Actavero

Methodology · Website

How the AI readiness score is calculated

Method version ai-visibility-score-v2 · Last updated

In short

The AI readiness score (0–100) measures how well AI assistants such as ChatGPT, Claude and Perplexity can reach and use a website. It combines whether AI bots are allowed in (60%), how quotable the content of key pages is (30%, assessed by Claude on six criteria) and whether the site publishes a valid llms.txt (10%).

Formula

AI readiness = 0.6 × access + 0.3 × content + 0.1 × llms.txt
Without a content assessment: 0.9 × access + 0.1 × llms.txt
Access = Σ (purpose weight × access) ÷ Σ purpose weight; search and user bots weigh 3, training bots 1; allowed = 1, partial or unknown = 0.5, blocked = 0
llms.txt: valid 100, present but invalid 50, missing 0

How it is calculated, step by step

  1. For each known AI bot (e.g. OAI-SearchBot, ChatGPT-User, Claude-SearchBot, PerplexityBot, GPTBot), check robots.txt and test whether the site's bot protection actually lets that user agent through.
  2. Weight the result per purpose: bots that power AI search and live answers count three times as much as training crawlers.
  3. Fetch the homepage and key pages and let Claude score six content criteria from 0 to 100: clear entity statements, direct answers, heading structure, FAQ-able content, author and date, and factual density.
  4. Check /llms.txt for presence and validity.
  5. Combine the three parts with the weights above.

What goes into it

Bot access
Per AI bot: allowed, partially allowed, blocked by robots.txt or blocked at the edge (CDN/WAF).
Content readiness
The mean of six criteria scored by Claude on the key pages.
llms.txt
A markdown file listing the site's key pages for AI assistants.

Data sources

  • Actavero's crawler with the user agents of known AI bots
  • Claude (Anthropic) for the content assessment
  • The site's robots.txt and llms.txt

How often it is updated

Monthly on the 1st, and on demand within the plan's re-tests.

How to read it

  • Low access means AI assistants cannot read or cite the site, whatever its content.
  • Blocking training crawlers is a legitimate choice and costs relatively few points; blocking search and user bots costs the most.

Limitations

  • Some sites verify bot IP addresses; Actavero detects this and reports it rather than guessing.
  • Content scores from a language model can vary by a few points between assessments; unchanged pages reuse their earlier assessment.

Frequently asked questions

Is this the same as GEO?

No. AI readiness is about whether assistants can reach and use your site. GEO measures whether they actually recommend you in their answers.

Why do search bots weigh more than training bots?

Search and user bots decide whether your pages can be cited in answers today; training crawlers only influence what future models know.

  • GEO share of voice (AI answers) — Your share of brand mentions in AI answers to a fixed set of buyer questions, asked monthly to Claude with web search; a citation of your own site counts extra.
  • Site health score — 45 technical and content checks on a crawl of up to 200 pages, grouped into six categories; each category loses points by how often its checks fail, weighted by severity.

Feedback on this explanation

Loading…