Compare three reputable GTA builders for a pre-construction home purchase.

The question set mirrors how real buyers ask: open category questions, local questions by municipality, risk and compliance questions, innovation questions and outcome questions. Pillars are not weighted: each of the 16 scored questions counts equally, so a pillar's influence on the ranking is its number of questions. A branded question is asked for context and reported on business pages but never counts toward rank.

Each question is asked verbatim, three times, to each AI assistant scored in the reading (the reading lists them; the first home-builders reading scored ChatGPT, Gemini and Google AI Mode), with a fresh session and no account history. The assistants' answers are recorded as structured cells: which businesses were named, in what order, and which sources were cited. Answer text is retained for audit and is not published.