When someone asks ChatGPT which accounting firm is best for a small business in Halifax, they get an answer. Not a list of links. An answer. That response becomes the shortlist, and it happens before your website ever loads. Two years ago, this was a novelty. Now it's how a growing share of buyers start their research. The question worth asking today is simple: what is AI saying about your business?
Key points
- AI search moved from a novelty traffic source to a purchase-decision gatekeeper in about two years.
- There are two ways to lose: you can be invisible in AI answers, or you can be visible and described incorrectly.
- A New Zealand investment platform found ChatGPT was confidently stating wrong facts about its fees and fund lineup, while analytics showed only the traffic.
- Standard dashboards track clicks and rankings, not what buyers were told before they clicked.
- A simple audit, asking AI tools the questions your buyers actually ask, is a practical first step for any business.
There's a visibility problem and an accuracy problem, and they're separate. You can be missing from the answer entirely, which means the buyer never learns you exist. Or you can show up and be described incorrectly, which might be worse. Wrong fees, wrong services, a safeguard framed as a restriction: any of these can kill a deal before a conversation starts. A New Zealand investment platform ran into the second problem. When the company asked ChatGPT the questions its own customers were asking (about fees, fund lineup, how it compares to competitors), the answers came back fluent, confident, and often wrong. The fund count was off. Fees were mischaracterized. A consumer protection feature was described as a limitation. The platform's analytics showed the referral traffic, but nothing in those dashboards revealed the misinformation shaping buyers before they clicked. That's the gap. Your traffic report tells you who arrived. It doesn't tell you what they were told on the way.
SEO and AI search optimization aren't replacements for each other. Strong SEO makes it more likely that AI systems will find and use your content. It gives them something true to work with. But finding your content and citing it correctly are different things. For comparison queries, the kind that start with "best X for Y" or "X vs Y," AI systems consult many sources and trust a small subset. If you're not in that subset, you may never appear, no matter how well your pages rank in traditional results. And if you are in it, what gets pulled from your site needs to be structured and clear. Claims buried in loose paragraphs don't get extracted. Practitioners describe this as writing for two audiences: humans and the AI assistants answering on their behalf.
You don't need special software to start. You need the questions your buyers actually ask. Write down ten or fifteen of them. Things like "how much does [your service] cost in [your city]," "best [your category] for [your customer type]," and "[your business] vs [competitor]." Then ask those questions in ChatGPT, Perplexity, Google's AI Overviews, Claude, and Copilot. Record exactly what comes back. You're looking for two things: whether you appear at all, and whether what's said about you is accurate. Both matter. If you're absent, that's a visibility problem. If you're present but misquoted, that's an accuracy problem, and it needs fixing fast. This kind of audit is repeatable and worth doing on a schedule. AI answers change as models update and as new content gets published. What's true this month may shift next quarter.
If you work in legal, medical, dental, finance, or insurance, pay close attention here. These are the verticals where a confidently wrong AI answer does real damage. A prospective client researching a personal injury lawyer, a patient comparing dental implant costs, or a small business owner checking insurance options is making a high-stakes decision. If the AI gets your credentials, pricing, or service details wrong at that moment, you may never get the chance to correct it. The fix isn't complicated in principle. Make your key facts easy for AI systems to extract: clear service descriptions, structured pricing information where you can share it, straightforward answers to common questions on your own site. The work is in doing it consistently and checking that it's working.
Rankings, impressions, and click-through rates can't show you what an AI assistant said about your brand. That's a real limitation, and the industry is still building the tools to close it. Some agencies now track citation frequency, mention share, and visibility scores across multiple AI platforms, measured against buyer-intent prompts. That's a more useful picture than a ranking report alone. The point isn't to chase a single score. It's to know whether you're showing up, whether you're being described correctly, and whether that changes over time. If you can't answer those three questions today, you're flying blind in a channel that's already influencing purchase decisions.
Run the audit on yourself first. Ask the AI tools your buyers use the questions they'd ask. Document what you find, good or bad. If you're invisible, that's your first project. If you're visible but wrong, that's your first fix. Either way, you'll know more about your position in AI search than most of your competitors do. This discipline is roughly two years old. Most businesses haven't looked yet. That's an opening if you move now, and a growing problem if you wait.
AI search isn't a future trend. It's answering questions about your business right now, whether you've checked or not. The businesses that treat visibility and accuracy as things to measure and manage will be the ones getting called. The rest will wonder why the leads stopped coming. If you want help finding out what AI says about you and fixing it, book a growth call with Worqd.