Research
What the engines actually do.
We run 80+ queries across 5 AI engines for every client, every week. That’s a lot of data about how answer engines behave. Here’s some of what it tells us.
Finding 01
Google rank doesn't predict AI mention.
Across the categories we track, ranking #1 in classic Google search had almost no relationship to being named in an AI answer for the same query. The engines weight different signals — citations, knowledge graph identity, and third-party sentiment — not blue-link position.
0.2
correlation between Google rank and AI mention across tracked queries.
Finding 02
Three names is the ceiling.
When AI answers a "best [category]" question, it overwhelmingly returns two or three named businesses, not a list. The competition isn't for a page-one slot — it's for one of a very small number of seats inside the answer itself.
2–3
businesses named in a typical high-intent AI answer.
Finding 03
Each engine has its own sources.
Perplexity and ChatGPT browse lean on the live web index and candid discussion; AI Overviews and Copilot parse schema and reviews in real time; the base models reflect training data. Winning everywhere means working all three layers, not one.
3
distinct source layers, each with a different time-to-win.
Finding 04
Mentions trigger branded search.
When an engine starts naming a business, we consistently see branded search volume rise within 24–72 hours in Search Console. The AI mention isn't the end of the funnel — it's the moment a buyer goes looking for you by name.
24–72h
lag between a new AI mention and a branded-search lift.
Methodology:figures are directional, drawn from Weadot’s own client query maps across ChatGPT, Perplexity, Claude, Gemini, and Copilot. We publish what we can verify and flag what we can’t.