You've probably heard some version of this advice already: get your schema markup right, feed ChatGPT the right prompts, make sure your FAQ page answers questions the way Google's AI expects. Treat AI visibility like SEO with a new interface, and the citations will follow.

That advice is mostly wrong, and it's costing businesses the exact visibility they're trying to buy.

AI visibility isn't a ranking problem. It's a worthiness problem.

Search engines used to rank pages by relevance and authority signals you could reverse-engineer. AI assistants like ChatGPT, Claude, and Google's AI Overviews do something different: they read what's actually written, judge whether it sounds like a real person with real experience, and decide whether it's worth repeating to someone who asked a question.

You can't schema-markup your way into that. You can only earn it.

Why the SEO playbook doesn't transfer

Traditional SEO rewarded structure. Keywords in the right places, backlinks, technical hygiene, a domain that had been around long enough to look trustworthy. None of that is irrelevant now, but it's no longer the main lever.

An analysis by Semrush of more than 300,000 AI-generated answers found something the SEO industry didn't expect: individual voices were being cited more often than the company pages sitting right next to them. A named person with a specific point of view was outperforming the polished corporate page it was published alongside.

That finding breaks the assumption most businesses are working from. If AI models preferred polished, authoritative-sounding company content, the winning move would be to keep doing what worked for Google, just harder. But they don't. They prefer specificity: a real opinion, evidenced with something that actually happened, from someone who clearly knows what they're talking about.

This is why so many businesses feel invisible in AI answers despite having a good, well-built website. The site is well built. It's just not saying anything an AI model would trust enough to repeat.

What AI visibility actually means

Here's the plainer definition: AI visibility is whether ChatGPT, Claude, or Google's AI names your business when someone asks a question your business should answer. Not "does it rank." Does it get recommended.

That's a different bar. Ranking rewards technical compliance, while recommendation rewards evidence that you're worth mentioning: a specific answer to a specific question, written by someone who clearly knows the terrain, not a generic overview that could belong to any consultancy in your sector.

Think about how you use these tools yourself. When you ask ChatGPT who's good at something, you don't want the SEO-friendly answer. You want the one that sounds like it came from someone who's actually done the work. AI models are trained on the same instinct, built to spot the generic answer and prefer the specific one, because that's what makes them useful to the person asking.

A worked example: the pivot that proved the point

I didn't arrive at this from studying AI search. I arrived at it backwards, through a product built for a different reason entirely.

Ghostart started to fix generic professional writing: the flat, forgettable LinkedIn posts and website copy that all sound like they came from the same template. We built a scoring system, the Beige-ometer, to measure how "alive" a piece of writing was: whether it had a real voice, a specific point of view, an actual person behind it, rather than safe corporate phrasing.

When AI search started reshaping how businesses get found, we expected to have to rebuild the product from scratch. We didn't. The exact writing that scored well on the Beige-ometer turned out to be the writing AI models were already choosing to cite. Specific, human, opinionated, evidenced.

We hadn't been building for AI visibility. We'd been building for authenticity, and the market caught up to the same standard.

That's the part most GEO advice misses. It treats AI visibility technical discipline to master. It's closer to an old editorial standard finally being enforced by a machine that can tell the difference.

What to actually do about it

If AI visibility rewards specificity over polish, the practical shift looks like this:

• Name real things. Real clients (with permission), real numbers, real dates, real decisions you made and why. Vague claims about "years of experience" give an AI model nothing to cite.
• Write in your own voice, not your brand's. A named person's opinion, backed by evidence, consistently outperforms a company's official position. If your content could have been written by any competitor, it's not visible; it's interchangeable.
• Answer the actual question, not the marketing version of it. If your customers are asking "who's good at X," your content needs to answer that directly, not circle it with brand messaging.
• Publish regularly, not once. AI citations decay. A single strong page doesn't stay visible forever if nothing else is being published to reinforce it.
• Get your team's expertise on the page, not just your own. A business with three specific voices publishing genuine expertise looks more credible to an AI model than one voice repeating the same points.

None of this is about gaming a system. It's about being specific enough, and honest enough, that a machine trained to spot the fake stuff has no reason to skip past you.

The actual takeaway

Stop asking how to rank in AI search. Start asking whether what you've published is specific enough that a stranger reading it would trust you over the ten other businesses doing the same thing.

That's the question AI models are already answering on your customers' behalf.

Pick one page or post this week: the one that's meant to represent your expertise best. Read it as if you'd never met you. Does it name anything real? Does it sound like a person, or like a category?

If you can't tell your business apart from a competitor's by reading your own homepage, neither can the AI deciding who to recommend.