Getting Found on Google Isn't the Goal Anymore — Getting Cited by AI Is

Type a question into Google today and there's a decent chance you never click a link to answer it. An AI Overview does it for you, right at the top of the page, built from a handful of sources it decided were worth summarizing. Ask the same question in ChatGPT or Perplexity instead of Google at all, and there's no results page to rank on — just an answer, assembled from whichever sites the model trusted enough to cite. Being "number one" on a page fewer people are scrolling past isn't the win it used to be. Being the source the answer is actually built from is.

The traffic that's disappearing from your analytics

This shows up as a strange kind of decline. Rankings hold steady, sometimes even improve, and traffic drops anyway — because the click that used to happen after a search now happens inside the search. Industry data on "zero-click" searches has been climbing for years; AI Overviews and chat-based answers accelerated a trend that was already underway, they didn't start it. A business watching only its keyword rankings can look completely healthy while the actual channel — people finding it and choosing it — quietly erodes underneath that metric.

Ranking and being cited are not the same skill

Traditional SEO optimizes for a results page: ten blue links, competing on relevance and authority signals a crawler can measure. Getting cited inside an AI answer is a narrower, harder target — the model isn't listing ten options, it's picking the two or three sources it's willing to summarize and stand behind. That favors content that states things plainly and specifically enough to quote, structured clearly enough for a model to parse without guessing, and backed by enough consistent signal elsewhere that the model has reason to trust it. Vague, keyword-stuffed pages that used to limp into page one now have nothing a model can confidently lift and attribute.

What actually earns a citation

Three things tend to separate the businesses AI models cite from the ones they skip. Clear, direct answers to specific questions — not a wall of marketing copy the reader has to mine for the actual point. Structured data that tells a crawler exactly what a page is (an article, a product, an FAQ, a review) instead of leaving it to infer. And consistency — the same facts, the same positioning, stated the same way across a site and the web around it, because a model is less likely to cite a source that contradicts itself from one page to the next. That last point is the same discipline behind why reach without recall is a leaking bucket — a model trusts what it recognizes as consistent for the same reason a person does.

Why this is a technical decision, not just a content one

The businesses getting cited aren't necessarily writing better sentences — they're running sites that a model can actually parse. That means structured data implemented correctly, pages that load and render fast enough to be crawled reliably, and information architecture that puts a clear answer where a crawler can find it instead of burying it under a design flourish. This is the same gap that quietly sinks Core Web Vitals scores: a site that looks finished and still fails the visitor who's actually using it is invisible to an AI crawler for the same underlying reason it's slow for a person — nobody built the foundation the surface is standing on.

None of this replaces ranking well on Google — that channel isn't going away. It's a second, stricter test running alongside it, and it rewards exactly the things a shortcut can't fake: a real, clearly stated position, and a site built well enough to carry it.

This is a Grow post — the last of our four stages: Define, Express, Build, Grow. If your rankings look fine but you've never checked whether AI answers actually cite you, Grow is where that gets fixed.