Somewhere in the last eighteen months, the search results page quietly stopped being a list of links and became an answer.

BrightEdge found that AI Overview presence across its tracked queries increased by 58% in one year, rising from approximately 30% to 48%. The precise prevalence varies significantly by dataset, industry, geography and search intent — a separate Semrush analysis of more than ten million keywords found AI Overviews stabilising around 16% in 2025 — but the direction of change is clear: more questions are being answered before a user reaches a conventional organic result.

For the person searching, this is a convenience. For the business whose entire visibility strategy depends on being clicked, it is something closer to a question most have not yet been asked to answer.

The instinct, when a shift like this appears, is to treat it as a tactics problem. Add some schema markup. Restructure a few headings. Wait for the algorithm to settle and see what happens. That instinct is understandable — and it is also the wrong first move, because it treats a decision problem as if it were a technical one.

Here is the decision underneath it: is your positioning clear enough that something with no loyalty to you — no relationship, no memory of your brand, no reason to be generous — could read your website once and describe accurately, in a sentence, what you do and who it's for?

Most businesses have never had to answer that question, because until recently, a human being was doing the interpreting. A visitor would land on a slightly confusing homepage, scroll around, infer meaning from context, forgive some ambiguity because a brand they were curious about was in front of them. AI systems extract — they don't infer. Ambiguity a curious human would forgive becomes information a machine simply cannot use.

The quality of decisions cannot consistently exceed the quality of the information, questions and assumptions from which they are made.

Most "AI visibility" fixes fail for the same reason most marketing fixes fail — they are answers to a question nobody actually asked first. Before you touch a heading tag, the real question is simpler and harder: if a stranger read only your homepage's opening paragraph, could they explain what you do, in one sentence, within ten seconds?

If you're not sure, that's the test worth running before any other.