Local SEO versus local AI search

What does local SEO actually rank?

Local SEO is a competition for a list. A buyer searches, Google returns a map with three businesses above the fold and more behind a click, and each one holds a position you can measure on any given day from any given point on the map. Move the measurement point a few kilometres and the list reorders, which is why local rank tracking is done on a grid rather than as a single number.

Google names the three inputs itself, and they have been stable for years. Relevance is how well a profile matches what was searched, distance is how far the business is from the searcher, and prominence is how well known it is. In practice that means a complete and active Google Business Profile, reviews in both volume and recency, business details that agree across the directories that mention you, and a page on your own site that plainly matches the search.

What does an AI assistant do instead?

An assistant does not return a list. It runs its own searches, often several of them from one question, reads a handful of pages, and writes a paragraph that names two or three businesses with a reason attached to each. There is no fourth place. You are either in the sentence or you are not, and the buyer usually stops reading there.

Because the assistant writes rather than ranks, the sources it read are the whole story. For local questions those sources are predictable. They are review platforms, the directories that cover the trade, local press and community threads, and the businesses own websites where those sites answer the question directly. A business with a thin site and forty good reviews is often named ahead of a business with a beautiful site and four.

The engines also differ in how much they know about where the buyer is, which is worth knowing before you interpret a result.

Google AI Overview

Sits on top of a real Google search, so it is genuinely local. The same question asked from two suburbs can produce two different sets of names.

ChatGPT and Perplexity

Accept a user location and use it when they search. City-level precision is real here, street-level is not.

Gemini

Has no location parameter in the API. Location can only reach it as text in the question, which is weaker, and anyone claiming city precision on Gemini is describing a rewritten prompt rather than a local search.

Why does a national average hide a local result?

Suppose you trade in eight cities and an assistant names you in two of them. Averaged, that is a visibility score of twenty five per cent, which looks like a brand that is mediocre everywhere. It is not. It is a brand that is winning outright in two markets and completely absent in six, and those two situations need opposite work. Defend the two and diagnose the six.

The average also hides movement. If you gain one city and lose another in the same month the national figure does not move at all, so the one report a business actually reads says nothing happened during the month that something did.

This is the reason to track by city rather than by country wherever the business is local. A country-level number is the right unit for a software company selling into Australia. It is the wrong unit for a business that serves three suburbs.

What carries over from local SEO?

Most of the groundwork does, which is the good news for anyone who has already done it.

Consistent business details

Name, address and phone number agreeing everywhere they appear. A model reading three versions of your address has three reasons to name a competitor instead.

Reviews, with recency

Volume still matters, but a page of reviews from two years ago reads as a business that may not be trading. Assistants quote recent sentiment far more readily than old sentiment.

Directory and local press coverage

The pages that already list businesses like yours are the pages an assistant reads. Being on them is the single most transferable piece of local SEO work.

Plain answers on your own site

A page that answers the question in the words a buyer used, in the first paragraph, is quotable. A page that opens with a paragraph about your journey is not.

What does not carry over?

Three habits from local SEO are either useless here or actively harmful.

Keyword position tracking has no equivalent, because there is no position. The nearest honest metric is how often you are named across a set of buyer questions, plus where in the answer you appear when you are named.

Proximity gaming does not transfer either. Registering an address you do not trade from, or stuffing a service and a suburb into the business name, is a Google Business Profile problem that can cost you the profile, and it buys nothing in an answer that is assembled from what other sites say about you.

Finally, the page built for the phrase rather than the question is dead weight. A thin page targeting the phrase "emergency plumber Parramatta" gives an assistant nothing to quote. A page that answers what an emergency call-out actually costs, how fast someone arrives, and what happens after hours, gives it three things.

How do you measure local AI visibility?

Start from the questions rather than the keywords. Write down what a buyer actually asks an assistant, in full sentences, including the ones that do not mention you at all, such as who is the best provider near me, who is open late, and what does this cost in this city.

Then run each question from the cities you trade in, on a schedule, and record two things. Whether you were named, and which sources the answer used. The second one is the actionable half. A month of answers tells you which review platform or directory is feeding your market, and that is where the next quarter of work goes.

Scout LLM does this with a city as a real location signal rather than a reworded question, keeps a separate daily series per city, and ranks the cited sources per market. The method matters more than the tool. If you do it by hand, do it on the same day of the week and keep the answers, because the trend is the finding.