Original research on the AI answer economy. We put real buyer questions to five AI engines — ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews — run every check multiple times for consensus, and publish what comes back: who gets recommended, which sources feed the answers, and what it means for how brands get found.
The direct answer, and the mechanics behind it — shortlists, sources, engine divergence and drift — explained with data from 631 measured AI answers. Plus the three moves that actually change your number.
120 scored answers about five Taiwanese B2B suppliers gave 2 recommendations — so we tested why. Three of the sites refused AI crawlers 22 times while serving a browser from the same IP every time, and their robots.txt says crawlers are welcome.
Every list like this is written by one of the tools — this one too. So we measured the category with our own scanner and published the numbers, starting with our own 0%. Plus the three questions that test any vendor in ten minutes.
307 answers about thirteen Taiwanese brands. Answers built from retrieved sources named the local brand 68% of the time; answers written from memory alone, 25% — and ChatGPT answered half the questions citing nothing at all.
210 answers to 18 buyer questions. Cursor leads everywhere except ChatGPT — where Copilot rules — and each engine leans toward its own family. Vendor blogs out-cite official sites, for the third category running.
The case for taking AI visibility seriously, made only with things we can show: 421 measured answers on shortlist concentration, why size doesn't guarantee a slot, source competition, engine fragmentation and drift.
186 answers about the AI-visibility industry itself. The category's leaders are nearly invisible to AI, the case for buying GEO is written by its sellers — and we're publishing our own 0%.
235 answers to 20 real buyer questions across four engines. Paddle out-recommends Stripe overall, Lemon Squeezy is invisible to ChatGPT, and two rival MoRs' blogs out-cite the official docs.