Evidence over vibes
AI search visibility, GEO and AEO: what the engines actually do, what the data shows, and what to do about it.
A results page ranks pages and lets you choose. An AI answer retrieves, selects and writes a few sentences that name a few brands for you.
There's no single citation algorithm. Retrieval, relevance, freshness and source type each play a part — and some of them are yours to influence.
AI knows a brand from memory built during training, or from pages fetched live. You can influence the second one quickly and the first one only slowly.
Each engine documents a little about how it finds sources and leaves most of the rest unpublished. Here's exactly what's confirmed, by whom, and what still isn't.
Anthropic, which makes Claude, runs three separate bots for training, search indexing and live user fetches. Each is controlled independently in robots.txt.
Gemini sits beside Google's search products, not inside them. Here's what Google documents about how it grounds answers in live search, and what's simply not published.
Perplexity retrieves live and always shows its sources, which makes it the easiest engine to check — and the one whose results track Google's rankings most closely.
AI Overviews and AI Mode both draw on Google's core search systems, but they cite different sources for the same question far more often than they agree.
Buyers don't type fragments into AI chat. They ask full, specific, often messy questions — and the type of question matters more than how many people search it.
AI search engines split one question into several related searches before answering. Coverage of those hidden sub-questions decides who actually gets retrieved.
You can't watch every engine daily with a small team. Start where your buyers actually ask, write content that travels, then check each engine on its own terms.
One-line definitions for the terms this blog uses most, each linked to where we cover it in depth.
A brand can be named without being cited, cited without being recommended, or recommended with no source at all. Measuring 'AI visibility' as one number hides which of these actually happened.
Five labels, one underlying job: get named, described and cited correctly wherever someone is searching for an answer.
ChatGPT can answer from what it learned during training or from a live web search it runs on demand. Only the second path involves choosing sources at all — and that choice leans heavily on Bing.
No. Clicks fall on queries where an AI summary appears, and that's real and measured — but SEO fundamentals are what AI features are built on, and the KPI stack is what actually needs to change.
AEO means writing so a direct-answer box or an AI engine can quote you. It is a real, older discipline with a new generation of surfaces — not a separate field from SEO.
AI search visibility is how often, how accurately and in what company AI engines name, describe and cite your brand. It measures the answer, not the ranking.
GEO is the practice of making a brand easier for AI engines to find, understand and cite. The research behind it is scoped, not a guarantee, and Google says that for its own features this is still SEO.