The method: six questions Taiwanese parents and students actually ask ("Which admissions consultants are most trustworthy for top US universities?", "Beyond cram schools, which strategy-led consultants exist?"), put simultaneously to ChatGPT, Gemini, Claude, Google AI and Perplexity - 30 checks, 27 substantive answers. Every brand mention and every citation recorded.
1 - The result: one giant, absent locals
| Brand | Named (/30) | Spread |
|---|---|---|
| Crimson Education | 15 | Named by all five engines - the only cross-engine consensus |
| IDP Education | 10 | Named by four engines |
| Others | 1 each | Scattered appearances |
| Taiwanese boutique consultants | — | Almost entirely absent |
Taiwan's best-known boutique consultancies - household names in parent circles - did not appear with any consistency. AI's map of the category barely overlaps with the parent-circle map.
2 - Who feeds the answers
Crimson's 15 mentions are no accident: its own site was cited 9 times - the only brand whose content directly feeds the answers. The cited page types are consistent: application guides, admissions data, school comparisons. The rest of the citation corpus is dominated by mass-produced education knowledge bases.
3 - Being read is not being recommended
The study's most counterintuitive finding: one education knowledge site was cited 7 times - and recommended zero times. AI reads its content as reference material but never files it as a service provider. Content earned it information authority without converting into recommendations. The lesson for anyone "doing content": guides alone are not enough - every page must also make unmistakably clear who you are, whom you serve and how you differ, or you end up in AI's encyclopedia drawer instead of its shortlist. We call this: being read ≠ being recommended.
4 - Why AI's list differs from the parent circle's
Parent-circle reputation is built in LINE groups, seminars and word of mouth - all invisible to engines. AI sees only retrievable pages: citable guides and data, third-party comparisons, pages that answer buyer questions directly. The bad news: offline fame earns nothing here. The good news: almost nobody is competing seriously yet - half the cited content is machine-produced knowledge-base filler, and actual practitioners are absent. Whoever enters first defines the category.
5 - Three moves for practitioners
1. Copy the proven playbook: the winning brand gets cited for guides, admissions data and comparisons - build the same on your own site.
2. Enter the sources AI already reads: the citation table is a map; genuine presence there beats any homepage rewrite.
3. Don't be an encyclopedia: every page should carry citable knowledge and a clear statement of who you are - then verify with fixed-question weekly rescans.
Want your own position measured? Run a free scan - one real buyer question, live →