The method
In July 2026 we used Citable, our AI visibility tool, to put 12 real customer prompts about Auckland builders to seven AI engines, including ChatGPT, Google's Gemini, and Perplexity. The prompts were the questions homeowners actually ask: "best builders in East Auckland", "trusted local builders in South Auckland with good reviews", "how much does a bathroom renovation cost in Auckland", "kitchen renovation builders near Howick". We recorded every business named, every source cited, and every attribute the engines used to describe them.
Finding 1: decision questions are crowded, cost questions are empty
Ask "who are the best builders in East Auckland" and the engines collectively name close to 40 businesses. Ask "home extension builders in Auckland Central": 38 named. The recommendation prompts are a contested market.
Then ask "how much does a bathroom renovation cost in Auckland" and the answer nearly empties out: two businesses named across all seven engines. Home renovation cost: six. The engines want to cite someone on cost, and almost nobody has published numbers worth citing. That is the most actionable gap in the entire study: publish honest cost content and you can own a question with almost no competition.
Finding 2: reviews are the currency of AI recommendations
The attribute the engines repeated most was review evidence, verbatim. One South Auckland builder was described as having "70+ 5-star Google reviews" by three separate engines. Another was called "the highest-rated builder on Google for East Auckland" with "5.0 stars, 30 reviews". The engines don't just count reviews, they quote them as the reason for the recommendation.
Finding 3: the engines trust platforms more than websites
The most-cited sources weren't builders' own sites. Builderscrack was used in recommendations 15 times across six engines. NoCowboys 11 times. The government's building.govt.nz LBP register appeared 10 times across four engines. Facebook pages were cited 17 times. If your business has no presence on the platforms an engine trusts, part of every answer is written without you.
Finding 4: attributes come from somewhere, and you can author them
Every description the engines gave was traceable: "award-winning" (repeated four times for one Howick builder, citing Master Builders awards), "covers Howick, Botany, East Tamaki, Dannemora, Flat Bush, Pakuranga" (lifted from suburb lists on a builder's site), "transparent pricing", and notably, one renovation firm was recommended partly because it publishes an "Auckland renovation cost guide". The engines describe you using the raw material you and your customers put on the record. Write the record deliberately.
What this means if you're a builder (or any local business)
- Review velocity beats review bursts. Consistent recent reviews get quoted as the reason to choose you.
- Publish real cost content. The cost prompts are nearly empty; honest numbers are an uncontested claim on them.
- Name your suburbs in plain text. Engines repeat suburb lists verbatim when recommending for location-specific prompts.
- Get onto the trusted platforms. Builderscrack, NoCowboys, your industry register, an active Facebook page: they are part of the answer whether you participate or not.
- Make your credentials machine-readable. Licences, awards and guarantees only become attributes if they exist as clear text and structured data.
The bigger picture
This was one vertical in one city in one month, and we'll re-run it quarterly. But the mechanics generalise: the engines reward exactly the evidence a careful customer would look for, published where machines can read it. AI search doesn't have a secret algorithm to game. It has a reading list, and the work is getting on it.
We run this exact measurement for clients as part of our AEO service, your trade, your city, your competitors, tracked monthly with Citable.
