Context Score
How well does your site explain itself?
A site that explains itself
to AI explains itself
to everyone.
Humans, search engines, AI assistants — they all need the same thing. Clear identity, connected content, and structured information. Context Score measures how well your site delivers it.
What it measures
5 categories. 100 points. One scan.
Identity
20 ptsDoes the site explain what it is? Organization schema, about page, meta descriptions, llms.txt, author entity.
Structure
20 ptsIs content findable? Sitemap, breadcrumbs, navigation depth, robots.txt, llms-full.txt.
Depth
20 ptsIs there substance? Answer pages, concepts, glossary, FAQ schema, blog, word count per page.
Connectivity
20 ptsDo pages reference each other? Internal link density, cross-section linking, orphan page detection, related content.
Discoverability
20 ptsCan external systems understand it? Schema diversity, speakable markup, SearchAction, social links, ai-plugin.json.
Sample output
Scan of visualizevalue.com.
Context Score: 91/100 | Grade: A | content
Identity 20/20 ████████████████████
Structure 17/20 █████████████████░░░
Depth 18/20 ██████████████████░░
Connectivity 18/20 ██████████████████░░
Discoverability 18/20 ██████████████████░░
20 pages crawled | 422 in sitemap | 36 avg links/page | 2357 avg words
Top opportunities:
→ Breadcrumbs on 8/20 pages — add to all content pages
→ Build FAQ pages — add content depth
→ 6/20 pages have related content sections — add to all
→ Speakable on 7/20 pages — AI doesn't know what to quoteConnected context
compounds.
Every page that links to another, every term that gets defined, every question that gets answered — makes the whole site more useful.
The Context Playbook
10 steps to build a site that explains itself.
Declare who you are
llms.txt, Organization schema, about page. Tell every system — human or machine — what this site is and who's behind it.
Map your territory
Sitemap, breadcrumbs, robots.txt. Map what exists, show where it lives, control who can read it.
Answer questions directly
/answers pages with the exact question as the H1. First paragraph is the answer. Body is the depth. FAQ schema on everything.
Define your terms
Individual glossary pages and concept pages. Not one long list — individual URLs with individual schemas. Each is a citation surface.
Cross-link everything
Articles link to concepts. Concepts link to courses. Courses link to workflows. The site is a web, not a list of pages.
Make content quotable
Speakable markup tells AI what to cite. Clear first paragraphs that stand alone as answers.
Build for every audience
/for/designers, /for/developers, /for/writers. Audience-specific entry points with tailored problems and recommendations.
Let machines read you
Schema markup on every page type. Diverse types: Organization, Person, Article, DefinedTerm, FAQPage, HowTo, Product.
Prove your work
Reviews, case studies, a changelog. Evidence that the work is real and ongoing.
Stay findable
Meta descriptions that answer questions, not taglines. sameAs links connecting identity across platforms. robots.txt addressing AI crawlers by name.
Run it
One command. No install. No account.
Scan a site
npx github:visualizevalue/context-score https://yoursite.comVerbose mode
npx github:visualizevalue/context-score https://yoursite.com --verbosePage-by-page details, orphan pages, cross-section map.
Requirements
Node.js 18+. Zero dependencies. Crawls up to 20 pages per scan. No API keys needed.
Built by following its own playbook
Every recommendation in the Context Playbook maps to something we actually built on visualizevalue.com. The score isn't theoretical — it's a checklist we followed.
422
pages in sitemap
20
answer pages
25
glossary terms
91
context score
Score your site.
Paste a URL. See what's connected and what's missing.
Get updates
New tools and features. No spam.