Three shifts separate GEO from classic SEO. Each one gets a section with a graphic here: on top, how it works in classic search; below, what changes in AI search. Once you understand these three, you understand the rest of the handbook.
Key takeaways
- From retrieval to synthesis: Search no longer delivers a list; it builds an answer from several sources.
- Weight instead of rank: Instead of a fixed spot, what counts is how strongly your content flows into the answer.
- Entities instead of pages: AI assesses your brand as a whole, not each page on its own.
- The foundation is still SEO: Only what was found beforehand makes it into the answer.
On this page:
- 1. From retrieval to synthesis
- 2. Weight instead of rank
- 3. Entities instead of pages
- Where the handbook goes next
1. From retrieval to synthesis
In classic search, Google finds matching pages and shows them as a list. Reading, comparing and deciding is up to you. In AI search, the system does this work: a language model retrieves several sources and combines them into one answer (grounding).
SEO · SEARCH FINDS query │ ▼ ┌──────────────┐ │ Index │ └──────┬───────┘ ▼ 1. link 2. link 3. link → you read, compare, choose GEO · AI ANSWERS prompt │ ▼ ┌──────────────┐ │ LLM │ └──────┬───────┘ ▼ one answer with sources → the system reads and chooses
SEO delivers links. GEO delivers answers.
In short: if you are in the list, you can be clicked. If you are in the answer, you get read.
2. Weight instead of rank
Rank 1 is a step on a ladder. In an AI answer, that ladder does not exist. Several sources flow in with different weights, and the same question can produce a slightly different answer tomorrow.
Visibility thus becomes a probability: how often are you named when many people ask similar questions? That is exactly what is measured by share of voice.
SEO · THE LADDER 1. source A ◀ takes it all 2. source B 3. source C GEO · THE MIX source A ███████ 35 % source B █████ 25 % source C ████ 20 % │ ▼ one blended answer
Visibility is no longer a position, but a probability.
In short: instead of “What position am I in?” the question becomes “How often do I appear, and how?”. How to measure that is shown in Measuring AI visibility.
3. Entities instead of pages
Classic SEO often thinks in individual pages, one per keyword. AI models think in entities: they connect everything said about your brand in different places into one picture. That picture is built from your website and from what others write about you.
SEO · SINGLE PAGES page A → "best CRM" page B → "CRM pricing" page C → "CRM review" GEO · ONE ENTITY BRAND ┌──────┼──────┐ offer pricing reputation └──────┼──────┘ ▼ one coherent picture
Watch out: if your pages contradict each other, for example on prices or services, the picture gets blurry. That makes it harder for AI to recommend you with confidence.
Where the handbook goes next
With these three shifts you have the foundation of part 1. The chapters before go deeper on each: How AI search builds answers on section 1, Visibility without rankings on section 2.
Next in the handbook: Part 2 turns to the technical side. If you would rather find out right away how your own brand is doing, jump to Measuring AI visibility.
> geo-glossary
Terms from this article
- Prompt
- A prompt is the question or instruction someone gives an AI assistant.
- LLM
- An LLM is a large language model that generates text word by word based on probabilities.
- Query Fan-Out
- With query fan-out, AI splits a question into several sub-searches and merges the results into one answer.
- Mention
- A mention is your brand being named in an AI answer, with or without a link.
