● USE CASE · AI-FIRST

Use case: AI-first website

NLWeb certified, dense entity graph, citations in ChatGPT and Perplexity. Your site speaking directly to AI agents.

The shift

The new channel: AI answers

ChatGPT with search, Perplexity, Gemini, Claude with tools, executive agents… The share of traffic arriving via an AI answer grows every quarter. This is not traditional SEO: these agents don’t look for keywords, they look for entities, verifiable data and executable actions. If your schema is thin, you don’t show up. Period.

AI agents turn your site into an API. If the API is coherent and rich, they invoke you. If it’s noisy or incomplete, they ignore you.

What is an “AI-first” site?

  • Entities with stable, resolvable @id (you can fetch the @id and get JSON-LD back)
  • Dense cross-linking: every Product points to its Brand, Organization, Reviews, Offers
  • Explicit actions: BuyAction, ReserveAction, SubscribeAction with target and deeplinks
  • Temporal context: datePublished, dateModified, validFrom, validThrough
  • Machine context: llms.txt, ai-policy, sitemap separated by type
  • NLWeb-ready: optional conversational endpoint over your own content

NLWeb Readiness Certificate

NLWeb is the emerging proposal for making websites “conversable” by natural-language agents. ValidGraph audits the criteria and, if you pass, issues a certificate valid for 90 days.

  • Quarterly automated verification
  • Embeddable badge for your footer
  • Public JSON at /.well-known/nlweb-readiness.json
  • Auto-renewal as long as you keep passing
  • Version history of the certificate

Dense entity graph

An AI-first site looks like this when you visualize it:

Organization ──┬── publishes ──→ NewsArticle
               │                       │
               ├── owns ──────→ Brand   ├── author ──→ Person
               │                  │    │
               │                  ↓    └── about ──→ Thing
               └── operates → Product ←─ isVariantOf
                                  │
                                  ├── offers ──→ Offer ──→ priceCurrency
                                  ├── aggregateRating
                                  └── potentialAction ──→ BuyAction

Real edges, not isolated entities. The graph is navigable. ValidGraph’s analyzer shows you exactly this graph and detects:

  • Orphans: entities with no incoming or outgoing connection
  • Islands: clusters disconnected from the main Organization node
  • Broken @id: references that don’t resolve
  • Low density: entities with fewer than 3 semantic properties

Compatibility by agent

AgentWhat it consumesWhat it rewards
ChatGPT (with search)JSON-LD + HTMLAuthor as Person, recent datePublished, sameAs
PerplexityJSON-LD + Open GraphCitation density, source diversity, resolvable @id
GeminiCanonical schema.orgComplete Product, Offer with priceCurrency, AggregateRating
Claude (with tools)Web fetch + JSON-LDpotentialAction, llms.txt present
Shopping agentsProduct + Offeravailability, gtin, shippingDetails, hasMerchantReturnPolicy

AI-first roadmap with ValidGraph

  1. Initial audit: site crawl + current AI Discoverability score.
  2. Strong identity: Organization with stable @id, sameAs to official profiles, verifiable logo.
  3. Property density: fill in description, keywords, about, audience for every key Article/Product.
  4. Cross-linking: connect Products to Brand and Organization, Articles to Person + Publisher.
  5. Actions: add potentialAction to transactional entities.
  6. Machine context: publish llms.txt, ai-policy, sitemaps by type.
  7. Continuous monitoring: alerts so a deploy doesn’t break the graph.
  8. NLWeb certification: once you cross the threshold, issue the badge.

Start getting cited by AI