Generative Engine Optimisation (GEO)

Traditional Search Engine Optimisation (SEO) competes for ten blue links by matching keyword queries and earning backlinks to climb a static SERP. Generative engines work differently: when a user prompts an assistant with a commercial or technical question, the system decomposes the query into multiple sub-searches (query fan-out), retrieves candidate pages via web search indexes, extracts dense passages into its context window, and synthesises a single cited response.

GEO optimises for every stage of that RAG pipeline. Technically, it ensures AI crawlers (`OAI-SearchBot`, `ClaudeBot`, `PerplexityBot`, `Google-Extended`) are explicitly permitted in `robots.txt`, exposes machine-readable site maps via `/llms.txt` and clean Markdown endpoints (`Accept: text/markdown`), and embeds JSON-LD schema (`Organization`, `Article`, `DefinedTerm`, `FAQPage`) to disambiguate entities. Editorially, GEO replaces fluffy 2,000-word intros with high-fact-density prose, verifiable primary statistics, dated regulatory specifics, and comparison tables that a synthesis model can lift cleanly.

For example, a UK B2B payroll platform rewrote its compliance guides with explicit HMRC threshold tables, primary statutory citations, and a root `/llms.txt` index; within eight weeks, its brand citation share across ChatGPT and Perplexity queries for UK enterprise payroll doubled.

Buyer research has shifted decisively upstream into conversational interfaces. When a CTO or procurement lead asks an AI assistant to shortlist three UK vendors that meet a specific regulatory or architectural requirement, brands absent from the model's retrieved context or training priors never make the shortlist, even if they rank on page one of classic Google search.

However, GEO is not a replacement for foundational technical SEO or genuine domain authority. Because live AI search engines rely heavily on underlying Bing, Google, and proprietary web indexes to fetch candidate pages before synthesis, a site with broken crawlability, thin generic AI-generated copy, or zero off-site entity corroboration on Reddit, GitHub, and trade press will be ignored by both classic and generative engines.