Where AI answers actually come from
When someone asks an AI assistant for a business recommendation, the engine does some mix of three things: recalls what it learned in training, runs live web searches on the user's behalf, and reads trusted structured sources. You can see the second one happening in your own Search Console, conversational queries suffixed with the searcher's location are AI assistants searching for users. Each pathway has inputs you control.
The six sources that write your AI reputation
1. Your Google Business Profile. The backbone of local recommendations. Verified, complete, active profiles get surfaced; unverified ones are invisible. Categories, service areas, hours, photos, and posts all feed the answer.
2. Your reviews. Count, rating, and recency, and the actual text. Engines quote review evidence as the reason for a recommendation, and they notice when a business stopped collecting reviews a year ago.
3. Industry platforms and registers. In our measurement of NZ trades, review platforms and the government licensing register were cited more often than any individual business website. Every industry has its equivalents. If the engines trust a platform and it doesn't know you, you're absent from part of every answer.
4. Your website's plain text. Engines lift suburb lists, prices, timeframes, and credentials verbatim, when they exist as readable text. Text hidden in images, sliders, or scripts that only render on click is invisible to most AI crawlers, which read raw HTML.
5. Structured data. LocalBusiness and FAQ schema tell machines exactly which text answers which question and which business it belongs to. It is the difference between hoping the engine parses your page correctly and telling it.
6. Consistency across all of the above. Same name, same phone, same service areas everywhere. Conflicting details make an engine less confident, and less likely to name you.
The order of operations
- Verify and complete your Google Business Profile, this week
- Start a weekly review habit: one text with a direct link after every job
- Rewrite your key pages to answer the questions customers actually ask, with real numbers
- Add FAQ and LocalBusiness schema so machines can parse it
- Claim and complete the platforms your industry's engines already cite
- Measure monthly: ask the engines what your customers would ask, and track who gets named
What doesn't work
Keyword stuffing reads worse to a language model than to a human. Fake reviews are a detectable pattern and a legal risk. "AI-optimised" tricks that hide text or manipulate prompts get filtered. The uncomfortable truth is that AI engines reward what a diligent customer would look for: evidence, published where machines can read it. There is no shortcut, but there is a head start, because most of your competitors haven't started.
We measure exactly which engines mention your business, and run the sequence above as a service: AEO for trades and professional services.
