SEO didn’t die. The answer moved.

In 2023, a homeowner who needed a roofer typed "roofer near me," looked at a map with three pins and a list of ten websites, opened four of them, and called one. The whole game for a local business was being one of those four tabs.

That still happens. But more and more, the same homeowner asks the question the way they would ask a neighbor, "who is the best roofer in Andover for a full replacement," and gets an answer instead of a list. A paragraph. Three names. A sentence about why each one. Sometimes a link, often no click at all.

What actually changed

Four things landed in roughly two years, and together they moved where the decision happens.

• Google started answering. AI Overviews put a written answer above the results for a growing share of questions, and AI Mode turned search into a conversation you can follow up on.
• ChatGPT learned to search. It went from a writing tool to something people ask for recommendations, and it now pulls live results from the web to answer.
• Perplexity and others made "ask, don’t search" normal. Answers with citations, not pages of links.
• Phones made it default. Voice assistants and search apps answer out loud or in a card, which leaves room for two or three names, not ten.

None of these killed Google search or SEO. They added a layer on top of it. The businesses that get named in that layer are mostly the ones that were already doing good SEO, plus a few specific things most sites never bothered with.

Ranking gets you considered. Being in the answer gets you hired.

Why this is harder for small businesses than it looks

A list of ten results is forgiving. You can be number seven and still get the call from someone who scrolls. An answer with three names is not forgiving. There is no number seven.

And the answer is not built from the same signals as the old list. An AI model is trying to answer a question confidently. It favors businesses it can describe clearly: what they do, exactly where, and what proof exists that they are good at it. A site that says "serving the greater area" and lists six services in one paragraph gives it nothing to be confident about, so it picks someone else.

What the models look for

Different tools weigh things differently, and none of them publish the recipe. But in the audits we run, the businesses that get named tend to have the same handful of things in common.

• A clear identity. The business name, what it does and where it is based, stated plainly and consistently across the site and the web.
• A page for the question. If someone asks about a service in a town, there is a page about that service in that town.
• Proof the model can read. Reviews with real detail, case studies or before-and-after pages, named by what and where.
• Structure. Schema markup that tells machines, in their own format, what the business is, what it offers and which towns it serves.
• Agreement across sources. The same name, phone and service area on Google, Yelp, directories and social pages. Models trust what they see repeated.

What this means for your business

If your site already ranks well, you are ahead. The work is adding the layer: making each page a direct answer, marking up what you already have, and building the proof pages and town pages the models look for.

If your site does not rank, the good news is that the fix is the same work. Structure that helps an AI model name you is the same structure that helps Google rank you. You do not have to choose.

The first step is not buying anything. It is finding out where you actually stand when someone asks.