The checks, vertical-tailored
What the audit checks for AI agencies.
Every audit, on every plan, scores Foundational SEO, Technical SEO and AI Visibility. Local SEO is scored when the pages we read show a physical location or a service area, such as a street address, a city and state or a Maps link. The check definitions are constant; the severity, the recommended fix, the competitor set, and the example code shift per vertical via the adapter.
Foundational pillar
Foundational
Keyword universe, content depth, on-page structure, internal link graph.
- h1-presence
Service landing pages should lead with a single, specific H1 ("Anthropic-certified RAG implementation for B2B SaaS") not a generic "We do AI". Specificity is the credibility signal a technical buyer reads first.
- passage-extractability
Every methodology post should carry 134-167 word "What is [technique]?", "When should you use [technique] over [alternative]?", "How do you evaluate [technique]?" answer blocks co-located under question-shaped H2 so ChatGPT and Perplexity can lift the passage verbatim.
- outbound-link-quality
Cite the arXiv paper that introduced the technique you describe, the schema.org spec for any structured-output claim, the Anthropic / OpenAI / Vertex docs for any model claim. Outbound authority to primary sources is the highest-leverage trust signal for the AI-savvy buyer.
Technical pillar
Technical
Crawl, render, schema validation, Core Web Vitals, render-parity diff.
- js-render-parity
If your agency site is a Next.js app routed with the App Router and your service pages live under client components, the raw HTML is empty. AI crawlers see nothing. The audit flags it (severity HIGH for ai_agencies) and recommends moving the page to a Server Component or static export with the content in the initial HTML.
- schema-validation
Organization + Service + Person + Article JSON-LD must validate against schema.org. For Service, populate serviceType (specific: "Retrieval-Augmented Generation", "Model Fine-Tuning", "Agent Orchestration"), provider (Organization), areaServed (Worldwide is fine for remote-first agencies), and offers if you publish pricing.
- core-web-vitals
Agency case-study pages are content-heavy and image-heavy (architecture diagrams, code snippets, screenshots). LCP below 2.0s requires image optimization, font-display swap, and (ideally) static export with edge caching. The audit measures lab + field data where CrUX coverage exists.
Local pillar
Local
On-site local signals: LocalBusiness schema, Maps link, phone and address, service-area pages.
- local-business-schema
Remote-first AI agencies do not need LocalBusiness schema. The audit marks local-pillar checks as N/A for ai_agencies (configured in the adapter's severity overrides), so absence carries no penalty. The pillar shows as "N/A" rather than 100/100 to avoid implying a passed check that does not apply.
- nap-consistency
N/A for fully remote agencies. If your site lists a HQ city, the audit reads the page for a complete phone, street address, and ZIP code. It does not compare your LinkedIn, Crunchbase, or partner-directory listings.
AI Visibility pillar
AI Visibility
AI crawler readability, schema for AI, passage extractability, entity graph.
- ai-crawler-readability
Render-parity diff per page. Raw HTML token count vs rendered DOM token count. AI engines (GPTBot, ClaudeBot, PerplexityBot) fetch raw HTML; any page where AI sees less than 70 percent of what humans see is flagged HIGH severity for ai_agencies and a Server Components / static-export fix is recommended per page.
- passage-extractability
A 134-167 word passage co-located under a question-shaped H2 is the format ChatGPT and Perplexity lift verbatim. For an AI agency, structuring case studies and methodology posts in this shape roughly doubles citation rate on technical-buyer queries.
- entity-graph
Organization sameAs to your Anthropic partner directory listing, OpenAI partner directory listing, Vertex AI partner directory listing, GitHub organization page, Hugging Face organization page, LinkedIn company page, and Crunchbase. AI engines triangulate AI-agency authority through this graph; partial graphs measurably suppress citation rate.