What is generative engine optimisation?

The shift GEO responds to

For twenty years, search returned ten links and the user chose. AI answer engines return one synthesised answer with a handful of citations, and the user often never clicks. Visibility now means being one of the sources the answer is built from, and being named in it. If a buyer asks an assistant for the best capability statement designer in Perth and the answer names three firms, being fourth in Google's classic results is worth very little.

How AI engines choose their sources

Each engine works differently in detail, but the pattern is consistent across them:

• Retrieval first. Most engines run a search behind the scenes, usually against Google or Bing's index or their own crawl, then read the top results. If a page is not retrievable for the query, it cannot be cited. Classic SEO is therefore the entry ticket.
• Extractable claims. The model looks for passages that answer the question directly: a definition, a number, a comparison, a list. Pages that bury the answer in marketing copy lose to pages that state it plainly.
• Entity confidence. Engines cite sources they can identify: a named organisation, a named author with credentials, consistent details across the web. Anonymous content is cited less.
• Corroboration. Claims that appear on several credible sites are trusted more than claims that appear once. Mentions on third-party sites matter as much as your own pages.
• Freshness for changing topics. For prices, regulations and technology, recent content wins.

What changes on a website

GEO does not require a new site. It requires the existing one to be legible to a machine reader and quotable by it.

Answer-first structure

Each page states its core answer in the first hundred words, then supports it. Headings phrased as the questions people ask. Definitions, numbers and comparisons in plain sentences rather than inside images or slides.

Structured data

Organisation, Person, Service, FAQPage and Article schema, kept accurate. It is the cheapest way to tell an engine who you are, who wrote the page and what it is about.

Author and organisation entities

Named authors with real credentials, consistent across the site, LinkedIn and industry profiles. An organisation page with the facts an engine wants to verify: location, founding date, leadership, sectors.

An llms.txt manifest

A plain-text file at the site root that tells AI crawlers which pages matter and what they contain. Explained in llms.txt explained.

Evidence density

Case studies with named clients and numbers, original data, and specific claims. Engines are built to prefer sources that look like evidence over sources that look like advertising.

Crawler access

Many sites block AI crawlers by default, often without knowing. If you want to be cited, GPTBot, ClaudeBot, PerplexityBot and Google-Extended need to be allowed in robots.txt. Our own robots.txt allows the crawlers that cite sources and blocks the ones that only harvest.

How GEO is measured

Rankings do not exist inside an AI answer, so the measure is presence: for a defined set of buyer questions, does the engine mention you, cite you, and describe you accurately? Tracked across engines over time, alongside classic organic traffic and revenue, because Google still pays the bills. We built SearchInsights AI to do exactly that tracking for our clients.

GEO and SEO are the same programme

Everything above is good SEO practice made stricter. A site that is fast, well-structured, authoritative and evidence-dense ranks in Google and gets cited by AI engines. A site that is thin, anonymous and blocked to crawlers does neither. The practical difference is emphasis: GEO puts more weight on extractable claims, entities and third-party corroboration, and less on the classic keyword-in-title mechanics. Our GEO service runs the two together for that reason.