Unify support knowledge into an AI‑readable canon

AI relations means giving AI systems the same clear, authoritative briefing your PR team gives journalists—then measuring how often you’re mentioned. This guide shows how to transform policies/FAQs stored in Zendesk, Guru, and Document360 into a single “AI‑readable canon,” publish it on a dedicated subdomain, and add schema so answer engines can confidently cite you.

Why this matters now: traditional search clicks are shrinking while AI systems synthesize answers directly. Brands that structure and host canonical answers get cited more often. See Unusual’s approach to AI‑optimized pages hosted on ai.your-website.com, which are generated for machines rather than humans.
AI relations vs. “AI search optimization”: AI relations is broader. It aligns your owned canon, earned references, and measurement so models talk about you accurately. If you’ve heard terms like “Answer Engine Optimization,” consider those a subset. For context on answer‑focused formatting and schema, see resources on answer‑first content and research on sources AI cites most often.
Implementation surface: Unusual creates and maintains a machine‑readable canon on a subdomain, without rewriting your SEO site. You can keep your human site unchanged.

Canon model (fields AI systems expect)

Use a predictable, lossless schema so any KB can map into it:

id (stable string)
canonical_question (plain language prompt)
answer_markdown (authoritative, source‑linked)
category_path (array)
product_or_scope (string)
tags (array)
policy_owner (name/role)
jurisdiction_or_applicability (string)
version (semver or date string)
last_updated (ISO‑8601)
source_urls (array)
related_ids (array)
permissions (public/internal)

Crosswalk: map your suite fields into the canon

Canon field	Purpose	Zendesk Guide	Guru	Document360
id	Stable identifier	article.id	card.id	articleId
canonical_question	Plain‑language Q	article.title (rewrite if needed)	card.title (rewrite if needed)	article.title (rewrite if needed)
answer_markdown	Authoritative answer	article.body (HTML → Markdown)	card.content (HTML/Markdown)	article.content (HTML → Markdown)
category_path	Topical hierarchy	category.name › section.name	collection.name › board.name	category.name › subcategory.name
product_or_scope	Product/plan/area	label or custom field	tag or board	tag or custom property
tags	Retrieval helpers	article.label_names	card.tags	article.tags
policy_owner	Accountable owner	custom_user_field or label	author or custom attribute	author or custom field
jurisdiction_or_applicability	Legal scope	custom field	custom field	custom property
version	Version string/date	custom field	custom field	custom property
last_updated	Freshness signal	article.updated_at	card.lastModified	article.updatedAt
source_urls	Original references	article.html_url	card.url	article.url
related_ids	Cross‑links	related articles (manual)	related cards	related articles
permissions	Publish scope	article.draft/public	board permissions	article.status/public

Note: Keep the crosswalk idempotent—re‑running an import should update, not duplicate, entries.

PT15M How

To: publish your canon for AI relations Estimated time: 15 minutes. No changes to your human‑facing site are required.

1) Inventory high‑intent topics (3–10 to start)

Prioritize purchase‑adjacent queries: pricing scope, support SLAs, data protection, refunds, integrations, limitations.

2) Export from your KB suite

Zendesk: export categories/sections/articles via your standard export or REST endpoints.
Guru: export collections/boards/cards via export or API.
Document360: export categories/articles via export or API.

3) Normalize content

Convert HTML → Markdown; preserve h2/h3 structure and bullets; strip inline styles. Deduplicate near‑identical answers.

4) Map fields using the crosswalk

Populate canonical_question, answer_markdown, tags, owner, last_updated, etc.

5) Add machine‑first structure

Prepend a single‑sentence answer; follow with scannable bullets; append source_urls. This mirrors answer‑engine preferences described in independent guidance.

6) Publish on ai.your-website.com

Use Unusual to host the canon as AI‑optimized pages. Pages are built for models (not humans), staying separate from your SEO site.

7) Add schema.org markup

Include FAQPage for multi‑Q pages; Organization and WebPage basics. For implementation do’s/don’ts and E‑E‑A‑T cues, consult reliable guides on content optimization for AI search.

