No. And most people optimizing for AI visibility right now don't realize this.
I went down a rabbit hole this week on how search behavior has shifted since AI chatbots went mainstream, and a few numbers stopped me.
Gartner predicted traditional search volume would drop 25% by 2026 because of AI chatbots. A mid-2026 study covering 35+ billion monthly searches found the real number is 29%. AI beat the prediction.
Zero-click searches, where you get an answer and never visit a site, now make up close to 60% of all searches. When an AI Overview shows up, that number jumps to roughly 83%.
Google still holds about 90% of traditional search. But when you count AI chat as part of "how people find information," Google's share of that combined pie has reportedly fallen from around 89% to under 60% in three years.
The bigger surprise
The "does one AI's trust transfer to all AI" question. It mostly doesn't:
- Different retrieval. Perplexity searches live almost every time. ChatGPT and Claude often answer from training data and only search when needed. Same question, completely different sourcing.
- Different indexes. These platforms don't all pull from the same search backend.
- Training data and live citation are separate channels. A model "knowing" about you from training isn't the same as it citing you live.
GEO is becoming a real discipline
Some call it GEO (Generative Engine Optimization), some call it AEO (Answer Engine Optimization). Same idea: instead of optimizing to rank in a list of links, you're optimizing to become the answer itself.
A few things that actually seem to move the needle:
- Put your direct answer in the first 100-150 words. One analysis found 44% of AI citations pull from the first third of a page, mostly the intro.
- Back claims with real numbers and named sources instead of hedging language.
- Brand mentions across the web now matter more than backlinks for getting cited by AI. One study found mentions correlate with AI citation about 3x more strongly than backlinks do.
- Keep content fresh. AI systems seem to favor recently updated pages.
This isn't a replacement for SEO, it sits on top of it. You still need a fast, crawlable, well-structured site. GEO is the next layer.
Where this is heading
It's also why I've been watching what tools like Landing Creator are building. Instead of manually writing hundreds of location or offer-specific pages, it generates them directly structured for both Google and AI citation, schema markup, FAQ blocks, internal linking, the works, and maintains an llms.txt index so AI crawlers can discover what a site actually offers. That's a fairly direct answer to the shift these numbers are describing.
Curious what others are seeing in their own traffic.
Top comments (2)
The training-vs-live-citation split is the part I’d make explicit in any benchmark. In a 100-query ChatGPT shopping test we ran, 75% of the top products matched Google Shopping’s top three. That does not prove a shared index, but it does show that retrieval behavior can look very different by surface and intent. I’d test your transfer question as a matrix: same prompt, fresh session, search on/off where available, then record mention and owned-domain citation separately. Otherwise a model’s prior knowledge can look like cross-engine trust when no live source was used at all.
That's a really important distinction to make explicit. Most visibility benchmarks don't separate those two channels, so you end up optimizing for "AI knows about me" when what actually matters is "AI retrieves and cites me live." The shopping overlap result is a good example of why that ambiguity is dangerous, 75% could mean shared retrieval or it could just mean those brands have strong enough training signal that no live search was needed at all. Two very different problems to solve. Will be more precise about that split in anything I write next on this.