Generative Engine Optimization: Get AI Agents to Cite You
AI agents and answer engines are changing how people find information. GEO — optimizing content to be cited by AI — is how small sites stay visible.
Generative Engine Optimization (GEO) is the practice of structuring your content so that AI agents and answer engines — ChatGPT, Perplexity, Google AI Overviews, Gemini, Claude — cite you in their answers. With traditional search click-through rates falling as AI answers replace result lists, GEO is how a small site stays visible to people who never click a blue link. Here's what it is and how to apply it.
Why GEO exists
For twenty years, the measure of online visibility was simple: rank high on a search results page, get the click. That's changing. When you ask an AI assistant a question, it often answers directly — and if it cites a source, it links it, but many people just read the answer and move on. The click may never happen.
The consequence is that traditional search volume is declining as people shift toward direct answers. The marketing research firm Yotpo and others point to the same trend, and Google itself now publishes guidance for optimizing for its generative AI features. That's the shift in one line: the answer engines are reading your page and deciding whether to cite it — and whether you get cited is a different skill from ranking in results.
GEO vs. SEO
GEO looks like SEO but has a different goal. SEO aims to win a ranking and get a click; GEO aims to get cited in an answer. Practically, that changes what "rank" means — there's no position to watch, only whether the AI mentions or links you when asked the right question.
| SEO | GEO | |
|---|---|---|
| Goal | Rank in results | Be cited in an AI answer |
| Metric | Position, click-through rate | Share of voice, citations |
| Audience | A person who clicks | A model that reads and cites |
| Key asset | Keyword hierarchy, backlinks | Clear, sourced, structured content |
| Success looks like | A blue link on page one | Your name in an answer |
The search engine itself now explains how it decides what to cite for generative AI: it runs a set of concurrent related queries (a "query fan-out") to gather context, then composes an answer. The practical implication is that a model cites content that's easy to retrieve, understand, and extract a clean answer from.
How to apply GEO
The good news is that GEO favors the same things that make content genuinely useful. You don't need to game a model; you need to make your content the kind that a model can parse and trust.
- Answer the core question in the first two sentences. Answer engines pull from the top of a page. Lead with the answer, then explain. This is the single highest-leverage change.
- Structure with headers and crisp sections. Models retrieve and quote cleanly from content with a clear hierarchy: H2 for each subtopic, short paragraphs.
- Include facts and specifics, not just claims. A model is more likely to cite a page with concrete, checkable information than one with vague assertions. Cite your sources.
- Use tables and lists where they fit. Structured data is easy to extract and reference. Comparing options in a table makes you the obvious citation for "compare X and Y."
- Make it citable by name. A model will say "according to [your name]" if your content is distinct and authoritative. A clear author, a strong take, and a consistent brand all help.
The industry guides agree on the same fundamentals — and because the bar is "be retrievable and trustworthy," the work compounds rather than chasing an algorithm update.
What this means for a personal agent or a small site
Here's the part that ties it to what we build at Wolffish. The current web is being read by agents — not just by people. A personal AI agent that researches for you, or an assistant that answers questions, is exactly the kind of system that pulls from content like this. If your business' customer asks an agent a question about your kind of product, the agent gets to decide whether you're the answer.
The practical upside: you don't need to out-engineer Google. You need to write clearly, answer directly, and be the kind of source an agent can trust. That's not a huge lift — mostly it's the discipline of leading with the answer.
Generative Engine Optimization — the takeaway
Agents are reading the web for you too — set up a research agent at wolffi.sh/start#research.
The takeaway
Search isn't disappearing; it's getting answered. GEO is the discipline of being the source that an AI cites when someone asks a question you're well placed to answer. Lead with the answer, structure clearly, be specific, and cite your sources — and you'll be cited too.
