Large Language Model Optimisation

Large language model optimisation (LLMO), also known as large language model optimization in US English, is the practice of structuring website content so that large language models — such as GPT-4, Gemini, and Claude — can discover, interpret, and cite it when generating responses to user queries.

What is LLMO?

LLMO emerged as a term focused specifically on optimising for large language models as the underlying technology, rather than for "answer engines" (AEO) or "generative engines" (GEO) as the user-facing product. This emphasis sits at the model layer: how LLMs process text, what they're trained on, and how retrieval-augmented generation works in practice.

The discipline covers content discoverability, semantic clarity, and extractability — ensuring that when an LLM encounters your page, it can parse the information, understand its context, and potentially use it to answer queries. Despite the different terminology, in practice it covers the same ground as AEO and GEO.

LLMO vs AEO vs GEO

All three terms describe the same fundamental discipline with different emphasis. LLMO emphasises the technology (the model itself), GEO emphasises the output (generative answers), and AEO emphasises the function (answering questions). The industry is converging around AEO as the umbrella term because it best describes what users experience.

If you're doing AEO, you're doing LLMO. There's no need for a separate LLMO strategy — optimising for answer engines inherently means optimising for the LLMs that power them. The naming distinction is mostly academic at this point.

Why LLMO Matters for SEO Professionals

Even if the terminology is redundant, the underlying concept is critical. LLMs are now the intermediary between users and your content. Traditional SEO metrics don't capture whether AI can find and use your pages — we tested this directly with our AI navigation research.

In 3,348 AI navigation tasks across 269 websites, we found that 91% of successful navigation completes within two clicks. Position in the DOM matters more than semantic relevance for AI agents. Search-first approaches either succeed instantly (90%) or fail badly (27%). These findings suggest that SEO success remains essential for AI discoverability, but it's no longer sufficient on its own.

How LLMs Find and Use Your Content

There are two main pathways. First, your content was in the LLM's training set, so it "knows" about it from pre-training. Second, the LLM searches the web in real-time and retrieves your content via RAG (retrieval-augmented generation).

For SEO professionals, path two is where you have control. This is where technical accessibility, content structure, and extractability matter most. Most websites are built for crawler access, but this mental model doesn't quite match how AI agents work.

Related Terms

• Answer Engine Optimisation — The umbrella term for optimising content for AI answer engines
• Generative Engine Optimisation — Focus on optimising for generative AI output
• Retrieval-Augmented Generation — The technical foundation for real-time LLM retrieval
• AI Crawling — How AI agents discover and index web content
• AI Search — The user-facing search experience powered by LLMs

Want to see how LLMs interact with your site? Compass audits how AI agents navigate and extract from your content.