GEO in Practice: Making Your Website Discoverable by AI Assistants (2026 Guide)

Bottom line first

As AI assistants become the first stop for a growing share of information queries, a website that AI engines cannot crawl and understand is invisible in those conversations — when users ask AI, you are simply not in the answer.

GEO (Generative Engine Optimization) is the systematic fix. This guide shares the 7-layer setup we shipped for the AI Enable Harness website, all verified in production.

1. JSON-LD structured data

AI engines rely on structured data to identify who you are, what the site is, and what each page covers. We emit three schema types on every page:

<script type="application/ld+json">
{ "@type": "Organization", "name": "...", "url": "...", "logo": "...", "email": "..." }
</script>
<script type="application/ld+json">
{ "@type": "WebSite", "name": "...", "inLanguage": "en" }
</script>
<script type="application/ld+json">
{ "@type": "WebPage", "name": "...", "description": "..." }
</script>

Keep the wording consistent across schema, page copy and llms.txt — inconsistent entity descriptions lower an AI engine’s confidence when citing you.

2. llms.txt: a site guide written for AI

A plain-text manifest at your site root that tells AI crawlers what the site is, where the core pages are, and how to attribute citations. Low effort, high leverage. We wrote a dedicated walkthrough: How to write llms.txt.

3. robots.txt: whitelist the AI crawlers

Blocking AI crawlers means forbidding AI citation. There are 13 user agents worth explicitly allowing — retrieval bots (OAI-SearchBot, PerplexityBot, Claude-SearchBot), session bots (ChatGPT-User, Claude-User) and training bots (GPTBot, Google-Extended, CCBot) each behave differently. Full list and a copy-paste config: The AI crawler whitelist.

4. FAQPage schema

FAQ blocks are the highest-density citation source for AI engines — a clean question-answer pair is exactly the shape an AI answer wants to quote. Pair the visible FAQ section with FAQPage JSON-LD (every post on this blog does both, including this one).

5. Sitemap

Auto-generate it, exclude error pages, and reference it from robots.txt. AI crawlers use it the same way search spiders do: discovery and freshness.

6. Canonical + Open Graph

Canonical tags prevent duplicate-content dilution across URL variants; OG tags control how shared links unfurl. Both feed entity consistency, which AI engines weigh when deciding what to cite.

7. BLUF content structure + semantic HTML

BLUF — bottom line up front. Every page and article should state its conclusion in the first screen, then argue it. Combine with semantic HTML (h1/h2, section, article, details) so extraction doesn’t depend on visual layout. LLM-based crawlers parse token-efficiently: the earlier and cleaner your key claims, the likelier they survive into the citation.

Rollout checklist

1. JSON-LD: Organization + WebSite + WebPage sitewide, FAQPage where relevant
2. llms.txt at site root
3. robots.txt whitelisting 13 AI crawlers
4. Sitemap generated and referenced
5. Canonical + OG on every page
6. BLUF restructure for key pages
7. After 1-4 weeks: query your brand and target long-tail terms in each AI assistant, and verify citations

Related reading

• How to write llms.txt: the AI crawler site guide
• The AI crawler whitelist: 13 user agents your robots.txt should allow