How we measure AI visibility

We score a website's AI visibility from 0 to 100 across four axes. The formula is open. Peer review, corrections, and suggestions are welcome — email methodology@shortcutsistem.com.

Version 1.2 · Last updated · Next review: 2026-10-17

The 4 axes

1. Crawler access (40 points)

For each of 7 known AI crawlers, we check robots.txt. Allow = points, Disallow = 0.

  • GPTBot — 7 points
  • ChatGPT-User — 6 points
  • ClaudeBot — 7 points
  • Claude-User — 6 points
  • Google-Extended — 4 points
  • CCBot — 4 points

2. Structured data (25 points)

We parse schema.org JSON-LD on the homepage and up to 3 product/inner pages.

  • Organization schema present — 5 points
  • sameAs links to ≥2 third-party profiles (LinkedIn, Wikidata, etc.) — 5 points
  • Product / Article / FAQPage schema on relevant pages — 5 points
  • BreadcrumbList — 3 points
  • Valid JSON-LD (passes Schema.org validator) — 4 points
  • llms.txt file present at root — 3 points

3. Entity consistency (20 points)

LLMs trust entities that look the same across the open web. We check:

  • Consistent NAP (name, address, phone) across website + 2 third-party sites — 7 points
  • Wikipedia or Wikidata entry (if eligible) — 5 points
  • About page with author/reviewer names and roles — 4 points
  • Public contact email with matching domain — 4 points

4. Freshness signals (15 points)

LLMs deprioritize stale content. We check:

  • Sitemap.xml present and updated within 30 days — 5 points
  • ≥1 content update within 90 days — 5 points
  • Visible "last updated" timestamps on key pages — 3 points
  • GitHub-style changelog or release notes — 2 points

What this does NOT measure

  • How often a site is actually cited by ChatGPT/Claude. We sample, not enumerate.
  • Content quality or factual accuracy. That's a different scoring system.
  • Page speed or Core Web Vitals. (Crawl budget matters less for AI than for Google.)
  • Backlinks. Backlinks help SEO; they help AI less than primary data does.

Peer review

Our methodology is reviewed quarterly by:

  • [Pending: independent AI/ML researcher — name to be added on confirmation]
  • [Pending: Indonesian university partner — name to be added on confirmation]

Disagree with the formula? Email us. Public corrections will be credited in the next quarterly update.

Version history

  • v1.2 (2026-07-17) — Added llms.txt to structured-data axis.
  • v1.1 (2026-04-12) — Adjusted entity-consistency weights after Q1 2026 calibration.
  • v1.0 (2026-01-30) — Initial release.