Somebody in your suburb just asked ChatGPT “who’s a good plumber near me” — and took the answer. Not a list of ten blue links to think about. One answer, with two or three businesses named. If yours wasn’t one of them, you didn’t lose the click. You never existed.
This is the quiet shift happening under local search right now. People still Google. But a growing share of “who should I call” questions now go to ChatGPT, Perplexity, and Google’s own AI Overviews — engines that don’t show a results page, they give a recommendation. ChatGPT alone serves hundreds of millions of users a week, and Ahrefs found that 28% of the pages ChatGPT cites most have zero organic visibility on Google. Read that again: AI engines reward different signals than classic SEO. Ranking well on Google no longer guarantees you exist in the other half of search.
The good news: you can check where you stand in about ten minutes, and most of the fixes are concrete, one-time, and unglamorous.
The 10-minute self-check
Open ChatGPT (and Perplexity, if you have five more minutes) and ask it the questions your customers actually ask. Not your business name — the category questions:
- “Who’s the best [your trade] in [your suburb/city]?”
- “I need a [dentist / plumber / Korean restaurant] in [area] — who do you recommend?”
- “Is [your business name] any good?”
Then look at three things:
- Are you named at all? If competitors come up and you don’t, you have a visibility problem, not a quality problem.
- Is what it says about you correct? Wrong hours, old address, a service you stopped offering — AI engines repeat whatever they last read about you.
- Where is it getting its information? Perplexity shows its sources directly. Usually it’s some mix of your website, Google reviews, directories, and local articles. Those sources are your levers.
Write down what you find. That’s your baseline — everything below is about moving it.
Why AI recommends one business and not another
AI engines don’t crawl your charm. When a model answers “who’s a good electrician in Henderson”, it’s drawing on what it can read, parse, and trust about the businesses in that area. In practice, the signals cluster into four groups:
1. Can AI crawlers even read your site? Every AI engine has a crawler — OpenAI’s OAI-SearchBot, Anthropic’s ClaudeBot, PerplexityBot. Your site’s robots.txt file (a small text file every website has) either lets them in or blocks them. We audit sites weekly and regularly find AI bots blocked by default — usually by a template or plugin the owner never touched. That’s invisibility you’re configuring yourself.
2. Structured data — the machine-readable version of your business. Under the visible page, well-built sites carry JSON-LD schema: a block of structured facts saying “this is a business, here’s its name, address, phone, opening hours, services, FAQ answers.” One Semrush test found proper schema lifted an LLM’s extraction accuracy on a page from 16% to 54%. Same content, three times more usable to the machine. Most small-business sites have none.
3. Consistency and citations. AI models cross-reference. If your name, address, and phone number (NAP, in the jargon) differ between your website, your Google Business Profile, and the directories you’re listed in, you look like two or three half-businesses instead of one solid one. Reviews matter here too — not just the star count, but review text the engines can quote: “they showed up same day” is a sentence a model can hand to the next customer.
4. Answer-shaped content. AI engines cite pages that answer questions directly. A page that says “Emergency call-outs: we answer 24/7 and reach most of West Auckland within 45 minutes” is quotable. A page that says “Welcome to our website! We pride ourselves on excellence” is not. Clear headings, plain factual sentences, real numbers, an FAQ — the same things that make a page useful to a human make it citable to a machine.
There’s also a newer convention worth knowing: llms.txt — a plain-text index of your site written specifically for AI models, sitting next to your robots.txt. Over 800,000 sites already ship one. It takes an hour to create and it’s the cheapest “please understand my business correctly” signal available.
What we’ve measured
This isn’t theory — we run a full audit-and-fix system on this exact problem (we’ve written up how the engine works and what it scores separately). Two recent runs, scored 0–100 on AI-readiness:
- A portfolio site: 45 → 61 after one fix pass — schema added, robots opened to AI bots, llms.txt created.
- A hospitality client at watda.co.nz: 50 → 66 — same categories of fixes, applied to a live business.
The pattern in both: nobody was doing anything wrong. They were doing nothing, because nobody had told them this layer existed. The first fixes are cheap and they move the number immediately; getting from “good” to “AI’s default recommendation” is slower and depends on reviews, citations, and content compounding over months. Anyone promising the second part in a week is selling something.
Do this this week
- Run the 10-minute self-check above. Baseline first.
- Check
yoursite.com/robots.txt— if you see AI bot names next to the word “Disallow”, that’s a five-minute fix for whoever manages your site. - Ask that same person one question: “do we have JSON-LD schema and an llms.txt?” The answer tells you a lot about the state of your site.
- Fix your NAP: same name, address, phone everywhere — site, Google Business Profile, Facebook, directories.
- Rewrite one page to be answer-shaped: your most-asked question, answered in the first two sentences, in plain English with real numbers.
None of this requires a new website. All of it requires someone actually looking.
Want the full picture instead of the self-check? We run a measured visibility audit — what Google and the AI engines see when someone looks for what you do, scored before and after. Ask us for one.