AI visibility. Ask ten people what it actually means and you'll get ten different answers, most of them borrowed straight from an SEO playbook.

AI visibility isn't about how often a system says your name. It's about whether that system trusts you enough to recommend you, and for what, specifically.

Those are two different problems. Chasing mentions and citation counts solves the first one. It does nothing for the second, and the second is the one that actually gets you recommended.

Get mentioned in ChatGPT answers. Track your citation count. Add the right schema markup and hope Google's AI Overview picks you up. That's the assumption baked into almost every guide on this topic, and it's the wrong starting point.

The assumption everyone makes

The common story goes like this: AI search is the new SEO, so the fix is the new version of old SEO tactics. Find the keyword equivalent for prompts. Stuff your website with the phrases people type into ChatGPT. Chase mentions the way you once chased backlinks.

I understand why this framing took hold. SEO gave marketers two decades of muscle memory, and it's tempting to point that same muscle at a new target. But AI answer engines aren't ranking pages against a query. They're synthesising an answer from whatever they judge to be credible, and then deciding who deserves the recommendation inside that answer.

That's a trust decision, not a ranking decision. And you can't game trust the way you can game a ranking algorithm, because the systems are specifically built to see through generic, keyword-stuffed content. Feed them thin, templated pages and they'll route around you to find the person who actually said something specific.

What the data actually shows

Semrush analysed over 300,000 AI-generated answers and found something that should worry anyone treating this as a checklist exercise: individual voices were outciting company pages. Not occasionally. Consistently. According to that analysis, a named person with a clear point of view was more likely to get cited than the polished, on-brand corporate page sitting on the same website.

That finding lines up with something I'd already built a product around, for a completely different reason. Ghostart started to fix human writing: to catch the flat, generic sentences that make a person's LinkedIn post sound like nobody wrote it. We called the scoring system the Beige-ometer, and it measured voice, emotion, rhythm and story. None of that had anything to do with search.

Then AI search arrived, and Semrush's evidence showed that the exact thing we'd been pushing people to do — write like a specific human with an actual opinion, not a brand template — was the literal mechanism these systems use to decide who's worth recommending. We didn't have to change what we believed. The market caught up to it.

Found for what?

The question worth asking

Here's the reframe that actually matters: stop asking "how do I get more AI visibility" and start asking "recommended for what?"

A consultancy with twelve generic service pages and a founder with zero public opinion might technically be "findable." Search engines index it, so it exists in the index. But if someone asks ChatGPT or Google's AI "who's good at pension transfer advice for small business owners," that consultancy has given the system nothing specific to recommend it for. There's no sentence anywhere that only they could have written.

Compare that to a sole adviser who's published three detailed, opinionated posts about the exact edge cases their clients keep hitting, the ones with real numbers, named scenarios, an actual argument. That person has given an AI system something to cite: a specific answer to a specific question, attached to a specific name.

This is the distinction most GEO advice skips over. It treats "appearing" and "being worth recommending" as the same goal. They're not. You can be technically visible and still be invisible in the way that matters, because visibility without a clear answer to "found for what" is just noise the system has no reason to surface.

What this means for your business

If you run an expertise-led business — a consultancy, an agency, an advisory practice — this changes where you should be spending effort. Less time on technical tweaks nobody can verify are working. More time answering the actual questions your clients ask, in your own words, with your own examples.

A few practical starting points:

• List the five questions your clients actually ask you before they buy. Not the ones your website already answers in generic language. The real ones, the ones that come up on discovery calls.
• Write the honest answer to each one, in your own voice, with a specific example or number. If you don't have a real statistic, don't invent one. A precise opinion beats a vague statistic every time.
• Put a real name on it. Founder-written or team-member-written content, published under an actual person, outperformed unattributed brand copy in the Semrush analysis above. Use that.
• Check what AI tools currently say when someone asks your exact niche question. If your name doesn't come up, that's a content gap, not a technical problem. A specificity problem.
• Publish on a schedule you can actually keep. AI systems reward fresh, consistent evidence of expertise, not a one-off pillar page from 2022.

None of this requires new software or a bigger budget. It requires the discipline to write down what you actually know, instead of what sounds appropriately professional. And none of it guarantees a citation — but it gives a system something real to cite you for, which is the part most businesses are currently skipping.

Start with one question

Think back to the last time a client asked you something on a call and you gave them a sharp, specific answer on the spot, then never wrote it down.

Write it up this week — your real answer, not the sanitised version. That's a better starting point for AI visibility than any schema audit, because it gives the system something true to recommend you for.