LLM-LD SOP

Appearance

Troubleshooting Fresh ​

Full Conformance Checklist ​

Level 1: Crawl-Ready ​

  • [ ] Schema.org JSON-LD in <head> of homepage
  • [ ] Schema.org JSON-LD in <head> of key pages (about, services, contact)
  • [ ] Correct @type for business (LocalBusiness, Organization, etc.)
  • [ ] potentialAction in Schema.org pointing to /ai-discovery
  • [ ] AI Discovery Page live at /ai-discovery
  • [ ] ADP contains llmld:DiscoveryPage JSON-LD
  • [ ] ADP links to llmdisco.com
  • [ ] ADP links to llmdisco.com/sitemap.xml
  • [ ] Footer link on main site: 🤖 AI-Optimized Website pointing to /ai-discovery
  • [ ] robots.txt blocks AI crawlers on main site (with Allow: /ai-discovery)
  • [ ] robots.txt allows AI crawlers on AI subdomain
  • [ ] Site submitted to llmdisco.com

Level 2: Ingest-Ready ​

  • [ ] All Level 1 items
  • [ ] llm-index.json accessible at /.well-known/llm-index.json
  • [ ] llm-index.json has valid @context array
  • [ ] llmld:meta with version and generated fields
  • [ ] llmld:site with name, type, description, and domains.primary
  • [ ] llmld:primaryEntity with @id, @type, and name
  • [ ] llmld:summary with one_liner, paragraph, and key_facts
  • [ ] llmld:pages with at least the homepage
  • [ ] entities.json accessible
  • [ ] knowledge-graph.json accessible

Level 3: Agent-Ready ​

  • [ ] All Level 2 items
  • [ ] llmld:actions.primary with at least one action
  • [ ] Each action has valid id, name, url, and type
  • [ ] Action endpoints are accessible without JavaScript
  • [ ] Action type matches actual function (schedule, purchase, signup, etc.)

Common Issues ​

Verification Fails: "No AI Discovery Page Found" ​

Symptom: Network verification can't find your ADP.

Checks:

  1. Is the page live at exactly /ai-discovery? (not /ai-discovery/ or /ai_discovery)
  2. Does the page return HTTP 200?
  3. Is /ai-discovery allowed in your main site's robots.txt?

Fix: Add Allow: /ai-discovery explicitly for every AI crawler user-agent in your main site's robots.txt.

Symptom: Network can't find the 🤖 AI-Optimized Website footer link.

Checks:

  1. Is the exact text 🤖 AI-Optimized Website present? (no variations)
  2. Does the link point to /ai-discovery?
  3. Is the link in the rendered HTML (not hidden by JavaScript)?

Fix: Ensure the footer HTML contains exactly:

html
<a href="/ai-discovery">🤖 AI-Optimized Website</a>

JSON Validation Error: "Invalid @context" ​

Symptom: Validator reports invalid @context.

Fix: The @context must be an array with both namespaces:

json
"@context": ["https://schema.org", "https://llmld.org/v1"]

Single string "@context": "https://schema.org" is not valid for LLM-LD.

JSON Validation Error: "Wrong @type" ​

Symptom: Validator reports wrong @type.

Fix: llm-index.json must use:

json
"@type": "llmld:AIWebsite"

ADP structured data must use:

json
"@type": "llmld:DiscoveryPage"

AI Subdomain Returns 404 ​

Symptom: ai.yoursite.com is not resolving.

Checks:

  1. Is there a DNS record for ai.yoursite.com?
  2. Has DNS propagated (can take up to 48 hours)?
  3. Is the CNAME or A record pointing to the right host?

AI Crawlers Still Hitting Main Site ​

Symptom: Server logs show GPTBot, ClaudeBot on main site pages.

Fix: Each AI crawler user-agent needs its own block directive. A wildcard User-agent: * rule does NOT automatically apply to named crawlers that have their own rules. Add explicit blocks for each crawler.

llms.txt Not Readable by AI ​

Symptom: AI systems can't fetch your llms.txt.

Checks:

  1. Is Content-Type: text/plain set?
  2. Is Access-Control-Allow-Origin: * set?
  3. Is the file on the AI subdomain, not the main site?
  4. Does the file end with EOF on its own line?

Schema.org Validator Shows Errors ​

Symptom: Google's Schema Validator reports errors.

Common fixes:

  • telephone must use the full number including country code: +1-555-123-4567
  • addressCountry must use 2-letter ISO code: "US", not "United States"
  • openingHoursSpecification.opens and .closes must be in HH:MM format: "09:00", not "9am"
  • potentialAction.target must be a full URL: "https://yoursite.com/ai-discovery"

Actions Not Working for AI Agents ​

Symptom: AI agents can't complete booking/purchase actions.

Requirements for agent-accessible endpoints:

  1. The action URL must return an accessible page/form
  2. The page/form must work without JavaScript for basic functionality
  3. The type field must match the actual action type
  4. For booking: scheduling page must be accessible without login

Verification Process FAQ ​

How do I check my conformance level? Submit your URL to llmdisco.com. The automated verification will report your conformance level.

What if verification fails? The verification report shows exactly what's missing or incorrect. Fix the issues and resubmit -- no limits on attempts, no waiting period.

How often does the network re-verify? The network periodically re-crawls member sites. You can also trigger manual re-verification from your listing page.

My listing shows Level 1 but I've implemented Level 2. What's wrong? Request a re-verification. The network assigned your conformance level at the time of submission; adding more features doesn't automatically upgrade your level.

Why can't I find my site in AI answers? LLM-LD makes your site AI-readable and discoverable, but does not guarantee AI systems will recommend you. Factors include: quality and specificity of your structured data, the AI system's training data and search behavior, competitive landscape in your category.

Data Quality Checklist ​

Poor data quality reduces the value of LLM-LD even when the technical implementation is correct.

  • [ ] one_liner is specific and differentiated (not "We help businesses succeed")
  • [ ] key_facts contain concrete information (numbers, certifications, specific offerings)
  • [ ] services list is complete and specific
  • [ ] service_area is accurate for local businesses
  • [ ] Contact information matches what's on your main site
  • [ ] Business hours are accurate and include special notes (holidays, extended hours)
  • [ ] Action URLs actually work and lead to functional pages
  • [ ] llm-index.json data matches Schema.org data (no contradictions)
  • [ ] No placeholder content (no "YOUR BUSINESS NAME HERE" left in files)
  • [ ] Files are kept current -- set a calendar reminder to review monthly

LLM-LD is licensed under CC BY 4.0. Created by CAPXEL.