When ChatGPT, Claude, Gemini, or Perplexity look at your website, they read the raw HTML your server returns — before any JavaScript runs. SearchAudit fetches exactly what those AI crawlers receive, then scores it from 0 to 100 across six weighted categories. The result is a single, explainable number: how readable, understandable, and citable your site is to AI search engines.
Crawlability (25%)
The heaviest-weighted category. We measure how much real, server-rendered text is present in the initial HTML response. Sites that render entirely in the browser return an empty shell to crawlers — no headings, no product copy, nothing to cite. Server-side rendering or static generation is the single biggest lever on your score.
Schema Markup (15%)
Structured data (JSON-LD) tells AI engines what your page represents — an organization, a product, a service — and the identity signals (name, description, logo, and sameAs authority links) they use to confidently attribute and cite your brand. Pages with a recognized entity type and complete identity properties are far easier for models to reference.
Semantic HTML (15%)
Clean document structure — a single <h1>, an unbroken heading hierarchy, landmark regions like <main> and <article>, descriptive image alt text, and a healthy ratio of readable text to markup — helps models parse what your page is actually about.
AI-Specific Files (15%)
Newer conventions AI crawlers increasingly look for: an llms.txt plain-text summary, a valid sitemap.xml, a robots.txt that explicitly welcomes GPTBot, ClaudeBot, and PerplexityBot rather than silently blocking them, and an ai-agents.txt policy file.
Meta Tags (15%)
The fundamentals that still matter: a descriptive title and meta description, a canonical URL to consolidate authority, and OpenGraph and Twitter card data so AI summaries and link previews render with real context instead of a bare URL.
Vibe-Code Issues (15%)
A dedicated check for the failure patterns common to sites built with AI app builders — client-only render shells, missing structured data, and default configurations that leave content invisible to AI crawlers. Fixing the rendering layer resolves most of them at once.
Every category rolls up into your 0–100 score with a full breakdown of where points were lost and the specific, copy-pasteable fixes to recover them. Run a free scan to see yours in under 60 seconds — no signup required.