Every scan returns an Agent Readiness Score from 0 to 100, built from individual checks grouped into five readiness categories. This page explains the categories and the checks behind them, how profile-aware weighting shapes the score, what each check status and severity means, and the evidence and remediation every result carries — plus the limits of what an MVP scan can and can't tell you.
Every check belongs to one of five categories. Together they describe how an AI crawler or agent discovers your site, reads its content, respects your access rules, finds the protocols it needs to act, and understands your products. The checks listed below are the exact set the scanner runs.
Can agents find your site's entry points — robots.txt, sitemaps, canonical URLs, and machine-readable discovery links?
Can agents read your content without running JavaScript — through server-rendered HTML, Markdown content negotiation, and llms.txt?
Do you give AI crawlers clear, explicit rules — which bots may access what, content-usage signals, and crawl-rate hints?
Can agents discover the protocols and APIs they need to act — MCP, OAuth metadata, API catalogs, and Agent Skills?
Can AI shopping assistants read your products, prices, and feeds, and find agentic-commerce signals?
The overall score is a weighted roll-up of the five categories. The scan profile you choose changes how heavily each category counts, so the result reflects what matters for your kind of site. A category weighted at 0% for a profile does not run and never penalizes the score.
| Category | All checks | Content site | API/Application | E-commerce |
|---|---|---|---|---|
| Discoverability | 25% | 35% | 20% | 20% |
| Content accessibility | 25% | 40% | 20% | 20% |
| Bot governance | 20% | 25% | 20% | 20% |
| Protocols | 20% | 0% | 40% | 15% |
| Commerce | 10% | 0% | 0% | 25% |
These category weights match the scoring engine. Within each category, individual checks are weighted further by severity (see below).
The 0–100 score also maps to a readiness level from 0 to 5 — a quick label for how far along a site is.
| Level | Score | Interpretation |
|---|---|---|
| 0 | 0–19 | The site is almost not ready for AI agents. |
| 1 | 20–39 | Basic discoverability. |
| 2 | 40–59 | Content is partially available to agents. |
| 3 | 60–74 | Good agent readiness for content use cases. |
| 4 | 75–89 | Ready for API and authenticated interactions. |
| 5 | 90–100 | An advanced, agent-ready site. |
Every check resolves to one of five statuses:
The requirement is satisfied.
Partial support, or an ambiguous result that only partly meets the requirement.
The requirement is not satisfied.
Not applicable to the selected profile, or excluded from a custom check scope. It carries zero possible score and never penalizes the result.
The check could not complete — usually a timeout, network error, or blocked response. A single failed check never aborts the scan.
Severity reflects how much a failing check matters. Within a category, a check's severity sets its weight, so higher-severity failures move the score more.
| Severity | Check weight | Example |
|---|---|---|
| Critical | ×3 | Site unavailable, scanner blocked, or a private-IP redirect. |
| High | ×2 | Missing robots.txt, sitemap, or Markdown/LLM representation. |
| Medium | ×1 | Missing content-usage policy, Agent Skills, or API Catalog. |
| Low | ×0.5 | Missing optional metadata or weak descriptions. |
Every result is evidence-based. Each check stores only what it needs to justify its status: HTTP status codes, selected response headers, the public files and URLs it discovered, and short content snippets. Full HTML is not stored by default, and the scanner only ever reads public pages.
For each failed or partial check, the report adds plain-language remediation guidance and a ready-to-send task for your developer — what to change, the expected result, and a copy-paste prompt for coding assistants. Reports and prompts are available in Russian, Ukrainian, and English.
The MVP scanner is deliberately narrow and safe:
The score reflects public signals at scan time, not a guarantee about future AI traffic. For the full limits, see the disclaimer.