For two decades the rules were simple: pick a keyword, win the position, count the clicks. That model is ending — not because Google is disappearing, but because a growing share of answers is generated before the user ever sees a link.
A different unit of work: intent instead of keywords
In SEO, planning starts with keywords and their volumes. In GEO, the starting point is the question a customer asks the model — often long, contextual and unlike anything in keyword research tools. A single question can combine several intents at once: comparison, budget, location, trust.
So instead of a keyword list we build a conversation map: what the customer wants to know at each stage of the decision, and what answer the model should give at that moment for your brand to have a place in it.
A different measure of success: citations instead of positions
Position #3 in Google is a concrete, measurable value. In a generative answer there are no positions — there is presence or absence. A brand can be named, cited as a source, or skipped entirely. It’s a binary game with high stakes, which is why we measure:
- share of answers — in what percentage of answers to key questions the brand appears,
- mention context — whether the model recommends, compares, or merely mentions,
- cited sources — which content (yours and others’) builds that answer.
“In GEO you’re not fighting for a spot on the list. You’re fighting to make the list unnecessary.”
A different content format: write so you can be quoted
Models pick content that is easy to verify and safe to repeat: clearly stated claims, definitions, dated and sourced data, direct answers to specific questions. A wall of text written “for the crawler” loses to a paragraph that answers the customer’s question in three sentences.
In practice this means rebuilding content: FAQ sections, summaries at the top of articles, structured data, consistent brand naming, and authors with real expertise. None of this hurts SEO — but only together does it make content citable.
Where to start
Not with writing new content — with checking what the models say about you today. An AI visibility audit shows which answers you’re in, which ones your competitors own, and which questions are up for grabs. Only then is it worth planning content and measuring progress.