Eight specialists, one target, one afternoon
Audit C is not one auditor with more checklists. It is eight specialist lanes running in parallel, each with its own tools: a forensics investigator with server archaeology, a technical lane driving a real crawler, a schema architect reading the JSON-LD graph, a backlink analyst with the full index, a competitor researcher with traffic data, a content quality scorer, an AI-visibility lane testing what machines actually see, and a conversion specialist scoring every call to action. Then a synthesis pass merges all of it into one prioritized register.
- Forensics, technical, schema, backlinks: the infrastructure squad.
- Competitors, content, AI visibility, CRO: the market squad.
- 12 agents, 94 cumulative findings, one synthesis.
Forensics flipped the biggest finding in the file
The forensics lane scored 71/100 confidence, the highest of any lane, because it stopped trusting tools and started testing reality. It requested all 691 broken backlinks like a browser would: every one returned HTTP 200. The host firewall blocks SEO crawlers, so every tool on the market reports a link apocalypse that does not exist. The same lane closed the sister-domain question for good: identical server subnet, identical PHP version, identical lead-form stack, identical 14-city list. Same operator, beyond reasonable doubt. The 301-redirect recommendation went from cautious to execute immediately.
- 691 dead links: all alive. Tool artifact, formally reversed.
- Sister domain: same subnet, same stack, same cities. Case closed.
- Wayback archaeology found the lost /gallery/ page that still earns links.
To AI crawlers, the homepage is an empty room
The AI-visibility lane fetched the site the way GPTBot, ClaudeBot, and PerplexityBot do, and got 127KB of HTML containing 1,471 visible characters: a caching plugin's compatibility script occupies the first 150 words of every page, and the real content never renders without JavaScript. The llms.txt file makes it worse: every URL in it points at the developer's staging server instead of the live site. The business is not losing the AI-answer race. It never made it to the starting line, for reasons that take about 30 minutes to fix.
- Server-side render score: 0/100 on the homepage.
- llms.txt points every URL at a staging server. Five-minute fix.
- Zero named authors, zero cited sources: nothing for an AI to quote.
The attack map: who TurnKey can actually beat
The competitor lane separated the field into giants and prey. The national franchises hold thousands of referring domains and are not the near-term fight. But the local leader holds the market with only 330 referring domains and generic city pages, and the discount player ranks on thin content for two cost keywords worth 1,000 searches a month each that TurnKey does not even address. The seniors and aging-in-place cluster, 5,400 monthly searches, has zero TurnKey presence despite walk-in tubs being a core service. The lane's verdict: a hyper-local New Orleans page plus a cost guide plus an accessibility hub beats the beatable tier with content alone.
- Cost-guide gap: two 1,000-volume keywords held by thin competitor pages.
- Seniors cluster: 5,400 monthly searches, zero TurnKey content.
- Proven local URL pattern: city + service nested pages. TurnKey has none.
Fix the ceiling first: the six-move opening
The synthesis pass ordered every finding by leverage and produced the opening sequence: repoint the llms.txt staging URLs (30 minutes), file the GMB merge to consolidate 41 reviews onto one profile, settle the NAP on one address and one phone everywhere, 301 the sister domain into the main site, build the missing New Orleans page, and rewire the schema identity so every page points at the merged listing. Six moves, mostly administrative, that remove the structural ceiling holding down every ranking. Content and links only start compounding after this.
- Days 1-30: the six structural moves above.
- Days 31-60: topical authority. Cost guide, accessibility hub, real city pages.
- Days 61-90: links to pages that earn money, not press releases to the homepage.