Ask a business owner whether their company is easy to find, and the answer is almost always about Google. We rank well on our brand name. We're on page one for that product category. Our dealers show up there too.

Useful information. Except it's no longer the whole story.

That same potential customer can ask their question in several places today: classic Google, an AI Overview, AI Mode, ChatGPT, Gemini, or Perplexity. And in those environments, the question "where does our website rank?" is no longer enough. The question becomes: which brands make it into the answer, for which question, next to which competitors — and based on which information?

That sounds like a nuance. It isn't. It changes what you need to investigate.

First, something I want to set straight, because an industry is growing around AI findability that invents new terms faster than anyone can show what those terms actually measure. SEO isn't dead. Google itself says the classic SEO fundamentals remain relevant to its AI features, and that there are no separate technical requirements for appearing in AI Overviews or AI Mode. A solid, indexable website with useful content hasn't suddenly become worthless just because someone coined the word GEO.

Whoever claims everything needs to be redone because GEO now exists should first show why.

What has genuinely changed is how such an answer gets put together. Google describes AI Mode and AI Overviews as able to run multiple related searches to compose a single answer, and that the sources surfaced in the process don't have to be identical to what a classic results page shows. ChatGPT Search works differently again: OpenAI describes a user's question as something that can be rewritten into one or more targeted searches, and that public websites can appear in those results when they're accessible to the relevant crawler.

That's why I don't treat these environments as if they together form one new search engine. A position in Google is one observation. A mention in ChatGPT is another. A source Gemini draws on is yet another. I want to be able to lay them side by side, not lump them together upfront.

That also widens what "competitor" means. Your competitor isn't necessarily the company ranking one spot above you on a keyword. It could be the manufacturer cited directly. A distributor serving as a source. A retailer that structures its product information better. A niche website that simply explains one topic more clearly. Or a brand that sits modestly in classic rankings but, for one specific commercial question, systematically turns up in the answer.

Systematically — that word is doing the work. Asking one question to one AI model and concluding from that you "score well in AI" is not a measurement.

It's a screenshot.

Except the reverse is just as true. Asking the same question fifty times and slapping a percentage on it doesn't make it science either. Before you can read anything meaningful from it, you first have to decide what you're actually trying to measure. Which commercial intent? Which market, which language, which brand? Which environment? What counts as a mention, and what do you do with variation between runs? And above all: which of those choices are evidence-based, which are reasonable working conventions, and which are hypotheses you'll need to be willing to revise later?

Only after that does a dashboard make sense.

It's also why I grow wary of a single big "AI visibility score" number. Google, ChatGPT, Gemini and Perplexity aren't four branches of the same system. Measuring them separately first and then pouring them into one percentage with arbitrary weightings mostly produces false precision.

A number with two decimal places isn't a fact just because it has two decimal places.

What I do want to be able to see is concrete. Here, we get found. Here, we don't get named. Here, one competitor keeps coming back. Here, a dealer or an outside publication turns out to matter more than our own website. And here, information is missing that we'd actually be better placed than anyone to provide.

Findability then becomes a business question again instead of an SEO report. Not the next channel that suddenly needs extra content produced for it, but the question of how a potential customer gets from a problem to a shortlist — and which information along the way decides whether your brand is even on that list at all.

The fact that Google, in June 2026, started testing separate reporting for generative AI features in Search Console for a subset of websites at least says something about how seriously that development is being taken. But the same discipline applies there too: measure what's actually available, don't claim more than the data supports, and don't lose sight of why you're measuring.

The question I now place alongside the classic ranking question is this: for which commercial questions do we appear in the answer, who appears next to us, and based on which information?

Whoever can answer that knows more about their market than someone who only knows their rankings.