Open ChatGPT and type the question your best customer would ask before they buy from you. Something like “best accounting software for a small construction business” or “most reliable HVAC company in Rochester”. Read the answer carefully.
If your business isn’t in it, that answer is being given to people right now, and you’re not in the room when it happens.
This is the part most owners don’t expect. In traditional Google search you could at least see your position and your traffic. In AI search, the assistant reads a handful of sources, writes a recommendation, and often the customer never clicks through to anyone. You can be losing a deal without a single line appearing in your analytics. A whole software category has grown up around measuring this, and comparisons of the best AI visibility platforms now list dozens of options — but before you spend anything, it’s worth understanding what you’re actually trying to find out.
Table of Contents
Start with the free check, not the subscription
You don’t need software to establish whether you have a problem. You need about twenty minutes.
Write down the five to ten questions a customer would genuinely ask an AI assistant before choosing a business like yours. Not your product name — the problem they have. “How do I find a good bookkeeper for an e-commerce store.” “Best CRM for a two-person sales team.” Then run each question through ChatGPT, Google’s AI Overviews, Perplexity and Gemini, and record two things: whether you’re named, and which websites the answer cites as sources.
That second column is the useful one. Those cited sources are the pages the AI trusted enough to build its answer from. If a competitor’s product page, a review site and a Reddit thread are the sources and none of them mention you, you now know precisely where the gap is.
Several vendors offer free one-off checks for Perplexity or Google AI Overviews presence, and a couple run permanently free tiers covering around ten prompts. Use those before you consider a paid plan.
Why you have to check more than one assistant
This is the mistake I see most often with small teams. Someone checks ChatGPT, finds the business mentioned, and concludes AI search is handled.
The engines mostly don’t cite the same pages. In our own analysis of roughly 600,000 prompts where at least two major AI engines answered the same question, only about one in ten cited pages appeared on more than one engine. ChatGPT was the extreme case — the overwhelming majority of pages it cites appear on no other engine at all.
The more encouraging finding: brand names travel much better than pages do. Around two-thirds of the time, if one engine names a brand in its answer, another one does too, and that rises further on “best X for Y” type questions. So the goal isn’t to win each engine separately with individual pages. It’s to become a business the category conversation recognises, which is a slower job but a more durable one.
What actually gets you cited
When we looked at what AI engines cite when someone asks them to recommend a tool in our own category, three patterns showed up that apply well beyond software.
Citations concentrate on a small group of sources. In our data the top 100 domains took roughly two-thirds of all citations, with thousands of other sites splitting the rest. This is not a long tail where everyone gets a slice. Being one of the trusted sources is most of the battle.
Community discussion matters more than owners expect. Reddit was the single most-cited domain in the set, ahead of every vendor’s own website. Not because Reddit is optimised, but because AI assistants treat practitioner conversation as evidence of what real users think. You can’t fake your way into that, but you can participate honestly in the places your customers already discuss your category.
Your own product pages do more work than your blog. By format, product and feature pages earned the largest share of citations — considerably more than comparison articles or how-to guides. If your service pages are thin, vague, or written as brochure copy, that’s the first fix. Say clearly what you do, who it’s for, what it costs, and where you operate.
When it’s worth paying for a tool
Manual checking works fine for a first look. It stops working when you need to know whether things are improving, because AI answers shift with wording, freshness and model updates, and your memory of last month’s answer isn’t evidence.
That’s what paid tools are actually selling: the same prompts, run on the same schedule, so you can see a trend instead of a snapshot. Most small businesses don’t need the multi-location and enterprise tiers, so ignore the top of the market entirely.
Three things to check before you pay for any of them:
- How many AI assistants does your plan actually cover? Not the enterprise tier — yours.
- Does it show you the source URLs behind each answer, or just tell you whether your name appeared?
- Does it end with something to do? A page to rewrite, a topic you’re missing, a site worth reaching out to. A tool that only produces charts will quietly become a subscription nobody opens.
The realistic timeline
Early movement can appear within a few weeks of fixing your core pages. Real, stable presence usually takes a few months, because it depends on how many trusted sources reference you — and that isn’t something you control directly. Anyone selling you a fixed timeline is guessing.
Start with the twenty-minute manual check this week. If you’re already named in most of those answers, you have less to worry about than you thought. If you’re not, at least you now know it, which puts you ahead of most businesses in your category.
