How the audit works — and what it cannot tell you.

What the scanner does

When you submit a domain, our scanner fetches four things from your public site: the homepage, robots.txt, llms.txt, and sitemap.xml. It follows up to three same-safety redirects, reads at most ~600 KB per resource, and identifies itself as CaliberAuditBot. It runs 24 checks using the published raw weights below and returns a normalized 0–100 score.

The raw weights total 102 points: 25 for crawler access, 15 for structured data, 32 for citable content, and 30 for social and trust signals. The displayed score is earned raw points ÷ 102 × 100, rounded to the nearest whole number. llms.txt is informational and carries zero weight.

The four categories

1 · Crawler access — 25 points

• robots.txt exists and parses (5)
• Key search and answer-system crawlers (GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, and Google-Extended) each allowed to crawl (3 each)
• XML sitemap present (5)
• llms.txt presence is reported as informational only — current evidence shows engines do not use it for discovery, so it carries 0 points

2 · Structured data — 15 raw points

• JSON-LD structured data found (5)
• Organization / LocalBusiness entity schema (4)
• FAQ / Product / Service schema (3)
• Open Graph tags (3)

3 · Citable content — 32 raw points

• Title tag (5) and meta description (4)
• H1 heading (3)
• Visible phone number (5) and address/location signals (5)
• Pricing signals (5) and FAQ-style content (5)

4 · Social & trust — 30 raw points

• At least two linked social profiles (10)
• Google Maps / reviews link (6)
• Visible contact email (4)
• HTTPS (5) and mobile viewport (5)

Scoring and grades

Score	Grade	Reading
85–100	A	Most audited signals are present; verify outcomes separately.
70–84	B	Many audited signals are present, with specific gaps to close.
55–69	C	Mixed signal coverage; several inputs need attention.
40–54	D	Substantial audited inputs are missing or inaccessible.
0–39	F	Most audited inputs are missing or inaccessible.

What this audit cannot tell you

• It does not measure actual citations or recommendations. The scan checks selected public signals — it does not observe whether search or answer systems currently recommend you. Citation observations are recorded separately, by hand, in client measurement logs, marked explicitly as measured or not observed.
• It reads the fetched HTML response. Content added only after client-side JavaScript runs may not be detected by this scanner.
• Pattern-based checks can miss unusual markup. A phone number rendered as an image, or schema injected client-side, may not be detected.
• It scans the homepage plus root files, not your entire site.
• No ranking guarantees. Nothing in this methodology predicts or guarantees placement in any answer or search result.

What happens after the audit

Inside the client workspace, findings run through a tracked loop: each becomes a remediation task with priority, owner, due date, and status history (todo → in progress → done, with blocked states recorded); content moves through idea → drafted → review → approved → published → verified live, and publishing requires the live URL; measurements are logged as explicitly measured, not observed, or not measured yet. Every change is written to a per-client activity history.

Responsible scanning

The scanner only fetches publicly accessible pages, respects a strict size and time budget, and is rate-limited per connection. Run it only on sites you own or are authorized to assess.