There’s a report that gets forwarded to every business owner, every month, by every marketing vendor in America. It shows keyword positions moving up and to the right, little green arrows next to phrases like “plumber near me,” and a summary line celebrating page-one placement. That report used to describe reality. Today it describes a channel that is quietly emptying out — and the green arrows are pointing at a room fewer and fewer customers ever enter.

This isn’t an argument that SEO is dead. It’s an argument that the metric is dying. Ranking position measures where you stand in a list. But the list is no longer where decisions happen. Decisions happen inside an answer — composed by Google’s AI Overviews, by ChatGPT, by Perplexity — and the only businesses in the room are the ones the machine chose to cite.

Section 01The Numbers Behind the Empty Room

For a decade, the deal was simple: rank high, get the click, win the customer. The first half still works. The second half is collapsing, and the newest clickstream research puts hard numbers on it.

The accelerant is easy to name. AI Overviews now appear on more than a fifth of Google searches, and when one is present, user behavior changes dramatically: Pew Research found that people click a traditional result on just 8% of searches with an AI Overview, versus 15% without one. Nearly half the click probability, gone — on the query types that are expanding fastest.

Here is the detail most reports skip, and the one that should reorganize your strategy: of the people who see an AI Overview, only about 1% click through to the sources cited inside it. The overview isn’t a preview of your website. For most customers, it is the visit. Whatever the machine says about your category is the entire brand impression — which makes the question “what does it say, and who does it quote?” the most commercially important question in organic marketing.

“The AI answer isn’t a preview of your website anymore. For most customers, it is the visit.”

Fiore AI — Growth Research

Section 02Ranking and Citation Are Different Games

The instinctive response is reassuring: “We rank well, so we’ll get cited.” Sometimes. But ranking algorithms and generative engines evaluate fundamentally different things, and conflating them is how well-ranked businesses end up invisible in AI answers.

Classic ranking is a popularity contest with a long memory — backlinks, domain age, engagement history. Citation is an extraction test with no memory at all: when the engine reads your page in the moment, can it safely lift a clear, current, verifiable answer? That test rewards a different set of behaviors:

1. Answer-first architecture. The direct answer within the first hundred words, then supporting depth — because extraction models weight what they can quote cleanly, not what you build toward.
2. Structured data that matches visible content. Organization, Service, and FAQ markup tell the engine exactly what you do; mismatched or missing schema leaves it guessing, and engines don’t cite guesses.
3. Verifiable specificity. Concrete figures traceable to named sources. Generative engines are becoming reflexive fact-checkers, and unverifiable claims are quietly filtered out of answers.
4. Entity consistency. The same name, services, and service area everywhere the model might look — your site, your profiles, your directories. Contradictions read as risk.

Notice what’s absent from that list: a ten-year-old domain and fifty thousand backlinks. This is the strategic opening. In our audit work across SMB and franchise sites, we consistently see well-structured pages on newer domains earning AI citations while their higher-ranked competitors — sites optimized brilliantly for a contest that’s ending — go unmentioned. Citation can precede ranking by months. For challengers, that’s not a footnote. That’s the whole opportunity.

Section 03What to Measure Instead

Retiring a metric only works if you replace it. The successor is citation share: of the questions your customers ask AI engines about your category and your market, in what percentage are you mentioned, quoted, or recommended — and is the framing accurate? It’s the AI era’s version of share of voice, and it’s measurable today.

The operational shift follows the measurement shift. Ask the major AI engines the twenty questions that precede a purchase in your category — cost, timelines, comparisons, “is it worth it” — and record who gets cited. That baseline is uncomfortable for almost everyone who runs it, which is precisely why it’s valuable. Then restructure your highest-intent pages answer-first, repair the schema, verify the claims, and re-run the baseline quarterly. Direction of citation share, not direction of green arrows.

None of this makes rankings worthless — transactional and navigational queries still deliver clicks, and Gartner’s projected 25% decline in traditional search volume by 2026 still leaves a very large 75%. But a metric doesn’t have to reach zero to stop deserving the top line of your report. It just has to stop describing where customers make up their minds. Rankings crossed that threshold. Citation share is what’s on the other side.

FAQFrequently Asked Questions