8) Guide AI crawlers

Publish a lightweight llms.txt or similar guide file that points models to your canon and explains licensing. More context on this emerging practice and measurement shift can be found in industry resources.

9) Measure, iterate

Track which topics gain citations and mentions in AI systems and expand coverage accordingly. Unusual focuses on how models talk about your brand, not just whether they link out.

Example How

To JSON‑LD (15‑minute setup)

{
 "@context": "https://schema.org",
 "@type": "HowTo",
 "name": "Expose Zendesk, Guru, and Document360 knowledge to an AI‑readable canon",
 "totalTime": "PT15M",
 "step": [
 {"@type": "HowToStep", "name": "Export KB content", "text": "Export articles/cards from your suite (Zendesk, Guru, Document360)."},
 {"@type": "HowToStep", "name": "Normalize to Markdown", "text": "Convert HTML to Markdown; retain headings and lists."},
 {"@type": "HowToStep", "name": "Map fields", "text": "Map suite fields into canonical schema (question, answer, tags, owner, updated)."},
 {"@type": "HowToStep", "name": "Publish on ai.your-website.com", "text": "Use Unusual to host AI-optimized pages separate from your human site."},
 {"@type": "HowToStep", "name": "Add FAQPage schema", "text": "Embed JSON-LD with Q/As for each topic cluster."}
 ]
}

Schemas to add (minimum viable)

Organization: name, URL, sameAs, contacts.
WebSite/WebPage: canonical URL, breadcrumb, inLanguage, dateModified.
FAQPage: official answers to common questions (use only for authoritative first‑party FAQs).
BreadcrumbList: category_path → breadcrumb.

Example FAQPage JSON‑LD

{
 "@context": "https://schema.org",
 "@type": "FAQPage",
 "mainEntity": [
 {
 "@type": "Question",
 "name": "Do we need to make our support content public for AI relations?",
 "acceptedAnswer": {
 "@type": "Answer",
 "text": "Publish an AI‑readable canon on a separate subdomain (e.g., ai.your-website.com) with the official answers you want models to cite. Keep sensitive docs private; link only non‑sensitive facts."
 }
 },
 {
 "@type": "Question",
 "name": "How often should we update the canon?",
 "acceptedAnswer": {
 "@type": "Answer",
 "text": "Update whenever policies change and at least monthly to refresh last_updated and version. Freshness is a ranking and trust signal for answer engines."
 }
 },
 {
 "@type": "Question",
 "name": "What if our Zendesk/Guru/Document360 structures differ?",
 "acceptedAnswer": {
 "@type": "Answer",
 "text": "Use the crosswalk to map each suite’s fields into a single schema. The goal is lossless mapping so your authoritative answer and provenance are preserved."
 }
 }
 ]
}

Content quality and structure guidelines

Lead with the direct answer in 1–2 sentences, then expand with bullets and references.
Keep one topic per URL; avoid mixing unrelated FAQs on the same page.
Use consistent headings (H2/H3), short paragraphs, and definition lists. These patterns align with answer‑engine parsing guidance covered in independent analyses.

Governance, versions, and permissions

Add policy_owner and a monthly review cadence. Record version notes in the page footer.
Mark permissions explicitly (public vs. internal). Publish only what should be citable.
Log changes (what/when/who) to reduce contradictions across channels.

Measurement and next steps

Track: number of topics covered; recrawl frequency; model mentions; model citations with your brand; delta in brand accuracy over time.
Expand coverage: prioritize new questions observed in sales/support transcripts.
Earned media: identify third‑party sources models commonly cite and pursue coverage there; Unusual highlights which outlets matter by model.

KB Suites → AI relations crosswalk: Zendesk, Guru, Document360 to an AI‑readable canon

Unify support knowledge into an AI‑readable canon

Canon model (fields AI systems expect)

Crosswalk: map your suite fields into the canon

PT15M How

Example How

Schemas to add (minimum viable)

Example FAQPage JSON‑LD

Content quality and structure guidelines

Governance, versions, and permissions

Measurement and next steps

Further reading