LLMO FAQ

Common questions about implementing the LLMO Framework. If yours is missing, open an issue — answers from the issue tracker get folded back into this page.

Does LLMO replace SEO?

No. LLMO and SEO solve different problems and run in parallel.

• SEO targets ranked search results.
• LLMO targets being cited, summarized, or directly answered by AI systems (ChatGPT, Claude, Gemini, Perplexity).

Most LLMO work — semantic HTML, JSON-LD, sitemaps, robots.txt — also strengthens SEO, so you do not pick one over the other.

How long does it take to implement LLMO?

• 30 minutes: minimum baseline (llms.txt, /ai/ Markdown, AI-allowing robots.txt, JSON-LD on key pages). See the Quickstart.
• A few weeks: reach 13/15 on the LLMO Framework. Most of the time goes into Authority and Citation signals across the site.

What should I implement first?

Start with Retrieval Signals (Component 3):

1. /llms.txt at site root
2. /ai/ Markdown summaries (and per-language if multilingual)
3. robots.txt explicitly allowing GPTBot, ClaudeBot, PerplexityBot, Google-Extended
4. sitemap.xml reachable from the root

Without these, Knowledge Clarity and Authority work cannot be discovered.

Do I really need /llms.txt and an /ai/ directory?

Both are recommended but optional.

• /llms.txt (per llmstxt.org) gives an AI a fast, structured map of the site — particularly valuable when AI agents fetch pages directly rather than via search.
• /ai/ Markdown gives crawlers and copy-paste users clean text without HTML chrome.

They cost very little to maintain and remove ambiguity for AI ingestion.

How do I block AI crawlers I do not want?

Use robots.txt with explicit User-agent directives. To opt out of OpenAI training crawls:

User-agent: GPTBot
Disallow: /

Major crawlers and their opt-out specs are listed in Research → Papers. LLMO is opt-in; blocking is always allowed.

How do I measure whether LLMO is working?

Three layers, tracked monthly:

1. Server logs — look for GPTBot, ClaudeBot, PerplexityBot user agents and which pages they fetch
2. AI prompt audits — query ChatGPT, Claude, Perplexity with industry-relevant questions and check whether your site is cited
3. Referral analytics — visits from chat.openai.com, claude.ai, perplexity.ai, etc.

Is JSON-LD required, or is plain HTML enough?

Plain semantic HTML alone is workable but underperforms. JSON-LD lets you express explicit facts — author, publisher, datePublished, sameAs identities — that AI systems consume without parsing prose. It is the cheapest single move that lifts both Structural Formatting and Authority Signals at once.

Does LLMO matter for B2B sites with low traffic?

Yes — arguably more. B2B traffic from AI search converts at much higher rates than from generic search (Go Fish Digital observed 25× higher conversion). When buyers ask AI assistants for vendors, being the cited source beats being on page 2 of Google.

How does LLMO relate to AEO and GEO?

LLMO is the umbrella.

Standard	Origin	Scope
AEO	Jason Barnard, 2018	Featured snippets, voice answers — no formal framework
GEO	Princeton/IIT Delhi/Adobe, KDD 2024	Generative search, academic
LLMO	This site, 2026	All LLM interactions: search, chat, RAG, agents

See LLMO vs SEO vs AEO vs GEO for the full breakdown.

Can I implement LLMO on a static site (no backend)?

Yes. Static sites are the ideal LLMO target — every component (llms.txt, /ai/ Markdown, JSON-LD, robots.txt, sitemap.xml) is a static file.

The site you are reading is a static Astro site published on GitHub Pages, scoring 15/15 on the LLMO Framework.

How often should I update content for Citation Signals?

AI systems weight freshness.

• Quarterly: primary reference pages (frameworks, guides, papers/research)
• Monthly: fast-moving topics (model releases, AI policy, tooling)
• Always: update both datePublished (creation) and dateModified (revision)

Microsoft’s October 2025 guidelines list freshness as one of three core principles. See Microsoft Guidelines.