There is a comfortable story circulating in the industry: that generative engine optimisation is just SEO with a few prompt tricks bolted on. Write some FAQ schema, sprinkle in question-shaped headings, and the chatbots will cite you. The story is comfortable because it requires no one to learn anything new. It is also wrong.
Surface versus substrate
SEO optimised a ranking function over documents. The function was opaque but stable: it consumed signals — relevance, authority, quality — and emitted an ordering. The entire craft was reverse-engineering that ordering.
GEO operates against a generation process, not a ranking function. There is no ordering to climb. There is a probability that a given statement is retrieved, trusted, and reproduced inside a synthesis you never see. The mathematics is different, the feedback loop is different, and — critically — the unit is different. SEO ranked pages. GEO concerns claims.
Why the difference is operational, not semantic
If the two were the same discipline with new vocabulary, the same tactics would transfer. They do not:
- Keyword density is meaningless to a model that embeds meaning, not tokens.
- Link equity does not move a claim through a retrieval index.
- Freshness hurts certain claim types — research durability beats recency, the opposite of the news-SEO playbook.
These are not refinements of SEO. They are reversals. A discipline whose best practices invert the predecessor’s best practices is not the predecessor “with prompts.” It is a new thing, and it deserves its own name, its own metrics, and its own research program.
The honest position
GEO is early. Much of what is sold as GEO today is repackaged SEO, and some of it is outright snake oil. The corrective is the same one that matured SEO two decades ago: measurement. Controlled corpora, reproducible runs, published method. That is the work this archive exists to do.
