Generative Engine Optimisation (GEO) explained — the new SEO for AI search
Generative Engine Optimisation (GEO) is the practice of structuring a website's content, schema and entity signals so that generative AI engines — ChatGPT, Perplexity, Claude, Gemini, Google AI Overviews — choose to cite the business when answering user questions. Where SEO optimises for ranked links, GEO optimises for citation inside a generated answer. The two share technical fundamentals (clean HTML, fast pages, schema), but GEO additionally rewards direct-answer paragraphs, FAQ structures, entity consistency and explicit crawler permissions for LLM bots.
What is GEO and why does it exist?
GEO — Generative Engine Optimisation — is a term coined in 2023 by researchers at Princeton and the Allen Institute to describe a new discipline: making content visible to generative search engines that don't return blue links, only synthesised answers.
It exists because user behaviour has shifted. Roughly one in three searches now ends inside an AI assistant rather than a search engine results page. If your business isn't structured to be cited inside that answer, you're invisible to a growing share of buyers.
GEO vs SEO — what's actually different?
Same fundamentals: crawlable HTML, fast load times, valid schema, internal linking, authoritative backlinks. Both engines need these.
Different content priorities: SEO rewards long-form pages with keyword density. GEO rewards short, specific, direct answers and FAQ blocks an LLM can lift verbatim.
Different permissions: SEO needs Googlebot. GEO needs GPTBot, ClaudeBot, PerplexityBot, Applebot-Extended, Google-Extended and CCBot — many of which are blocked by default in modern hosting templates.
Different metrics: SEO measures ranking position and click-through. GEO measures citation share — how often your brand appears in a generated answer for relevant prompts.
The five-step GEO playbook for SMBs
1. Allow the bots. Audit robots.txt and explicitly allow the major LLM crawlers.
2. Add structured data. Organization, LocalBusiness, Service and FAQPage schema on every commercial page (full breakdown in our schema markup guide).
3. Rewrite openers as direct answers. A 40–60 word plain-English summary at the top of each page.
4. Build FAQ blocks. 4–6 conversational Q&A items per page, marked up with FAQPage JSON-LD.
5. Publish a llms.txt. A root-level summary file that tells assistants who you are and which URLs matter.