Is your business invisible in AI answers? How to check
Absence from AI answers makes no sound — no lost ranking, no traffic dip, no alert. Here’s how to find out where you stand, and why one ChatGPT session can’t tell you.
When your search rankings slip, you see it: traffic falls, a report goes red, somebody asks questions. When you slip out of AI answers, nothing happens at all. The buyers who would have called simply hear other names, and your dashboards register a quiet month. That silence is the problem this article solves — a straight way to check where you stand before the silence gets expensive.
The warning signs
- Leads mention AI less than your market does. If buyers in your category increasingly say “I asked ChatGPT” and none of yours do, someone else is absorbing those conversations.
- Your competitors show up in casual checks and you don’t. Not proof on its own — single answers vary — but a pattern across several tries is a signal worth measuring properly.
- Your online facts disagree with each other. Mismatched names, addresses, categories or dead links across listings are exactly what makes an engine skip a business. If you know your citations are messy, assume your visibility reflects it.
- You’ve never checked. The most common sign. Most local businesses have no idea what the engines say about their category — which also means the ones who measure first hold an advantage.
Why one query proves nothing
Before you open a chat window and draw conclusions, know this: the same question, asked twice, returns different names. The largest public test of AI recommendations found the identical prompt produced the identical brand list less than once in a hundred runs — we’ve written up the full evidence. A single flattering answer is not proof you’re visible, and a single absence is not proof you’re invisible. Only frequency across repeated runs means anything.
Write down the questions buyers actually ask
Not your brand name — buyers who know your name aren’t the prize. The discovery questions: “best [your service] in [your city]”, “who should I hire for…”, “[service] near me with good reviews”. Ten is enough to start.
Ask them repeatedly, across engines
Each question, several times, in fresh sessions, on more than one engine — ChatGPT, Claude, Perplexity, Gemini at minimum. Log every business named. Yes, this is tedious by hand; that’s the honest cost of a real answer.
Count frequencies, not appearances
The output that matters is a table: each business, and the share of answers that named it. Your position in that table — including a zero — is your actual AI visibility, stated honestly.
Read the citations while you’re there
Engines that show sources are handing you the map: the directories, review sites and articles they trusted. Every cited source where you’re absent or inconsistent is a concrete fix.
Or let the scan do all of it
The free scan runs ~18 buyer questions, repeatedly, across six engines — including Google’s AI Overviews and AI Mode — and reports your mention rate with confidence ranges, who was named instead, and the cited sources. A human reviews every report before it sends.
What to do with the result
If you’re visible: protect it. The work that put you there — consistent listings, steady reviews, quotable pages — decays quietly if ignored, and monthly measurement is how you notice drift before buyers do. If you’re invisible: the fix list is knowable and mostly free — start with the playbook. Either way, check whether your market is one we already measure publicly; if it is, your standing is a click away.
Maybe — one answer is one draw from a shuffling distribution. If repeated runs name you consistently, yes, you’re genuinely visible. That’s the part worth verifying.
Common, and fixable. It usually means the sources engines trust say too little about you, or disagree. The citations in your category’s answers tell you exactly where to start.
Monthly, with the same questions — frequent enough to catch drift, spaced enough that movement is real rather than noise. It’s the cadence our own leaderboard runs on.