AI visibility is whether an AI system trusts you enough to recommend you by name when someone asks a question you could genuinely answer well. Most founder-led businesses are focusing on the wrong thing here, and it's costing them the exact clients who'd become their best customers.

The visibility trap most businesses fall into

When GEO first started getting attention, the instinct was predictable: treat it like SEO 2.0. Find the new lever, work it. Add FAQ schema. Publish another "top 10 questions about X" page. Stuff in the keywords your competitors rank for and wait for AI Overviews to notice.

I understand the instinct, since it's the muscle memory of two decades of SEO, and it's not entirely wrong. But it treats AI visibility when it's actually a trust problem.

A consultancy owner I'd describe as typical of this pattern spent months building out generic explainer content. Technically correct, keyword-aware, structurally sound. Ask ChatGPT who's good at their specific niche, though, and they still don't come up. The content is findable. It isn't recommendable.

What AI visibility actually measures

AI systems like ChatGPT, Perplexity and Google's AI don't rank pages the way a search engine does. They synthesise an answer from sources they judge specific and credible enough to cite. That's a different bar than "does this page contain the right keywords."

An analysis by Semrush of over 300,000 AI-generated answers found something that should worry anyone still writing corporate-page content: individual, named voices outcited generic company pages, and not by a small margin. The systems trust a real person saying something specific more than a brand saying something safe.

That makes sense once you think it through. A generic page hedges, covering every angle so nobody's offended and nothing's wrong. An AI model trying to answer a specific question has no use for a hedge, because it needs a source that actually commits to an answer.

Being findable and being worth recommending are two different achievements, and only one of them shows up in an AI answer.

Why "worth recommending" beats "found everywhere"

Here's the question I'd ask instead of "how do I get more AI visibility": found for what, exactly?

I built the scoring system behind Ghostart, the Beige-ometer, years before GEO existed as a category. Its whole job was to catch writing that sounded like nobody in particular, the kind of professional content that reads fine and says nothing. That was an aesthetic argument then. It's a business argument now.

The same flat, hedge-everything writing that made a LinkedIn post forgettable is now the exact writing an AI model has no reason to cite. The problem didn't change, but the stakes did: generic writing used to just underperform, and now it's actively invisible to the systems your customers are asking instead of Google.

That's the reframe: AI visibility isn't a new discipline bolted onto content marketing. It's the old discipline of sounding like a real person with a real point of view, except now there's a machine checking whether you actually did it.

How to check where you actually stand

Skip the dashboards for a minute and do this instead:

• Ask the actual question. Type the exact question your best client had before they hired you into ChatGPT or Perplexity. Not a keyword. The real sentence they'd type.
• See who gets named. If it's not you, look at who is. Read what they published. It's rarely the most heavily worked-over page — it's usually the most specific one.
• Check your own content against that bar. Does anything on your site answer that exact question in a voice a competitor couldn't have written? Or does it read like it could belong to any firm in your sector?

Most businesses fail this test not because they lack expertise, but because none of that expertise made it into anything publishable and specific.

What to do with what you find

If the check above came back thin, the fix isn't more content. It's more specific content, published under a real name.

• Answer the exact questions clients ask you privately. Not the polished FAQ version. The messy real version, with your actual reasoning attached.
• Publish under individuals, not just the company page. Given what the Semrush data shows about named voices, a founder or team member's byline on a specific answer carries more weight than the same words on a generic services page.
• Use real numbers and real examples wherever you can. Vague claims give an AI model nothing to cite confidently. A specific figure, a named example, a stated opinion — all of these give it something to work with.
• Recheck every few months. AI citations aren't permanent. If you go quiet, a competitor's fresher, more specific answer replaces yours.

None of this requires a bigger content team or a new tool budget. It requires deciding that the safe, generic version of your expertise isn't serving you anymore, and it probably never was.

Open ChatGPT right now and ask it the question your last good client asked before they found you. Whatever answer it gives, you'll know exactly what to fix first.