Introduction
If your organic traffic has quietly flattened or dipped over the past year, and you can't fully explain why from your rankings alone, you're not imagining it. The buyer journey itself has changed shape — and a meaningful slice of it no longer touches your website at all.
This isn't a story about Google losing its dominance overnight. It's a story about a second, parallel research channel that didn't exist a few years ago, that now intercepts buyers earlier in their decision, synthesizes an answer on their behalf, and decides — often without telling you — whether your brand gets mentioned in that answer.
Here's what actually changed, and what it means for how you measure visibility going forward.
The data: AI engines now handle a real share of informational queries
Google still processes the overwhelming majority of searches. But industry tracking through 2026 consistently points to AI platforms capturing roughly 12 to 20% of informational query volume — the research, learning, and "what's the best way to..." questions that used to flow almost entirely through a search bar.
That shift looks different depending on where you measure it:
- Google's AI Overviews appeared in roughly 18% of searches in early 2025 and climbed to about 26% by late 2025, and they're no longer confined to informational queries — they expanded from 8% to 18% of commercial queries as well.
- Research, learning, and commercial-investigation queries — the exact category that historically sent visitors to blogs, comparison pages, and product content — are the segments AI platforms are capturing fastest.
- B2B buying has moved even further: 94% of B2B buyers now use generative AI tools somewhere in their purchase process.
None of these numbers mean Google is going away. They mean a second channel now exists where buyers form opinions, narrow shortlists, and sometimes make decisions before a traditional search ever happens — and that channel runs on different rules than the ones most marketing teams have spent a decade optimizing for.
A purchase decision in 2020 vs. 2026
It's worth walking through what actually changed at the behavioral level, because the shift is easy to underestimate from a dashboard.
2020: A buyer researching, say, project management software types "best project management software" into Google. They get ten blue links. They click into three or four — a review site, a comparison article, two vendor homepages. They form their own opinion by reading and comparing. Every one of those visits is a session you can see in your analytics, a page you optimized, a CTA you A/B tested.
2026: That same buyer opens ChatGPT or Perplexity and asks, "what's the best project management software for a 10-person design team?" The AI engine doesn't return ten links — it synthesizes one answer, pulling from multiple sources at once. Pew's analysis of Google's AI Overviews found that 88% of summaries cited three or more sources, with only 1% relying on a single source. The buyer reads a confident, comparative answer naming two or three tools, asks a follow-up question to narrow further, and may go straight to a vendor's pricing page — having never visited a single one of the sources the AI actually pulled from to write that answer.
The research phase didn't disappear. It moved inside the AI conversation, where you have no analytics, no page to optimize in the traditional sense, and often no idea the conversation happened at all unless you're checking whether you're cited.
Zero-click and zero-visit: when AI answers without sending anyone anywhere
"Zero-click search" used to describe a narrow case — a featured snippet that answered a question well enough that the user never needed to click through. In 2026, it's closer to the default behavior.
The numbers are stark. Zero-click searches reached record levels in 2025, with 58.5% of U.S. searches and 59.7% of EU searches ending without any external click, and the zero-click rate jumped to an average of 83% specifically when an AI Overview appeared. A controlled 2026 study found an even sharper effect at the individual query level: a randomized field experiment measured a 38% drop in clicks to websites when AI Overviews were present.
Conversational AI engines push this further into what's better described as zero-visit search — not just skipping a click on one query, but resolving the entire research phase of a purchase decision across several back-and-forth turns, with the website never entering the picture until (if ever) a final, narrowed decision point.
This is the uncomfortable part for anyone used to measuring marketing performance through sessions and pageviews: a buyer can be deeply influenced by your content, your category positioning, and your competitive standing — entirely through an AI answer — without a single visit landing in your analytics. The influence is real. The traffic evidence isn't.
What it means to be a "cited source" vs. a "ranked page"
Ranking and citation look similar from a distance — both are forms of being chosen — but they reward genuinely different things, and confusing the two is why a lot of teams with strong SEO are caught off guard by weak AI visibility.
| Outcome | What earns it | Who decides |
|---|---|---|
| Ranked page | Backlinks, keyword relevance, site authority, crawlability | Google's algorithm; the user compares links |
| Cited source | Clear facts, recognizable entities, structured data, third-party coverage | AI model trust when constructing an answer |
A ranked page earns its position through backlinks, keyword relevance, site authority, and crawlability. Google's algorithm decides you deserve a spot on page one, and the user does the comparing.
A cited source earns its position by being something an AI model trusts enough to lift from directly when constructing its own answer. That depends on different signals: how clearly your content states facts, whether your entities (brand, product, people) are recognizable to the model, whether you have structured data that removes ambiguity, and — critically — whether your content shows up in the kind of third-party coverage AI models weight heavily. A 2026 analysis of more than 25 million links found that 84% of AI citations come from earned media — editorial coverage in independent publications — rather than brand-owned content or paid placement.
That last point reframes the whole problem. Ranking is something you can largely engineer on your own domain. Citation is something you partly earn off your domain, through how the broader web talks about you, combined with how easy your own pages make it for an AI model to extract a clean, citable fact once it gets there. A page can rank well and still never get cited, if a model can't parse a confident, specific claim out of it.
This distinction is exactly why AI Search Optimization (AISO) and SEO aren't the same discipline wearing different clothes — they optimize for different evaluators with different criteria, even when the underlying content overlaps. A full breakdown of how AISO and SEO compare covers where the two strategies align and where they diverge.
The new metric that matters: share of AI citations
If clicks and rankings only tell part of the story now, the question becomes: what should you actually be measuring?
The closest analog to "share of voice" or "share of search" in this new landscape is share of AI citations — out of all the times an AI engine answers a question relevant to your category, what percentage of the time does it mention or recommend you instead of (or alongside) your competitors?
It's a harder number to get than a rankings report, because there's no public dashboard for it the way there is for SERP positions. In practice, measuring it means systematically asking the AI engines your buyers actually use — ChatGPT, Gemini, Claude, Perplexity, Google AI Overviews — the questions your buyers actually ask, and tracking whether you show up, how you're framed, and who shows up instead of you.
A few ways to start building that picture:
Run a baseline visibility check. An AI Visibility Score gives you a 0–100 read on how likely each major AI engine is to surface and recommend your site right now, broken down across the dimensions — entity clarity, structured data, FAQ readiness, citation potential — that actually drive whether you get cited.
Benchmark against the competitors you're actually losing citations to. Competitor Analysis runs side-by-side comparisons across the same dimensions and engines, so you can see exactly where a rival is winning the citation and what's different about their content.
Track it over time, not as a one-off. Share of AI citations moves as you publish, as competitors publish, and as the underlying models update — treat it as an ongoing metric, the same way you'd treat search rankings.
The takeaway
The buyer journey didn't get shorter in 2026 — it got a second track. Some buyers still search, click, and compare the way they did in 2020. A growing share now ask, get a synthesized answer, and act on it without ever generating a session you can see. Traditional SEO metrics weren't built to capture that second track, which is exactly why "good rankings, flat traffic" has become such a common and confusing pattern for marketing teams this year.
Measuring rankings tells you how visible you are to people still searching the old way. Measuring share of AI citations tells you how visible you are to everyone else.