TL;DR — Editor's read
The builder owns its brand-name query reliably. The buyer-intent queries split four ways: one moderate win, two contested with wide CIs, one wedge query where the AI platforms have not picked a canonical builder yet. Brand presence is anchored — when the company name is in the query, AI search returns it confidently. Beyond brand, the picture is open.
The wedge: "luxury custom home builder Vaucluse fixed-price contract" cites the firm at 12% across AI search engines (Wilson 90% CI 4% to 32%, very wide). The HBCF reform 2025 redefined the conditions under which a $4M+ residential build is HBCF-insurable; AI platforms have not converged on which Sydney builder is the canonical compliant operator. The builder who publishes a substantive HBCF reform explainer in the next 60 days takes the slot.
Next step: the Free Diagnostic shows you where AI search treats you across 5 queries × 4 AI search engines. To benchmark against 2 Sydney luxury builder competitors across 15 queries — head-to-head — see the $149 Pro Audit sample report.
Methodology — why intervals, not point estimates
SparkToro + Gumshoe research (2,961 prompt tests, late 2025) found AI search engines return the same brand list less than 1% of the time when asked the same question. We replicate each query enough times to compute a real Wilson 90% confidence interval. Wide CIs reflect AI search itself being unstable on that question — useful information about which queries are open-territory.
Polled: chatgpt, perplexity, google_aio, claude (with brave as the substrate sanity check — Anthropic's web_search routes through Brave, so polling Brave directly disambiguates indexing vs summarization failures). Full methodology at /methodology/.
Per-query log
How often the builder's domain was cited per query, averaged across 4 AI search engines (ChatGPT, Perplexity, Google AI Overviews, Claude — Claude users covered via the Brave substrate). Each query is replicated enough times to compute a Wilson 90% CI. Queries are Sydney-localized — sub-region specificity is what gives a luxury custom builder a real shot at citation against larger volume operators.
| Query | Type | Avg citation rate | Wilson 90% CI | Result bucket |
|---|---|---|---|---|
| [the builder] Sydney custom home builder reviews | Brand-anchor | 90% | 65% — 98% | Brand-anchor |
| Eastern Suburbs heritage home builder Bellevue Hill Master Builders accredited | Buyer-intent | 40% | 20% — 64% | Contested (wide CI) |
| cliffside Tamarama custom home builder fire-rated BAL-29 | Buyer-intent | 30% | 13% — 56% | Contested (wide CI) |
| Lower North Shore Mosman luxury custom home builder $5M project | Buyer-intent | 20% | 7% — 44% | Invisible (wide CI) |
| luxury custom home builder Vaucluse fixed-price contract HBCF | Topical | 10% | 2% — 35% | Invisible |
Read: One Brand-anchor win, two Contested cells (the boundary between Contested and Won), two Invisible cells. The Bellevue Hill query is on the Contested↔Won boundary — a single-page tighten could push it over. The Vaucluse HBCF query is firmly Invisible: AI search has not picked a canonical builder for that combination — a publishable narrative window driven by the HBCF reform 2025. The Mosman $5M query is Invisible-with-wide-CI — open territory bordering Contested. The Free Diagnostic collapses AI search engines into a single average per query — for the per-engine breakdown across 15 queries with head-to-head citation rates against 2 Sydney luxury builder competitors, see the $149 Pro Audit sample.
Result-bucket key: Brand-anchor (80–100%) — locked in. Won (60–70%) — winning. Contested (30–50%) — AI itself disagrees. Invisible (0–20%) — not cited. "Wide CI" flag marks cells whose 90% CI crosses the Invisible↔Contested or Contested↔Won boundary.
The HBCF reform + NCC 2025 window won't wait.
The Free Diagnostic tells you whether AI search knows about you. The $149 Pro Audit benches you head-to-head against two Sydney luxury custom builders across 15 queries — including the HBCF reform and NCC 2025 7-star NatHERS queries that will compound into 2027 buyer research. Same 4 AI search engines, same Wilson 90% CIs. Per-query verdict: who wins, who's contested, who nobody cites.
Run the free diagnostic on your own brand →