Here's a scenario playing out in trades and professional firms across the country right now. A plumbing company ranks on page one for its main service keywords. Its Google Business Profile is solid. Traffic looks healthy in the monthly report. Then the owner asks ChatGPT a question a real customer might type: 'Who's a good plumber in Halifax?' The assistant names three competitors. The company isn't mentioned at all. That gap between traditional search performance and AI-assisted discovery is one of the quiet problems of 2026. Customers haven't stopped using Google, but a growing share of them now start with ChatGPT, Perplexity, Gemini, or Google's AI Overviews, especially when they don't know which business to call. If those systems don't recognize you, you're absent from a conversation you never knew was happening. The good news: you can measure this yourself in about an hour, and the results tell you something your keyword dashboard can't.
Key points
- A business can hold page-one Google rankings yet be invisible when buyers ask AI assistants for local recommendations.
- The most valuable test questions are category recommendations, where the buyer asks for 'best [service] in [city]' without naming any brand.
- AI platforms, including Google, OpenAI, and Perplexity, publish guidance confirming their systems draw on web content and cite sources, which means businesses can influence but never control AI answers.
- AI SEO and answer engine optimization build on conventional local SEO foundations like content, structured data, reviews, and citations; they don't replace them.
- Repeat the test weekly under identical conditions and treat the results as directional signals, not guarantees.
Traditional SEO metrics measure how you perform when someone searches. AI visibility measures something else entirely: whether an assistant can identify your business, describe what you do, and recommend you when a buyer asks an open-ended question. These two outcomes don't always move together, and the mismatches cut both ways. Some businesses rank well on Google yet never appear when someone asks ChatGPT or Perplexity for a recommendation. Others get recommended by an AI without their website ever being cited, which means the assistant is pulling its impression from reviews, directories, or third-party content rather than anything the business published. The reason is mechanical, not mysterious. AI assistants synthesize answers from web content, and the platforms themselves say so. OpenAI explains that ChatGPT Search draws on web sources with citations you can inspect. Perplexity describes its approach as multi-source synthesis with references. Google's guidance on AI search experiences emphasizes original, useful content as the raw material. None of these platforms accept submissions or paid placement. They discover what's already published, which means the quality, clarity, and consistency of your online evidence is what shapes whether you show up.
A practical testing method has been circulating among marketers this year, and it works just as well for a roofing contractor as it does for a software company. The idea is simple: write 20 questions a real buyer might ask, run them through ChatGPT, Perplexity, and Google's AI features, and record what comes back. The questions fall into three groups, and each group tells you something different. First, brand recognition. Ask the assistant directly: 'What does [your business name] do?' or 'Is [your business] a good choice for [service]?' This tells you whether AI systems have enough accurate information about you to describe your services and who you serve. Wrong or vague answers here are a red flag that your business facts are thin or inconsistent across the web. Second, and this is the money set, category recommendations. These are questions where the buyer doesn't name anyone: 'best HVAC companies in [city],' 'who should I call for a burst pipe in [town],' 'experienced personal injury lawyer near me.' For most trades and professional firms, this is exactly how new customers arrive. They don't know you yet. If your competitors appear in these answers and you don't, that's the visibility gap that matters most. Third, informational questions from the research phase: 'how long does a roof replacement take,' 'what does an accountant cost for a small business,' 'signs your furnace needs replacing.' These show whether AI systems treat you as a source of expertise, and whether your content is being read and reused when people are still deciding.
A one-off test tells you very little, because AI answers shift between sessions. To get signal instead of noise, keep the conditions consistent. Start a fresh conversation for every question. Reuse the exact same wording each time. Keep your language and location settings unchanged. Log the date, the full answer, which businesses were named, and which URLs were cited. Then repeat the whole set weekly. After three or four weeks, patterns emerge that a single session would never show. You'll see how assistants describe your services in their own words, which competitors consistently appear alongside you, and which third-party sources keep getting cited. That last one is often the most useful finding, because it tells you where the AI is actually forming its opinion of your category. If answers keep leaning on a local directory or a handful of review sites, those are the places your business information needs to be accurate and complete. One caution worth stating plainly: this testing is directional. You're sampling a system that changes, sometimes week to week. A missing mention today isn't a verdict, and a glowing recommendation today isn't a lock. What the test gives you is a trend line, and trends are what you can act on.
The instinct, once a business sees it's absent from AI answers, is to look for a trick. There isn't one, and anyone selling guaranteed AI citations is selling something the platforms don't offer. What the evidence supports is less glamorous and more durable. AI systems discover and reuse the same foundation that good local SEO has always required: useful original content, accurate and consistent business information across the web, structured data that machines can parse, a healthy review profile, real citations in relevant directories, and a technically accessible website that AI crawlers can actually reach. A recent UK case study illustrates the layered approach. Prestigious Coatings, a uPVC and kitchen spraying specialist in Lancashire, worked with its agency through brand identity, a website build, then technical SEO, locally targeted service pages, structured data, Google Business Profile development, and authority building, with AI and answer engine optimization layered on top. The reported results included page-one Google rankings across Lancashire, year-on-year organic growth, stronger Maps visibility, and a growing presence in AI-powered search. The sequence matters. The AI visibility was built on top of the fundamentals, not instead of them. That's the honest framing for this whole discipline. Generative engine optimization and answer engine optimization are extensions of conventional SEO, and treating them as separate services, or as replacements, is a mistake. You're improving the evidence layer that both Google and the AI assistants draw from.
If you run a business where customers find you by searching a category rather than a brand name, HVAC, plumbing, electrical, roofing, law, accounting, clinics, real estate, this applies to you directly. Your future customers are asking assistants 'who's good at this near me' before they've heard of anyone, and the shortlist those answers produce is becoming a real competitive boundary. If you'd rather not spend your weeks running prompt tests and logging answers, the test still serves a purpose: it's a diagnostic. Run it once to see where you stand, or hand it to someone who manages your marketing. The output, a documented record of how AI systems currently describe your business and your competitors, is a far more useful starting point than another traffic report. The market is moving quickly on this. Established agencies with decades in traditional SEO are now packaging AI visibility services explicitly, which tells you demand among small businesses is real. That also means the businesses that establish accurate, well-structured online evidence early will be harder to displace later.
Your rankings report measures the search box. It doesn't measure the conversation happening in ChatGPT, Perplexity, or Google's AI Overviews, and that conversation is increasingly where first contact happens. Twenty questions, run weekly under consistent conditions, will show you whether AI systems know your business, how they describe it, and who they recommend instead of you. The fixes, when gaps appear, are the same fundamentals that have always powered local visibility: clear content, accurate business facts, real reviews and citations, and a technically sound website. Influence, not control. Directional signals, not guarantees. But for an owner-operator who wants to be on the shortlist when it matters, that's a measurement worth an hour of your week.