Generative engine optimisation

Traditional SEO aims for a ranked link. GEO aims for a citation inside a generated answer, where the buyer may read your point without ever visiting your site.

Google and the researchers who named GEO disagree about whether it’s a separate discipline. The term comes from a 2023 paper by researchers from Princeton and IIT Delhi, presented at KDD 2024. It found that adding citations, quotations from relevant sources and statistics to content raised its visibility in AI answers by up to 40%, and that results varied by subject. Google’s position, in its May 2026 guide to generative AI features, is that optimising for its AI search is still SEO, because those features draw on Google’s search index and its core ranking systems. Both are right about different engines. For Google’s AI features, good SEO covers most of it. For ChatGPT, Claude and Perplexity, the rules differ enough to need their own work.

Why ranking in Google isn’t enough

The engines outside Google don’t simply reuse Google’s results. Ahrefs found that on average only 12% of links cited by ChatGPT, Gemini and Copilot rank in Google’s top 10 for the same prompt. Even inside Google, the link between ranking and citation has weakened: Ahrefs measured 76% of AI Overview citations coming from top-10 pages in July 2025, and 38% when it re-ran the study in 2026.

The crawlers differ too. A 2024 study by Vercel and MERJ found that none of the major AI crawlers, including OpenAI’s and Anthropic’s, rendered JavaScript. Googlebot does. A page Google reads perfectly can be close to empty for the crawlers behind ChatGPT and Claude.

What GEO involves

Area	What to do	Why it matters
Access	Let AI crawlers through robots.txt and your CDN, and serve main content in the HTML	A crawler that is blocked or can’t render the page has nothing to cite
Evidence	Add named statistics, quotations and citations to sources	The GEO paper’s strongest finding
Answer structure	Open each section with a direct, self-contained answer	AI tools lift passages, not whole pages
Coverage	Answer the follow-up questions around each topic	Google says its AI features may split a question into subtopics and search each one
Entity clarity	Describe your company the same way on your site, profiles and the third-party pages that mention you	AI tools combine what several sources say about you
Measurement	Track citations by engine, on a schedule	No single tool covers every engine

What Google says you don’t need

Google’s guide includes a section on GEO and AEO myths. Its AI features documentation says you don’t need special schema, AI text files or extra markup to appear in AI Overviews or AI Mode. The guide does flag one requirement most site owners miss: a site must be included in Search generative AI features in Search Console to be eligible to appear in them. Check that setting before anything else.

Measuring GEO

Each engine needs its own measurement. Search Console’s Generative AI report shows how often your pages appeared in Google’s AI features, as impressions only. Bing Webmaster Tools reports Copilot citations page by page. For a Canadian taxation SaaS product, that’s where we track its Copilot citations. ChatGPT, Claude and Perplexity offer no site-owner reporting, so I track them with my own tool, which runs a fixed set of buyer questions through each engine on a schedule and records who gets cited.

GEO and AEO

The two terms overlap, and people use them loosely. In this glossary, answer engine optimisation covers structuring content so any engine can lift a direct answer from it. GEO covers the wider job of getting cited across generative engines, including access, evidence and measurement. Google’s guide treats both as part of SEO.

Common mistakes

• Treating GEO as a replacement for SEO. Google’s AI features still draw on its search index.

• Assuming a page that ranks in Google is visible to ChatGPT or Claude.

• Blocking AI crawlers by accident through a CDN or firewall setting.

• Publishing claims without named sources, statistics or examples, which is the opposite of what the GEO research found works.

• Buying special markup or AI text files that Google says it doesn’t use.

• Measuring one engine and assuming the others behave the same way.