CROSS-CATEGORY: Two Independent Audits Published the Same Day Say the AI Software-Discovery Channel Is Both Farmed and Unreliable
Trellner and Haus Research both published on 2026-09-02 without citing each other, attacking opposite ends of the same pipe. Trellner shows the supply side is manufactured (215,128 templated pages across three interlinked domains, 59.8% of grounded sources outside the top 100k). Haus shows the retrieval side is broken (34.7% of figure-bearing citations don't support their number, and the "entry price" question type passed only 62.6%). For anyone selling software, the practical read is that the answer engine now sitting between a buyer and a category page can be gamed by templated pages and cannot be trusted on your pricing, so a vendor pricing page needs to be machine-fetchable and unambiguous or the model will cite someone else's number for it.
↳ Follow the thread