative health content that ranks — while hedging claims, flagging for review, and staying YMYL-safe.
Chaining skills into a full agency deliverable
Market research → sitemap → content → schema → reporting. How outputs become inputs across a stack.
Location pages that rank without thin-content penalties
The line between a useful local page and a doorway page Google punishes — and how to stay on the right side.
HIPAA-safe review responses for healthcare
Why you can't confirm someone is a patient — even when they said so first — and how to respond anyway.
The same skill in Claude, Cursor & ChatGPT
One markdown file, five tools. A practical look at portability and the one gotcha per platform.
Why every skill should return structured output
JSON-first outputs aren't a developer indulgence — they're what lets one skill's result feed cleanly into the next.
Input gating: why a good skill asks before it answers
The single design choice that separates infrastructure from a prompt — and why a refusal to start is a feature.
Measuring SEO equity, not just traffic
Organic dependency, branded demand, and durable value — the metrics that survive when the ad budget stops.
The untracked phone call is killing your attribution
Why most local attribution is wrong, and how to fix the blind spot before you reallocate a single dollar.
Spotting content decay before rankings fall off
Aging pages don't crash — they erode. How to find decay early and decide what to refresh, retire, or leave alone.
Patients search symptoms, not procedures
Mapping problem-first searches to the services that solve them — responsibly, without diagnosing the reader.
FAQ schema and the voice-search opportunity
Valid FAQPage JSON-LD, conversational phrasing, and why the first sentence of every answer matters most.
Clearing the cost objection for elective care
High-ticket, cash-pay decisions hinge on one fear. How to address financing without sounding like a pitch.
Google Business Profile posts that actually do something
Most practices post filler or nothing. What to post, how often, and why fake offers will burn you.
Read your reviews for patterns, not panic
One angry review is noise. Seven mentioning the same wait time is a fixable operational problem.
Turning workflows into SOPs that scale
The difference between an agency that grows and one that can't: documentation a new hire can actually follow.
The human-review gate is not optional
Why regulated-content skills produce drafts, never publishables — and how to build the review step into the workflow.