GEO Decision System™
The Method, Applied · Real Sites

The nine layers, pointed at live businesses.

Proof isn't only my own scorecard. This is the GEO Decision System™ run end to end on real home-services sites — a flagship five-phase demonstration audit, plus three field audits across the cleaning vertical. Every business is anonymized; the method is the point.

These are self-directed audits. Several of the sites never engaged me — I audited them to pressure-test the system end to end and to map how AI answer engines actually treat this vertical. That is exactly why every business here is anonymized by sector and market: the findings are real, the logos are withheld.

And one thing stays deliberately out of view — the remediation. You'll see the diagnosis and the headline numbers; the fix roadmap, schema package, and sequencing are the work itself. This page shows how I think, not the answer key.

◆ — Flagship Audit

A five-phase audit of a live home-services site.

A live North American home-services website, audited cold across all six AI answer engines. Chosen as the flagship because it exercises every layer of the system — from entity resolution to a production-ready action brief.

102
Pages inventoried & classified
33.8
Avg citation-probability / 100 · 10 priority queries
14
Schema types missing from the stack
6
AI engines cold-tested
Phase 1
Entity & cold baselineResolve the business as an entity, then cold-test where it is cited today across ChatGPT, Perplexity, Google AI Overviews, Gemini, Claude, and Copilot.
Phase 2
Content inventoryFull page-level crawl and classification of all 102 pages by intent and citation role — what each page is actually for, in the eyes of an answer engine.
Phase 3
Citation-probability scoringScore ten priority queries for how likely the site is to be cited. The portfolio averaged 33.8 out of 100 — the gap, quantified.
Phase 4
Schema-stack deltaMap structured data present versus required. Fourteen schema types were missing; the delta was written up as ready-to-paste JSON-LD.
Phase 5
Execution Action BriefA prioritized roadmap sequenced by citation impact and effort — the bridge from diagnosis to a build a team can execute.
What it proved
A site can be well-built for humans and still score in the low thirties for AI citation. The gap wasn't content quality — it was entity clarity and structured data. That is the pattern the whole system is designed to close.
▖ — Field Audits

Three anonymized audits from the cleaning vertical.

Representative patterns from the field, across three markets. Businesses anonymized by sector and region. Findings shown; fixes withheld.

US · Residential Cleaning

Invisible beyond the brand name

L0 · L6 · L7

Recognized when named. Absent for the category.

  • Inconsistent name / address / service-area signals across site and profiles
  • No Organization schema anchoring the entity
  • Service area never stated as machine-readable data

Asked cold for cleaners in its own city, engines returned directories — never the business.

AU · Bond / End-of-Lease

Strong content the machines can't read

L1–L4 · L6

Depth on the page, nothing structured underneath.

  • Detailed service and inclusion lists locked in plain HTML
  • No Service or FAQPage schema on any offer
  • Pricing and guarantees unparseable as structured data

Engines summarized aggregator listings instead of the site's own richer content.

UK · Domestic Cleaning

No authority for the answer layer

L0 · L5 · L7

Template pages, no citable proof.

  • Thin, near-duplicate location pages with no unique substance
  • No review or trust signals surfaced to engines
  • Zero third-party corroboration of the entity

Every relevant answer defaulted to trade directories with stronger citation footprints.

Every audit runs the same nine layers, L0 → L8 →

Want this run on your brand?

The same nine layers, the same cold six-engine test — pointed at your business.

Start with a Snapshot →